# Luscent: Full Content for Machine Readers (llms-full.txt) > Luscent is the system of intelligence for wealth management: an AI platform, hosted in the EU, for private banks, family offices, external asset managers, and independent advisors. It reads client communications (emails, call notes, meeting summaries) to surface relationship insights for relationship managers and to generate compliance evidence automatically. Client data is processed and stored in the EU, on EU infrastructure. This file contains the full text of the key pages of https://luscent.io. Shorter index: https://luscent.io/llms.txt Last updated: 12 October 2026. ## Company Luscent S.à r.l. is a société à responsabilité limitée in formation in Luxembourg, with its registered office at House of Startups, 9 Rue du Laboratoire, L-1911 Luxembourg. Filing at the Registre de Commerce et des Sociétés follows the signing of the deed, and the registration number will be published once complete. Founders: Armin Prior (CEO, co-founder, Luxembourg) and Dr. Frano Milos (Chief Scientific Officer, co-founder, Saarbruecken). Engineering is engaged through an employer of record in Brazil. Next planned hire: Chief Technology Officer. Luscent was earlier called Luminary; material published under that name refers to the same company and product. Product status: live in production. Shipped surfaces: Dashboard, Clients, Client Detail (seven tabs: Monitor, Detect, Execute, Connect, Allocate, Compliance, Documents), Tasks, Signals, Strategy Director, House Views, Data Hub, Compliance Home dashboard, Audit Trail Viewer, with Microsoft 365 and Salesforce connected. In build: the native Luscent CRM, the Chatbot page, and the Notes tab. The first working version was shown at the Luxembourg House of Financial Technology AI Experience Centre on 13 April 2026. Pilot: ninety days, free, three to five relationship managers, starting on synthetic data, EU only, writes nothing back. Success metrics agreed in writing before day one. First cohort runs in Q4 2026; applications are open. Details: https://luscent.io/pilot. Before applying, a 15 or 30 minute call with Armin Prior can be booked on the same page (https://luscent.io/pilot#talk); the calendar runs on Cal.com. A suitability evidence readiness check (eight questions; runs in the browser, nothing sent or stored) is at https://luscent.io/readiness. Solutions pages: https://luscent.io/solutions/private-banks · https://luscent.io/solutions/family-offices · https://luscent.io/solutions/external-asset-managers · https://luscent.io/solutions/independent-advisors --- # Luscent Product: Signals, Health Scores and Evidence > Markdown mirror of https://luscent.io/product for machine readers. Canonical page: https://luscent.io/product ## Your book, understood. Luscent reads the communications your firm already has, shows every relationship manager the next right conversation, and generates the suitability evidence as the work happens. ## The book 127 relationships, ranked by what needs attention (illustrative, synthetic data): | # | Client | Health | What moved | AUM | |---|---|---|---|---| | 01 | De Vries Holding | 54 | Business sale mentioned | 61.200.000 EUR | | 02 | Klein Family Office | 61 | Review declined twice | 42.000.000 EUR | | 03 | Weiss & Partners | 74 | Suitability review due | 128.400.000 EUR | | 04 | Mercier Patrimoine | 81 | Target market breach | 87.300.000 EUR | | 05 | Lindqvist Family | 84 | Next generation active | 33.900.000 EUR | | 06 | Aaltonen Holding | 88 | Reporting date met | 214.600.000 EUR | | 07 | Bergmann Privat | 92 | No change | 76.500.000 EUR | ## The score: one number per relationship, and you can open it Every relationship carries a Health Score from 0 to 100, rebuilt daily from the communications your firm already produces. Low means attention, not risk. The number is only worth having because it resolves: every score opens onto the four or five things that moved it, and each of those opens onto the messages they were drawn from. That is the difference between a score a relationship manager argues with and a score a supervisor can hold a conversation about. It survives being questioned, because the questioning is the point. ## Use case 01 · Retention (for the head of the desk) ### The client who is leaving tells you first. A departure is a sequence, not an event. The sequence is in the mailbox for months before it is in the AUM report, and it is legible: replies get shorter, the client starts opening the exchanges, a competitor gets named, a review gets declined twice. The distance between the first legible signal and the transfer request is the whole opportunity. Luscent surfaces the sequence while it is still a sequence. What the relationship manager does about it is theirs. ## Use case 02 · Growth (for the relationship manager) ### The money is already in the conversation. Liquidity events, successions, an inheritance, a business sale, a tax question that only makes sense if something is about to happen. Clients say these things in passing, to someone who is reading for something else, and they do not come back. Opportunities are ranked by what they are worth and how sure the system is, so a book of 127 relationships produces a list of three. ## Use case 03 · Evidence (for the compliance function) ### The file writes itself, while the advice happens. Suitability files get assembled afterwards, from memory, unevenly across a population of relationship managers. The reconstruction is the risk: it is slow, it is uneven, and at an inspection it is the thing that is hardest to defend. Compliance outcomes come from rules, never from a probability, and the rule that fired is on the record with its source. ## The Signal, opened A Signal is not a notification. It is a recommendation with its working attached: what was read, the rule that gated it, the model that read it, and whose decision it is. 1. Type and confidence, on the face. AI-derived confidence is a number. Compliance confidence is always 100, because a rule produced it. 2. The evidence, attached. The messages it was drawn from, as the model received them. Nothing to take on faith. 3. The rule that gated it. Deterministic and versioned. No Signal reaches a relationship manager without passing one. 4. The model and version. Which model read the messages, and that it ran in the EU. The serving boundary, on the record. 5. The decision. Drawn as controls because they belong to a person. The system recommends; it never acts alone. 6. The audit entry. Written when the decision is: who, when, and what changed. ## How it works: seven stages, and no path around them The order is the product, not the list. Consent cannot be recorded after the reading. Identifiers cannot be removed after the model has already seen them. A rule cannot be checked against an output nobody explained. No user setting skips a stage. The one stage not yet on every path is redaction: it runs on the document paths today, and the rest is in build. Every Signal, every time, in this order: 1. Consent: lawful basis recorded before anything is read 2. Redact: pseudonymised on document paths 3. Infer: EU-hosted, EU providers only 4. Rules: deterministic, no model discretion 5. Explain: evidence and rule attached to the output 6. Decide: the relationship manager, always 7. Audit: audit log of what happened and who Compliance Signals carry a confidence of 100 because a rule produced them. Signals derived from communications carry a number, and the number is always on the face of the Signal rather than in a footnote. ## What the model is given The shape survives. The identity does not. Redaction is stage two of the spine: it runs on the document paths today, and extending it to every language-model call is in build. ## What it will not do: four constraints, stated as constraints 1. It does not send your clients' words to a US model. Inference is hosted in the EU by EU providers only. No OpenAI, Anthropic or Google API sits in the path of client data. 2. It does not start on your real clients. The pilot runs on synthetic data first. Real correspondence enters once the firm has seen what the system does with the invented kind. 3. It does not act on its own. Every Signal is a recommendation. The relationship manager decides, and the decision is recorded against the client with the person who took it. 4. It does not ask you to take its word. Every Signal carries the rule that gated it, the messages it was drawn from, and the model and version that read them. ## The first month: four steps, each with something true at the end of it | Step | What happens | |---|---| | Connect | Microsoft 365 and Salesforce, authorised by the firm, scoped to what the firm chooses to share. Nothing is written back and nothing is migrated. | | First look | Every relationship carries a Health Score and a reason. The first conversation is about whether the ranking matches what the desk already believes. | | First Signal acted on | One Signal, one decision, one record. False positives are triaged weekly with the compliance function, and the rate is one of the success metrics agreed before day one. | | Compounding | A relationship representation accumulates from your own communications. Month six is worth considerably more than month one, and no amount of capital shortens that for anyone. | ## Where it lands: five surfaces, one of them daily | Surface | What it is | |---|---| | Dashboard | The prioritised feed. What needs attention today, in order. | | Clients | The book, in three views, filtered however the desk works. | | Client Detail | Seven tabs: Monitor, Detect, Execute, Connect, Allocate, Compliance, Documents. | | Signals | Book-wide inventory, searchable by type, urgency and date. | | Tasks | What a Signal turned into, and who owns it. | ## What it covers: product regulation only Two regulations are deliberately absent from this table. DORA and NIS2 describe how Luscent operates as your supplier rather than what the product produces, so they sit in the Trust Center instead. | Regulation | Reference | Status | What Luscent produces | |---|---|---|---| | MiFID II | Article 25 | Live | Suitability evidence written at the point of advice, with its source attached | | GDPR | Articles 28 and 32 | Live | Lawful basis recorded, document paths pseudonymised | | EU AI Act | Article 50 | Live | Model outputs labelled and sourced, a named person decides | | CVM and ANBIMA | Res. 30, 178, 35, 19, 50 | Live | Brazilian resolutions run in the same rules engine | | FinSA and FinIA | Not shipped | Roadmap | Swiss coverage is on the roadmap, not in the product today | | LGPD | Data handling | Acknowledged | Recorded in the way Brazilian client data is handled | | UK Consumer Duty | Not shipped | Roadmap | On the roadmap and not in the product. We would rather say so | | WpHG, WpDVerOV and MaComp | Not shipped | Roadmap | German conduct rules for banks and investment firms, not in the product today | | FinVermV (§34f GewO) | Not shipped | Roadmap | German rules for independent financial investment brokers, not in the product today | This describes what Luscent does, not what your firm is obliged to do. It is not legal or compliance advice. The platform generates evidence and surfaces Signals; decisions remain with the firm and its professionals at all times. The pilot is ninety days, free, with three to five relationship managers, starting on synthetic data. Success metrics are agreed before day one, so at the end there is something to judge rather than an impression. --- Luscent S.à r.l. is a société à responsabilité limitée in formation in Luxembourg, at House of Startups, 9 Rue du Laboratoire, L-1911 Luxembourg. Luscent stores and processes client data in the European Union only: storage on Scaleway in Paris, and the data and AI services on Hetzner in Germany. Sub-processors, certifications and security controls are documented in the Trust Center: https://luscent.io/trust-center --- # Luscent Trust Center > Markdown mirror of https://luscent.io/trust-center for machine readers. Canonical page: https://luscent.io/trust-center > Last reviewed 4 October 2026. This page states where Luscent runs, what it holds, who can reach it, and what is not yet true. Luscent is a seed-stage company and does not hold a security certification. Everything below is written so that a reviewer can check it rather than take it on trust. ## 01 At a glance The ten facts most reviewers open this page for. Every figure on this page traces to Luscent's infrastructure security map. Where a control is partial, this page says partial. | Aspect | Detail | |---|---| | Client data residency | European Union only | | Client data store | Scaleway PAR-2, Paris, France. Single region | | Data and AI services | Hetzner, Germany (Falkenstein, Nuremberg) | | AI inference | EU-hosted, EU providers only | | Non-EU model APIs in the client-data path | None | | Sub-processors with access to client data | Three | | Encryption at rest | AES-256 on the data store. Trace store partial | | Encryption in transit | TLS 1.3 public. Internal partial | | Client notification after a confirmed breach | Within 24 hours | | ISO/IEC 27001 | Controls aligned, certification not held | ## 02 Where your data lives ### All client data stays in the European Union Luscent stores and processes all client data in the European Union. Its client data store runs in a single region, Scaleway PAR-2 in Paris, France. Its data and AI services and its model trace store run on Hetzner in Falkenstein and Nuremberg, Germany. Luscent's own code, identity and internal communications run inside the Microsoft EU Data Boundary and hold no client data. Two Luscent sub-processors are not established in the European Union. Tailscale, a United States company, provides the encrypted network between Luscent’s own machines. Traffic is end-to-end encrypted with keys Tailscale does not hold, so Tailscale receives connection metadata only, even when its relays carry traffic that cannot connect directly. Mailgun sends Luscent's outbound email, and the sending domain is provisioned in Mailgun's United States region rather than its European one. Luscent states this here rather than leaving it to be discovered in a questionnaire. ### Data classification Luscent classifies data in four levels and maps each level to a zone. Client data is class C4. Its primary store is Scaleway PAR-2 in Paris; Luscent's data and AI services and its model trace store run on Hetzner in Germany. | Class | Contents | Zone | Protection | |---|---|---|---| | C1 Public | Marketing site, open-source contributions | Microsoft EU | TLS in transit | | C2 Internal | Internal documentation, architecture decisions | Microsoft EU | TLS, encrypted at rest | | C3 Confidential | Business communications, contracts, investor material | Microsoft EU | TLS, encrypted at rest, information protection labels | | C4 Restricted | Client relationship data, communication metadata, Signal outputs, financial summaries | Scaleway PAR-2; Hetzner, Germany | TLS 1.3 public, AES-256 at rest on Scaleway. Hetzner hops partial (section 03) | ## 03 How data moves ### Encrypted where it is stored, partial between hosts Client data reaches Luscent over TLS 1.3. Inside Scaleway, Luscent services talk over WireGuard on a private mesh, and client data is written to a managed PostgreSQL database and to object storage encrypted with AES-256 at rest. Backups are encrypted to the same standard as their source. The data and AI services and the model trace store run on two Hetzner hosts in Germany that are not on that mesh. Default-deny firewalls admit only Luscent's own production address. Encryption on the hops to and from those hosts, and at rest in the trace store, is marked partial below until it is confirmed. | Scope | Standard | Implementation | |---|---|---| | In transit, public | TLS 1.3 | Managed certificates with automatic renewal, HSTS enforced | | In transit, internal | WireGuard | ChaCha20-Poly1305 across the Tailscale mesh, between Luscent's Scaleway services | | In transit, to and from Hetzner | Partial | Default-deny firewalls admit only Luscent's own production address. Not on the WireGuard mesh | | At rest, database | AES-256 | Scaleway managed encryption, keys held in Scaleway KMS | | At rest, object storage | AES-256 | Server-side encryption, private bucket policy, versioning enabled | | Backups | AES-256 | Encrypted to the same standard as the source | | At rest, model trace store | Partial | Self-hosted on Hetzner, Germany. Encryption at rest to be confirmed | Encryption keys for data at rest are held in Scaleway's key management service. Customer-managed keys are not available today and are on the roadmap after the Series A. ### Where the record itself lives A firm that already runs a CRM keeps it, and Luscent reads from it. A firm without one will be able to use the Luscent CRM, which Luscent builds in-house, which is in build today and which is included for Luscent clients at no extra cost. It is deliberately lightweight, it is designed to run inside the same EU deployment, behind the same API boundary as the rest of the platform, and it is not a third-party product. Choosing it will add no sub-processor to the list in section 05. ## 04 AI and your data ### Inference is EU-hosted, from EU providers only Inference is hosted in the European Union by EU providers only. No OpenAI, Anthropic, Google or other non-EU model API sits in the path of client data. ### Training Luscent does not use client data to train or fine-tune models. Luscent consumes models as served infrastructure. ### Human review Luscent staff do not read client data as a matter of course. Access to production data is limited to investigating a fault or responding to a client request, is performed by named individuals under the access controls described in section 07, and is logged. ### Deterministic where it matters Luscent's compliance checks run on a deterministic rules engine, not on a language model. A rule produces the same result on the same input every time, which is the property an auditor needs and the property a probabilistic model does not have. Language models are used to read and summarise communications; they do not decide whether an obligation is met. ### EU AI Act Luscent is a provider of an AI system under Regulation (EU) 2024/1689. The Article 50 transparency obligations applied on 2 August 2026. The high-risk regime was rescheduled by Regulation (EU) 2026/1744 to 2 December 2027 for Annex III systems and 2 August 2028 for Annex I systems. Luscent's documented Article 50 position is available to clients and prospects on request. Luscent does not describe itself as compliant with an obligation before that is true. [Where AI is used in Luscent, and who decides](https://luscent.io/trust-center/ai-transparency) ## 05 Sub-processors and people ### Six sub-processors. Two outside the European Union. Luscent publishes its complete sub-processor list here and updates it when it changes. Jurisdiction is the second column because it is the second question. | Sub-processor | Jurisdiction | Purpose | Data | Agreement | |---|---|---|---|---| | Scaleway (Iliad Group) | France (PAR-2) | Compute, managed database, object storage, model inference | C4. Client relationship data and Signal outputs | Article 28 terms in Scaleway's data processing agreement | | Mistral AI | France (EU API endpoint) | Language model inference | C4. Client-derived text sent for inference | Article 28 terms in Mistral AI's data processing agreement | | Hetzner Online GmbH | Germany (Falkenstein and Nuremberg) | Hosting for Luscent's data and AI services and for observability, including the language-model trace store. Planned: encrypted backup copies | C4. Requests handled by the data and AI services, and client-derived text in model traces. Backup copies will be encrypted with keys held outside Hetzner. | Hetzner data processing agreement (Article 28), concluded before client data reaches Hetzner | | Microsoft | Microsoft EU Data Boundary | Identity (Entra ID), source control (Azure DevOps), internal communications (Microsoft 365) | C1 to C3. Code, internal documents, business communications. No client data. | Microsoft Products and Services Data Protection Addendum | | Tailscale | United States | Encrypted mesh network between Luscent services | Network metadata only. Traffic is end-to-end encrypted over WireGuard with keys Tailscale does not hold, including when a Tailscale relay carries it. | Tailscale data processing agreement | | Mailgun (Sinch) | United States | Outbound email sent from mailer.luscent.io | Recipient address and message content for email Luscent sends from that domain. | Mailgun data processing agreement | ### Where Luscent's people are Luscent's leadership is in Luxembourg and Germany. Luscent's engineering team is engaged through an employer of record in Brazil. Luscent states this plainly because a reviewer will establish it in ten minutes and should not have to. What bounds that access is written down and testable. Engineering access to production runs through Microsoft Entra ID with multi-factor authentication enforced, and through tag-based access control on the Tailscale mesh. No engineer holds direct access to the production database. Production changes reach production only through the deployment pipeline. Development and staging environments contain no client data. Access is revoked from a single checklist on departure. Personal data of Luscent staff is processed by the employer of record. Client data is not. ## 06 Certifications and regulation ### Luscent does not hold a security certification Luscent has mapped its controls to ISO/IEC 27001:2022 and that work is continuous. Luscent does not hold the certificate and does not describe itself as certified. Certification requires an accredited external audit, which is planned after the Series A. A SOC 2 Type I report is planned for 2027 and a Type II report for 2028. | Standard | Status | Target | |---|---|---| | ISO/IEC 27001 | Controls aligned, certification not held | Certification planned post-Series A, 2027 | | SOC 2 Type I | Not held | Planned 2027 | | SOC 2 Type II | Not held | Planned 2028 | ### What Luscent is, and is not | Question | Answer | |---|---| | What Luscent is | A technology provider: software that wealth managers use on their own Client communications | | What Luscent is not | A regulated financial institution. It holds no CSSF authorisation and gives no investment, legal or compliance advice | | Who decides | The client firm and its professionals, at all times. Luscent produces evidence and Signals; it decides nothing for the firm | | Under DORA | An ICT third-party service provider to the financial entities it serves | | Under the GDPR | Processor for client data, on the firm's documented instructions; controller for its own business data | | Where it is based | Luxembourg, in formation, and a member of the Luxembourg House of Financial Technology (LHoFT) | ### Regulation that applies to Luscent as a company These are obligations on Luscent as your supplier. DORA appears in this table and in no other, because DORA describes how Luscent operates as an ICT third-party service provider. It is not a feature of the Luscent product and Luscent does not present it as one. | Regulation | Instrument | Status | What it means here | |---|---|---|---| | GDPR | Regulation (EU) 2016/679 | Applies | Luscent is a processor of client data and a controller of its own business data. | | DORA | Regulation (EU) 2022/2554 | Applies as ICT third-party provider | Luscent supports its clients' Article 28 register and due-diligence obligations. Evidence available on request. | | EU AI Act | Regulation (EU) 2024/1689 | Applies as provider | Article 50 transparency obligations applied on 2 August 2026. Luscent's Article 50 position is documented and available on request. | | NIS2 | Directive (EU) 2022/2555 | Monitored | Applicability follows client sector and Luxembourg transposition. | ### Regulation the product helps you evidence These are obligations on you. Luscent produces the record that evidences them. The distinction between this table and the one above is deliberate: a supplier that blurs the two is describing its own compliance as though it were yours. | Regulation | What Luscent produces | |---|---| | MiFID II | Suitability evidence and audit trail | | GDPR | Lawful basis records, retention, subject-request support | | EU AI Act | Transparency and human oversight | | FinSA and FinIA | Nothing yet. Swiss coverage is on the roadmap | | WpHG, WpDVerOV and MaComp | Nothing yet. German conduct rules are on the roadmap | | FinVermV (§34f GewO) | Nothing yet. German broker rules are on the roadmap | | UK Consumer Duty | Nothing yet. UK coverage is on the roadmap | | Retail Investment Strategy (EU) | Nothing yet. Expected 2029 | | CVM and ANBIMA | Resolutions 30, 178, 35, 19 and 50 | | LGPD | Brazilian data-protection acknowledgment | ## 07 Security controls ### What is in place, and what is not Luscent is a seed-stage company. Its technical controls are ahead of its organisational ones, which is the normal shape at this size and is stated here rather than smoothed over. A control that is partial is labelled partial. ### Identity and access | Control | Status | Implementation | |---|---|---| | Multi-factor authentication | In place | Enforced for every Luscent staff account through Microsoft Entra ID. | | Least privilege | In place | Administrative infrastructure access is held by two people. Engineers cannot reach the production database directly. | | Network access control | In place | Tag-based access lists across the Tailscale mesh. | | Offboarding | In place | Identity deactivation, mesh removal and repository revocation run from a single checklist on departure. | | Periodic access review | Planned | Quarterly recertification of privileged access. | ### Engineering and change | Control | Status | Implementation | |---|---|---| | Peer review | Partial | Branch protection with pull-request review is being extended to every repository that deploys to production. | | Deployment path | Partial | A pipeline-only deployment path is being re-established for the current codebase. | | Environment separation | In place | Development, staging and production are separate instances on separate network tags. Development and staging hold no client data. | | Infrastructure as code | Partial | Core infrastructure is declared in version control. Edge configuration and part of the frontend hosting are being brought under it. | | Dependency auditing | Partial | Dependency advisories are checked in continuous integration. A dedicated scanning platform is planned. | | Secure coding standard | Partial | Static analysis and type-level enforcement are in place. A written standard and formal threat modelling are planned. | ### Monitoring and operations | Control | Status | Implementation | |---|---|---| | Application and infrastructure monitoring | In place | Metrics and logs with alerting on critical errors and resource thresholds. | | Network access logging | In place | Every mesh connection event is recorded. | | Time synchronisation | In place | All instances synchronised, all log timestamps in UTC. | | Central security event platform | Planned | Log aggregation into a single security monitoring platform is planned. Daily review and critical-error alerting are the current controls. | ### Policy and organisation | Control | Status | Implementation | |---|---|---| | Information security policy | Partial | Drafted. Formal approval is scheduled for the first board meeting after the RCS filing. | | Named security roles | In place | A security lead, an infrastructure owner, an application security owner and an executive accountable for every external security claim. | | Segregation of duties | In place | Two-person integrity for production change. | | Confidentiality undertakings | In place | Signed by every member of staff and every contractor. | | Security awareness training | Partial | Security onboarding is in place. Recurring formal training is planned. | | Device management | Partial | Full-disk encryption is enforced. Centrally managed devices are planned. | ## 08 Privacy and your rights ### Luscent is your processor, not your controller For client data, you are the controller and Luscent is the processor under Article 28 of Regulation (EU) 2016/679. Luscent processes client data only on your documented instructions. For Luscent's own business data, such as the details of the people it contracts with, Luscent is the controller. Luscent signs a data processing agreement with every client. The template is available before any commercial commitment, so that your legal review can start at the same time as your technical one rather than after it. Write to legal@luscent.io for the template. ### Data subject requests Where a data subject exercises a right against you as controller and the data sits in Luscent, Luscent assists with access, rectification, erasure and portability within the timeframes set by the data processing agreement. Requests go to privacy@luscent.io . ### Brazil Luscent acknowledges the Lei Geral de Protecao de Dados for Brazilian client data and maintains the corresponding records. Brazilian engineering personnel access is described in section 05. ### Retention Client data is retained for the term of the agreement and deleted or returned on termination as set out in the data processing agreement. Operational retention periods are below. | Record | Retention | |---|---| | Database backups and point-in-time recovery | 30 days | | Application and infrastructure logs | 30 days | | Network access logs | 90 days | | Identity and sign-in audit logs | 30 days | ## 09 Resilience ### Recovery objectives Luscent runs a high-availability managed database in Paris with automatic failover, and backups run automatically with 30 days of point-in-time recovery. Luscent's client data store runs in a single region and Luscent does not claim cross-region replication or failover. Recovery objectives for the loss of the Paris region are not yet stated. | Scenario | Recovery point objective | Recovery time objective | |---|---|---| | Single instance failure | No data loss | Under 5 minutes | | Database failure | Under 1 minute | Under 15 minutes | | Loss of the Paris region | Not yet stated | Not yet stated | | Data corruption | Under 24 hours | Under 2 hours | | Complete infrastructure loss | Not yet stated | Not yet stated | ## 10 Incident response ### Clients hear within 24 hours Luscent commits to notifying affected clients within 24 hours of confirming a security incident that affects their data. Article 33 of the GDPR allows 72 hours for notification to a supervisory authority. Luscent sets its own commitment 48 hours inside that ceiling, because a client who learns late cannot act at all. | Phase | Commitment | Detail | |---|---|---| | Detection | Continuous | Automated monitoring and alerting | | Classification | Within 1 hour | Severity assigned, P1 to P4 | | Containment | Within 2 hours for P1 | Within 8 hours for P2 | | Client notification | Within 24 hours | More restrictive than the 72 hours allowed by GDPR Article 33 | | Root cause analysis | Within 5 business days | Technical review | | Post-incident report | Within 10 business days | Written and shared | ### Reporting a vulnerability Security researchers should write to security@luscent.io . Luscent asks for a reasonable period to remediate before public disclosure, and does not pursue legal action against good-faith research conducted on those terms. ## 11 Documents ### Available now Luscent does not gate the two documents a reviewer needs first. The sub-processor list is on this page and the data processing agreement is available before any commercial commitment. The rest arrive by return of email rather than through a portal, because at Luscent's size a portal would be theatre. | Document | Access | Notes | |---|---|---| | Data processing agreement | On request | Article 28 template, available before commercial commitment | | Sub-processor list | Public | Published on this page and updated when it changes | | Architecture and data-residency overview | On request | Zone architecture with regions and data classes | | DORA Article 28 information pack | On request | For a client's register of information and due diligence file | | EU AI Act Article 50 position | On request | Documented provider assessment | | Security questionnaire response | On request | Completed against the reviewer's own format | Everything marked "On request" can be requested in one email to security@luscent.io (subject: Security pack request). A person replies. ## 12 Questions ### What Luscent holds **Where does Luscent store client data?** Luscent stores and processes all client data in the European Union. The client data store is Scaleway PAR-2 in Paris, France, a single region; Luscent's data and AI services and its model trace store run on Hetzner in Germany. No client data is stored or processed outside the European Union. **Does client data ever leave the European Union?** No. Luscent stores client data, processes it and runs model inference on it entirely within the European Union, and no non-EU model API sits in the path of client data. Two Luscent sub-processors are established outside the European Union. Tailscale, a United States company, provides the encrypted network between Luscent's own machines and receives connection metadata only, because that traffic runs peer to peer rather than through Tailscale servers. Mailgun sends Luscent's outbound email, and the sending domain is provisioned in Mailgun's United States region rather than its European one. **What client data does Luscent hold?** Luscent holds client relationship records, communication metadata, the Signals and Health Scores the platform derives from them, and financial summaries such as assets under management. Luscent classifies all of this as C4, its most restricted class. Its primary store is Scaleway PAR-2 in Paris; Luscent's data and AI services and its model trace store run on Hetzner in Germany. ### The Luscent CRM and integrations **What is the Luscent CRM, and is it a third-party product?** The Luscent CRM is built in-house by Luscent. It is not a third-party product, and Luscent does not bundle or resell another vendor's CRM. It is in build and not yet part of the shipped product. It is included for Luscent clients at no extra cost. It is deliberately lightweight: it is designed to hold client records, activity history and relationship data for firms that do not already run a CRM, on the same EU infrastructure as the rest of the platform, behind the same API boundary, and it introduces no additional sub-processor. **Does Luscent require a firm to replace its existing CRM?** No. Luscent is an intelligence layer over the systems a firm already runs. A firm with an existing CRM such as Salesforce keeps it, and Luscent integrates with it. A firm without one will be able to use the Luscent CRM, which is in build and included for Luscent clients at no extra cost. The system of record stays wherever the firm wants it. **Which systems does Luscent connect to?** Luscent integrates with Microsoft 365 and with Salesforce, and is building its own in-house CRM for firms that do not run one. Each connection is authorised by the firm through the provider's own consent flow and is scoped to what the firm chooses to share. ### AI **Which AI models does Luscent use, and where do they run?** Inference is hosted in the EU by EU providers only. No OpenAI, Anthropic, Google or other non-EU model API sits in the path of client data. **Does Luscent train models on client data?** No. Luscent does not use client data to train or fine-tune models. Luscent consumes models as served infrastructure. **Does a language model decide whether a firm is compliant?** No. Luscent's compliance checks run on a deterministic rules engine, not on a language model. A rule produces the same result on the same input every time, which is the property an auditor needs and the property a probabilistic model does not have. Language models are used to read and summarise communications; they do not decide whether an obligation is met. **How is Luscent classified under the EU AI Act?** Luscent is a provider of an AI system under Regulation (EU) 2024/1689. The Article 50 transparency obligations applied on 2 August 2026. The high-risk regime was rescheduled by Regulation (EU) 2026/1744 to 2 December 2027 for Annex III systems and 2 August 2028 for Annex I systems. Luscent's documented Article 50 position is available to clients and prospects on request. ### Who can reach it **Who at Luscent can access client data?** Access to production systems is limited to named individuals under enforced multi-factor authentication and tag-based network access control. Administrative infrastructure access is held by two people. No engineer holds direct access to the production database. Luscent staff do not read client data as a matter of course: access happens to investigate a fault or to answer a client request, and it is logged. **Where are Luscent's engineers located?** Luscent's leadership is in Luxembourg and Germany. Luscent's engineering team is engaged through an employer of record in Brazil. Engineering access to production is governed by enforced multi-factor authentication and tag-based network access control, and no engineer holds direct production database access. Development and staging environments contain no client data. **Does Luscent support single sign-on?** Yes. Luscent authenticates users through Microsoft Entra ID single sign-on, and session policy is configurable per client firm. **Which sub-processors have access to client data?** Three. Scaleway processes client data as Luscent's infrastructure provider in France; Hetzner hosts Luscent's data and AI services and its model trace store in Germany; and Mistral AI (France) is the primary language model inference provider for client-derived text, through its EU regional endpoint, with Scaleway's EU inference service as the fallback. Microsoft holds Luscent's own code, documents and internal communications and does not receive client data. Tailscale carries network metadata only, because traffic between Luscent services is end-to-end encrypted with keys Tailscale does not hold, including when its relays carry it. **Does Luscent use any United States sub-processor?** Yes, two. Tailscale is a United States company and provides Luscent's encrypted internal network. Tailscale receives connection metadata only. Client data does not reach Tailscale servers, because traffic runs peer to peer over WireGuard directly between Luscent's own instances. Mailgun sends Luscent's outbound email, and the sending domain is provisioned in Mailgun's United States region rather than its European one. ### Certifications and regulation **Is Luscent ISO 27001 certified?** No. Luscent has mapped its controls to ISO/IEC 27001:2022 but does not hold the certification. Formal certification is planned after the Series A, targeted for 2027. Luscent will not describe itself as certified before an accredited auditor has issued the certificate. **Is Luscent SOC 2 compliant?** No. Luscent does not hold a SOC 2 report. A SOC 2 Type I report is planned for 2027 and a Type II report for 2028. **Does Luscent hold a DORA registration?** DORA does not create a registration for ICT providers of Luscent's size. Luscent operates as an ICT third-party service provider to financial entities and supports its clients' obligations under Article 28 of Regulation (EU) 2022/2554, including the information required for their register of information. DORA is a statement about how Luscent operates, not a feature of the Luscent product. **Is Luscent a controller or a processor of client data?** For client data, the client firm is the controller and Luscent is the processor under Article 28 of Regulation (EU) 2016/679. Luscent processes client data only on the firm's documented instructions. For Luscent's own business data, such as the details of the people it contracts with, Luscent is the controller. ### Contracts, exit and incidents **Can Luscent sign a data processing agreement?** Yes. Luscent signs a data processing agreement under Article 28 of the GDPR with every client. The template is available on request before any commercial commitment. **What happens to client data when an agreement ends?** Client data is retained for the term of the agreement and is deleted or returned on termination, as set out in the data processing agreement. Backups age out of the 30-day point-in-time recovery window that applies to the production database. **How quickly does Luscent notify clients of a security incident?** Within 24 hours of confirming a security incident affecting client data. This is more restrictive than the 72 hours that Article 33 of the GDPR allows for notification to a supervisory authority. **What is Luscent's legal entity?** Luscent S.à r.l., a société à responsabilité limitée in formation in Luxembourg, with its registered office at House of Startups, 9 Rue du Laboratoire, L-1911 Luxembourg. The company is in formation and its registration number will be published here once the filing with the Registre de Commerce et des Sociétés is complete. **Who does a security researcher contact?** security@luscent.io. Luscent asks researchers to allow a reasonable period for remediation before public disclosure and does not pursue legal action against good-faith research conducted under those terms. ## 13 Contact Each address reaches a monitored mailbox, not a form. Security and vulnerability reports: security@luscent.io Privacy and data subject requests: privacy@luscent.io Contracts and data processing agreements: legal@luscent.io Legal entity: Luscent S.à r.l. (in formation) Registered office: House of Startups, 9 Rue du Laboratoire, L-1911 Luxembourg Luscent S.à r.l. is in formation. Its registration number will be published on this page once the filing with the Registre de Commerce et des Sociétés is complete. This page was last reviewed on 4 October 2026 and is updated when the underlying architecture, sub-processor list or certification status changes. --- # AI Transparency: Where AI Reads, and Who Decides > Markdown mirror of https://luscent.io/trust-center/ai-transparency for machine readers. Canonical page: https://luscent.io/trust-center/ai-transparency Language models read and summarise communications; they do not decide whether an obligation is met. No client-facing output leaves the platform without a relationship manager approving it. ## Where AI is used, and where it is not | Task | How it is done | Who decides | |---|---|---| | Reading Client communications | Language models read and summarise them | Nothing is decided here | | Raising a Signal from what a Client wrote | A language model, shown with a confidence | The Relationship Manager, after checking the source | | Deciding whether a compliance rule is met | A deterministic rules engine, never a language model | The rule, which you can read | | Drafting anything meant for a Client | A language model drafts it | A named person, before anything leaves | ## What every Signal carries Every Signal carries the rule that gated it, the messages it was drawn from and the model that read them, kept seven years. The same holds on every tier. - **Compliance Signals show 100%.** They come from deterministic rules, not from reading language. A rule either fired on the Client's data or it did not. - **Other Signals show a confidence.** It says how sure Luscent is that it read a change correctly. Use it to decide how closely to check the source. - **Microsoft 365 and Salesforce.** Read and never written. Your systems of record stay exactly as they are. ## Where inference runs, and what it learns Inference is hosted in the European Union by EU providers only. No OpenAI, Anthropic, Google or other non-EU model API sits in the path of client data. Luscent does not use client data to train or fine-tune models. Luscent consumes models as served infrastructure. The providers are named in the Trust Center's sub-processor list: https://luscent.io/trust-center#subprocessors ## People, not scores Luscent supervises Client files and the firm's obligations, never the Relationship Manager. There is no RM-level score, ranking or rate anywhere in the product. The Health Score belongs to the Client relationship. ## The EU AI Act Luscent is a provider of an AI system under Regulation (EU) 2024/1689. The Article 50 transparency obligations applied on 2 August 2026. Luscent's documented Article 50 position is available to clients and prospects on request. Luscent does not describe itself as compliant with an obligation before that is true. ## Questions **Does Luscent use client data to train models?** No. Luscent does not use client data to train or fine-tune models. Luscent consumes models as served infrastructure. **Where does inference run?** Inference is hosted in the European Union by EU providers only. No OpenAI, Anthropic, Google or other non-EU model API sits in the path of client data. The providers are named in the Trust Center's sub-processor list. **Can a language model decide a compliance outcome?** No. Luscent's compliance checks run on a deterministic rules engine, not on a language model. A rule produces the same result on the same input every time. Language models are used to read and summarise communications; they do not decide whether an obligation is met. **Does it score our Relationship Managers?** No. Luscent supervises Client files and the firm's obligations, never the Relationship Manager. There is no RM-level score, ranking or rate anywhere in the product. The Health Score belongs to the Client relationship. **Where is Luscent's EU AI Act position?** The Article 50 transparency obligations applied on 2 August 2026. Luscent's documented Article 50 position is available to clients and prospects on request. Luscent does not describe itself as compliant with an obligation before that is true. --- # Luscent FAQ: The Questions That Actually Get Asked > Markdown mirror of https://luscent.io/faq for machine readers. Canonical page: https://luscent.io/faq Short answers on data, evidence, procurement, and what is actually shipped. If the one you need is not here, ask and it will be. ## What it does, and what it is not **01 Is this a CRM?** No. A CRM records what happened. Luscent reads the same substrate and tells you what it means and what has been missed. A native CRM exists as a separate Luscent product for firms that do not have one. It is in build, and it is included for Luscent clients at no extra cost; firms on Salesforce keep Salesforce. **02 How is Luscent different from Microsoft 365 Copilot?** Copilot is a general assistant across Microsoft 365: it drafts, summarises and searches for any employee. Luscent does one job for wealth managers. It reads the Client communications your firm already records, keeps a Health Score for each Client, raises Signals, and assembles MiFID II suitability evidence that a named Relationship Manager approves. The two can run side by side. **03 Does it replace the relationship manager?** No, and it is built so that it cannot. No client-facing output leaves Luscent without a person approving it. Every Signal is a recommendation, and the relationship manager decides. **04 Does it write anything back into our systems?** No. Nothing is written back into Microsoft 365, Salesforce or any system of record. Luscent reads, and your systems of record stay exactly as they are. ## Data, and where it goes **05 Where is client data processed and stored?** On European infrastructure. Client data is stored on Scaleway in Paris, a single EU region, and the data and AI services run on Hetzner in Germany. Identity and mail run on Microsoft's EU platform. **06 Does client data leave the EU at any point?** No. There are no international data transfers, and no US model API sits in the client-data path. **07 Which models do you use, and are they American?** No. Inference is hosted in the EU by EU providers only, and no OpenAI, Anthropic or Google API sits in the client-data path. Where they are served is the claim that matters. **08 What does the model actually see?** On the document paths, identifiers are pseudonymised before the model call. Extending that to every language-model call is in build, and it is validated before it reaches production. Inference is EU-hosted, by EU providers only, with no US model API in the client-data path. ## Compliance and evidence **09 How is a compliance outcome produced?** By a deterministic rules engine, never by a model. A compliance outcome carries the rule that fired and no confidence score at all, because a regulatory result is not a guess. Signals produced by a model carry a confidence, and the two are drawn differently everywhere on purpose. **10 Can we see why a Signal was raised?** Yes. Every Signal carries the rule that gated it, the source messages it was drawn from, and the model and version involved. A Signal acted on records the decision, who took it, and when. **11 How does this affect our DORA outsourcing assessment?** Luscent is a vendor under your process, not a DORA feature. You run the outsourcing assessment under CSSF 25/882 and DORA Article 28, and we supply a DORA Article 28 information pack and register inputs on request. DORA is company readiness on our side; it is not something the product does for you. ## Getting started **12 What does the pilot involve?** Ninety days, free, three to five relationship managers, starting on synthetic data. Success metrics are agreed before day one, so at the end there is something to judge rather than an impression. **13 Do we have to give you production data to evaluate it?** No. The pilot starts on synthetic data, and you decide if and when real correspondence enters it. **14 What do we need in place before we start?** Microsoft 365, with or without Salesforce. That is the whole dependency today. **15 What happens if we stop?** Data exports and a GDPR data processing agreement are standard, and source-code escrow is available. ## Where the company actually is **16 Who else is using it?** No completed pilots yet. The first working version was shown at the Luxembourg House of Financial Technology AI Experience Centre on 13 April 2026, and the first cohort runs in the fourth quarter of 2026. Applications are open. **17 What is shipped today, and what is not?** Product version 1.4. Signals generate live on real books in production, the Signal lifecycle is complete, and the suitability gate is auditable with every override recorded. The surfaces are Dashboard, Clients, Client Detail, Tasks, Signals, Strategy Director, House Views, Data Hub, the Compliance Home dashboard and the Audit Trail Viewer, with Microsoft 365 and Salesforce connected. In build, and described as in build: the native Luscent CRM, the Chatbot page and the Notes tab. Multi-seat access control is release 1.5. The second engine for structured portfolio data is release 2.0, and it is not dated. **18 How do you version releases?** A release ships when its bar is met, not on a date. Every release, what it changed and when it landed is in the changelog. Anything not listed there has not shipped, whatever else you may have been shown. ## Procurement, and who does what **19 What does our IT team have to do?** Two things, and neither is a build. Read the Trust Center, and when you choose to connect, authorise read-only access to Microsoft 365, and to Salesforce if you run it, through your own consent flow and scoped to what you choose to share. The first four weeks run on synthetic clients, so neither step waits on a client record. **20 Who has to sign, and who else is involved?** The firm signs the SOW, not an individual relationship manager. Around it: a sponsor from compliance or private banking leadership, three to five named relationship managers, your compliance function for the weekly false-positive triage, and half an hour a week with one named person on our side. **21 Do we need our own AI or data team?** No. Luscent runs the environment and the models, and nothing is hosted on your side. Your side brings people who know the book, not engineers: the relationship managers who use it, the compliance function that triages it, and a sponsor who takes the read-out at week eight and the review at the end. ## Asked before signature **22 You are an early-stage company. What happens to our data if you stop?** The pilot starts on synthetic data, so in its first weeks there is nothing of yours to lose. Under an agreement, Client data is kept for its term and deleted or returned on termination, as the data processing agreement sets out, and backups age out of a 30-day recovery window. Data exports are standard and source-code escrow is available, and because Luscent writes nothing back, your own systems stay complete without it. **23 Does it score our Relationship Managers?** No. Luscent supervises Client files and the firm's obligations, never the Relationship Manager. There is no RM-level score, ranking or rate anywhere in the product. The Health Score belongs to the Client relationship. --- Answers are kept in step with the Trust Center. Where the two ever differ, the Trust Center is the one that counts: https://luscent.io/trust-center --- # Luscent vs a CRM > Markdown mirror of https://luscent.io/compare/luscent-vs-crm for machine readers. Canonical page: https://luscent.io/compare/luscent-vs-crm A CRM records what happened. Luscent reads the communications your firm already has and tells you what they mean, what has been missed, and what you can prove. Most firms need both and keep the CRM they have. A firm without one will be able to start on ours. ## Is Luscent a CRM? No. A CRM is a system of record: it holds what someone entered. Luscent is an intelligence layer over the systems a firm already runs. A firm on Salesforce keeps Salesforce, and Luscent integrates with it. A firm without a CRM will be able to start on the Luscent CRM, which is in build and included for Luscent clients at no extra cost. ## What each one does with the same Client | Question | A CRM | Luscent | |---|---|---| | What it holds | Contacts, accounts, activities and pipeline: what someone chose to enter, when they had time to enter it. | The emails, call notes and meeting summaries the firm already has, read alongside the CRM record. | | How a change reaches the RM | When someone logs it, or at the next periodic review. | As a Signal, raised from the Client's own words: a business sale mentioned, a review declined twice. | | The view of a relationship | Fields, stages and the date of the last activity. | A Health Score from 0 to 100, rebuilt daily. Every score opens onto the things that moved it, and each of those onto the messages it came from. | | What to do next | Tasks that someone created. | The next right conversation for each Relationship Manager, ranked by what needs attention. | | Suitability evidence | Notes written after the fact, rebuilt from inboxes when a reviewer asks. | Evidence written as advice happens, with the exchanges attached. A deterministic rule: it fired or it did not, and you can read why. | | Does it change the record? | It is the record. | No. The connections to Microsoft 365 and Salesforce are read-only, and Luscent writes nothing back into them. | | Who decides | The Relationship Manager. | Still the Relationship Manager. A named person approves before anything reaches a Client. | ## When a CRM is enough, and when it is not **A CRM alone is enough when:** - You need a single record of contacts, accounts and activity - The question is what happened, and someone logged it - Pipeline and reporting are the job **Luscent adds to it when:** - What a Client means is in the mailbox, not in the fields - A cooling relationship should be visible before the AUM report shows it - A reviewer asks what the firm knew, and when, and you want to show them ## How Luscent works alongside your CRM 1. **Connect.** Microsoft 365 and Salesforce, each authorised by the firm through the provider's own consent flow and scoped to what the firm chooses to share. 2. **Read.** Luscent reads the communications, scores every relationship and raises a Signal when something moves. 3. **Act.** The Relationship Manager decides what to do. The evidence is filed as the work happens, and a named person approves before anything reaches a Client. ## No CRM yet? Start on the Luscent CRM A firm without a CRM will be able to start on the Luscent CRM: a deliberately lightweight record of Clients, their contacts and their MiFID suitability data, built in-house. It is in build today, and it is included for Luscent clients at no extra cost. - **Clients:** Name, Client type, Client phase, Health Score, Owner, MiFID client category - **Contacts:** Name, Email, Phone, Client, Politically exposed person - **Suitability (MiFID):** Knowledge level, Loss-bearing capacity, Investment horizon, ESG preferences - **Built in-house.** Not a third-party product. Luscent does not bundle or resell another vendor's CRM. - **No new sub-processor.** Choosing it adds no sub-processor to the Trust Center list. - **Moves with you.** CSV import with a review step before anything is written, and CSV export when the firm moves to a larger CRM. ## Questions **Does Luscent replace my CRM?** No. Luscent is an intelligence layer over the systems a firm already runs. A firm with an existing CRM such as Salesforce keeps it, and Luscent integrates with it. The system of record stays wherever the firm wants it. **We have no CRM. Do we need one first?** No. A firm without a CRM will be able to start on the Luscent CRM, built in-house: Clients, their contacts and their MiFID suitability data, with the Health Score on the Client record. It is in build today and included for Luscent clients at no extra cost. When the firm moves to a larger CRM, its records export as CSV. **Which systems does Luscent connect to?** Microsoft 365 and Salesforce. Each connection is authorised by the firm through the provider's own consent flow and is scoped to what the firm chooses to share. **Is the Health Score a lead score?** No. It is a reading of one existing relationship, from 0 to 100, rebuilt daily from the communications your firm already produces. Low means attention, not risk, and every score opens onto what moved it. **Can we try it without touching our CRM?** Yes. The pilot runs for ninety days, free, with three to five Relationship Managers. It starts on synthetic data, stays in the EU and writes nothing back. --- # Luscent vs building it in-house > Markdown mirror of https://luscent.io/compare/luscent-vs-in-house for machine readers. Canonical page: https://luscent.io/compare/luscent-vs-in-house A firm with a data team can build its own way to read Client communications, and some should. Either way, the question at the end is the same: can you show what the firm knew, and when? **Should we build this ourselves?** You can. A firm that builds it takes on the connections to mail, calendar and CRM, the reading of communications, the scoring, a rules engine for suitability, the evidence trail and the AI governance around all of it, and keeps each of them current as the rules change. Luscent is that system already, hosted in the EU, with a free ninety-day pilot that starts on synthetic data. ## What a build has to include, and where Luscent stands | Part | Luscent | |---|---| | Connections: mail, calendar and CRM, read-only and scoped by the firm | Live | | Reading the communications: language models hosted in the EU, by EU providers | Live | | A score per relationship: Health Score from 0 to 100, each change traced to its source | Live | | Suitability rules: MiFID II, GDPR, the EU AI Act, CVM and ANBIMA | Live | | More jurisdictions: Swiss, UK and German rules | Roadmap | | The evidence trail: written as the work happens, exchanges attached, retained seven years | Live | | Named approval: a person approves before anything reaches a Client | Live | | A record for firms without a CRM: the Luscent CRM, included at no extra cost | In build | ## When building fits - You have a data and AI team with room beyond its current roadmap. - Your communications estate is unusual enough that no product will fit it. - You want to own the model choice and every line of the evidence design. ## When Luscent fits - The evidence question needs an answer before an internal build would ship. - Your team should spend its time on Clients, not on maintaining a rules engine. - You want to judge it on your own book first, without production data. ## Questions - **Do we have to move any data into Luscent?** No. Luscent reads Microsoft 365 and Salesforce where the data already sits, through connections the firm authorises and scopes, and writes nothing back into them. - **Does Luscent use our data to train models?** No. Luscent does not use client data to train or fine-tune models. - **What happens to our data if we stop?** Under an agreement, Client data is kept for its term and deleted or returned on termination, as the data processing agreement sets out, and backups age out of a 30-day recovery window. Data exports are standard and source-code escrow is available. - **How do we compare the two fairly?** Run the pilot: ninety days, free, three to five Relationship Managers, starting on synthetic data, with success metrics agreed before day one. Then put the same ten vendor questions from the white paper to your internal build. --- # Luscent Integrations: What Luscent Connects To, and What It Does Not Yet Source: https://luscent.io/integrations Luscent reads the systems your firm already runs and writes nothing back. Two connections are live today. The rest are marked for what they are: in build, or on the roadmap with no date promised. | System | What it carries | Status | Note | |---|---|---|---| | Microsoft 365 | Mail, calendar, Teams | Live | Read-only. Pilots start here. | | Salesforce | Client records and activity, including Financial Services Cloud | Live | Read-only. Scoped separately for a pilot. | | Luscent CRM | Client records, activity history and relationship data | In build | Built in-house for firms that do not run a CRM. Included at no extra cost. | | Core banking | Positions and balances | Roadmap | Not shipped. No date is promised. | | Portfolio management | Holdings and returns | Roadmap | Not shipped. No date is promised. | | Consolidation | Positions across custodians | Roadmap | Not shipped. No date is promised. | | Custodian platforms | Positions and statements | Roadmap | Not shipped. No date is promised. | | Advisor platforms | Holdings and valuations | Roadmap | Not shipped. No date is promised. | ## How a connection works - **Reads, never writes.** Nothing is written back into any connected system. Your systems of record stay exactly as they are. - **Authorised by the firm.** Each connection runs through the provider's own consent flow and is scoped to what the firm chooses to share. - **Stays in the EU.** Client data is stored on Scaleway in Paris; Luscent's data and AI services run on Hetzner in Germany. ## Questions **Which systems does Luscent connect to?** Luscent integrates with Microsoft 365 and with Salesforce, and is building its own in-house CRM for firms that do not run one. Each connection is authorised by the firm through the provider's own consent flow and is scoped to what the firm chooses to share. **Does Luscent require a firm to replace its existing CRM?** No. Luscent is an intelligence layer over the systems a firm already runs. A firm with an existing CRM such as Salesforce keeps it, and Luscent integrates with it. A firm without one will be able to use the Luscent CRM, which is in build and included for Luscent clients at no extra cost. The system of record stays wherever the firm wants it. **Do we have to move any data into Luscent?** No. Luscent reads Microsoft 365 and Salesforce where the data already sits, through connections the firm authorises and scopes, and nothing is written back into them. **Does Luscent write anything back into our systems?** No. Luscent reads, and nothing is written back into any connected system. Your systems of record stay exactly as they are. **What does a pilot connect?** Pilots start on Microsoft 365, with three to five relationship managers. A CRM connection is scoped separately with the firm. --- # Ten Questions to Ask Any Vendor That Reads RM Communications Source: https://luscent.io/recorded-not-rated#checklist (Appendix B of the white paper "Recorded, Not Rated") Before an AI system reads your firm's recorded Client communications, put these ten questions to the vendor. Each comes with what a good answer looks like. It applies to every vendor, Luscent included. 1. **What is the documented intended purpose of the system?** A good answer: a written statement, consistent across the instructions for use, the contract and marketing, that names Client relationship management and suitability evidence as the purposes and expressly excludes monitoring or evaluating employees. 2. **Does any output exist, even internally, per employee?** A good answer: "No." Architecture diagrams show outputs keyed to Client identifiers; the vendor can show that no per-RM table exists in the data model; management views aggregate at firm level. 3. **Does the system perform profiling, of whom, and why?** A good answer: Client profiling occurs (relationship health, suitability indicators), with the legal basis and a DPIA template. Staff profiling does not occur, and the vendor explains how this is enforced. 4. **Where is your Article 6(3) and Annex III self-assessment?** A good answer: a versioned document, shared under NDA, that analyses 4(b) against the Commission's draft examples, states whether the filter is relied on, and commits to updating it when the final guidelines are published. 5. **Which Article 50 obligations apply, and how do you meet them?** A good answer: mapping of Client-facing chat (disclosure at first interaction), generated text (machine-readable marking or a reasoned exemption, Code of Practice status), and confirmation that no emotion recognition is used. 6. **Where are models hosted and data processed?** A good answer: named hosting provider and region, model families used (for example open-weight models on EU infrastructure, as one option), no training on customer data without a contract, and the transfer position if any processing leaves the EEA. 7. **What do you provide to support our DORA obligations?** A good answer: the information needed for the buyer's register of information (DORA Article 28), acceptance of mandatory contract terms (Article 30), incident notification cooperation, and audit and access rights. The buyer owns these obligations; the vendor enables them. 8. **Who are your sub-processors?** A good answer: a current list with location and function, advance notice of changes, and a right to object. 9. **What do you log, and how long do you keep it?** A good answer: logs sufficient to trace every Signal and suitability prompt, retention aligned with the firm's MiFID II schedule, and access controls that stop logs being used to reconstruct RM performance. 10. **How do we exit?** A good answer: export of all Client-level outputs and evidence links in open formats, deletion certification, a transition assistance period, and no lock-in of the underlying recordings, which stay in the firm's archive. --- # The Luscent Manifesto: Proof Should Be a By-Product > Full text of https://luscent.io/manifesto ## Understanding a client and proving you understood should be one act. A manifesto for European wealth management. For twenty years, European wealth management has been sold two different systems for one job. One holds the relationship: contacts, positions, logged calls, the tidy record of what happened. The other holds the proof: suitability files, review notes, the paper trail an auditor will eventually ask for. Between them sits a person, typing the same thing twice. That split was a technical limitation that hardened into a belief. Firms came to accept that knowing a client well and being able to prove it were separate kinds of work, one valuable and one merely necessary, and that the second would always be paid for in evenings. It has never been true, and it is now expensive. Every European rule since MiFID II has raised the standard of proof and left the method to the firm. Meanwhile the thing that would satisfy that standard, the actual record of what was said and when, sits unread in mail, call notes and meeting summaries that no system was ever asked to understand. The knowledge is already in the building. The proof is already in the knowledge. What is missing is a system willing to read it. ## The work Afterwards: assembled from memory. Reconstructed afterwards, the evidence is one late block and a compliance officer taking your word for the middle. As it happens: produced as the work happens, there is a mark under every conversation at the moment it can still be checked. Same work, same people, one diagonal replaced by a vertical. ## What follows from it ### 01 Evidence should be a by-product, not a project. If proving the decision costs a separate afternoon, it will be done late, done thin, or reconstructed under pressure a year later. Evidence produced while the work happens costs almost nothing, because the work was going to happen anyway. ### 02 Nothing is asserted without its source. Every Signal opens onto the exchange it came from. Every Health Score opens onto the movement behind it. A number a firm cannot trace is not intelligence, it is an opinion with a typeface. ### 03 The person decides, and the system proves they did. No client-facing output leaves Luscent without a Relationship Manager approving it. Judgement is not the bottleneck to be automated away. It is the part a regulator, and a client, is actually asking about. ### 04 Sovereignty is architecture, not a setting. Client data processed and stored in the EU. Language models, hosted in the EU by EU providers, reading the communications. Compliance detection on a deterministic rules engine, because a rule either fired or it did not, and a supervisor will not accept a probability as an answer. None of this can be switched on afterwards. ## Notes 1. MiFID II, Directive 2014/65/EU, applicable since 3 January 2018. Firms must keep records sufficient to demonstrate suitability, for five years as a minimum. 2. EU AI Act, Regulation (EU) 2024/1689. The Article 50 transparency obligations have been in force since 2 August 2026. Record what happened. Understand what it means. Prove you knew. One motion. This is what a system of intelligence is for, and it is the whole reason Luscent exists.