[00] Use case 04 / 04 · Relationship DriftSynthetic data throughout

Relationships rarely end loudly. They drift.

Replies slow down, a review is moved twice, another firm's name appears in a thread. Each on its own is noise. Luscent reads the communications the firm already has, shows the pattern in the Health Score with the messages behind it, and puts the next right conversation on the relationship manager's list.

The case, in four factsRelationship Drift

  • TriggerA pattern across 90 days of client communications
  • ShowsThe Health Score trend and the messages behind each Signal
  • ReturnsA Task with a drafted next step
  • DecidesThe relationship manager; nothing reaches the client without them

One finding, one decision at a time.

Client · MonitorDemo · synthetic
Clients / Klein Family Office · Luxembourg · São PauloHealth Score 75
Health Score · last 13 weeks
75Neutral · from 76
At risk below 60
Signals · this client

Reading 90 days of email, calendar and CRM activity. No open Signals yet.

Klein Family Office, Luxembourg. Anne Keller's client for six years. Nothing looks wrong on any single day.

Reading
Step 1 of 6: Ninety days, read quietly
[01] The storyToday · the moment · with Luscent

From noise to the next conversation.

When a client relationship slips, Luscent shows it in the Health Score, links the messages behind it and lists the next conversation for the relationship manager.

  1. 01 · Today

    Nothing looks wrong on any single day.

    A reply takes a week. A review moves twice. A forwarded thread names another firm. Each is noise on its own, and together they are a relationship going quiet.

  2. 02 · The moment

    The pattern, with its evidence.

    The Signals are linked into one pattern, and the Health Score moves, with the messages behind the change beside it.

  3. 03 · With Luscent

    The next right conversation.

    A Task with a drafted note reaches the relationship manager, who decides what to say and when. The follow-up is on the record.

[02] Who gets whatFour seats

One workflow,four people who stop chasing it.

Relationship manager

  • The relationships that need attention, with the reason
  • A drafted note to start from, never sent without you

OutcomeTime goes to the right conversation.

Compliance officer

  • Contact and follow-up recorded as they happen
  • Every Signal traceable to its source message

OutcomeClient care that can be shown.

Head of advisory

  • The book's Health Score trend in one view
  • Cover for a colleague's clients with the history attached

OutcomeFewer relationships go quiet unnoticed.

Management

  • Fee income retained where a cooling relationship is caught early
  • A signal to act on, not an exit interview

OutcomeRetention acted on early, not explained late.

[03] The business caseSame model as the full estimate

Fee income kept where a relationship was cooling.

Relationship Drift works on retention: relationships noticed early enough to act on. This is the part of the estimate on the pricing page it addresses.

  • 01Cooling relationships surfaced with their evidence
  • 02The next step on the relationship manager's list
  • 03The follow-up recorded, not remembered
  • PilotMeasured on your own book in the pilot, against success metrics agreed in writing before anything is connected.
Estimate · your firmModelled, not measured
Your firm
Fee income retained
900.000 EUR a year
20 % of at-risk relationships kept
Relationships at risk a year
12
Fee income per client
375.000 EUR
The working

12 relationship managers × 10 clients × 10 % annual attrition × 375.000 EUR fee income (75 bps on 50 m) × 20 % of at-risk relationships kept. The same model and assumptions as the estimate on the pricing page. Modelled from time-and-motion assumptions, not measured client results.

The full estimate, on the pricing page
[04] BoundariesAnd how AI is used here

What it always does,and what it never does.

Always

  • Links every Signal to the messages behind it.
  • Groups related Signals into one pattern instead of three notifications.
  • Shows the Health Score trend, not only today's number.
  • Puts the next step on the relationship manager's list as a Task.

Never

  • Contacts the client. No client-facing message leaves without the relationship manager's approval.
  • Predicts that a client will leave. It shows what changed and where to look; the relationship manager judges what it means.
  • Reads beyond what the firm connects: Microsoft 365 and Salesforce, read-only.
How AI is used · Reads

A language model reads the connected messages into Signals, each shown with a confidence and linked to its source.

How AI is used · Decides

The relationship manager judges what a Signal means. Nothing reaches the client without their approval.

How AI is used · On the record

Every Signal, the messages behind it and the Task it created. The score belongs to the relationship, never to the relationship manager.

The EU AI Act's Article 50 transparency obligations applied on 2 August 2026. Luscent's documented position is available on request. Where AI reads, and who decides

[05] The ruleOur reading, not legal advice

Our reading of MiFID II (Directive 2014/65/EU), Article 24(1)

“... an investment firm act honestly, fairly and professionally in accordance with the best interests of its clients ...”

The duty does not pause when a client goes quiet. A relationship that has drifted is where advice goes stale and service slips unnoticed; seeing it early is how the firm keeps acting in the client's interest instead of finding out at the transfer request.

Our reading, not legal advice. The rules engine applies the firm's own configuration of these rules, and the relationship manager decides.

[06] QuestionsAsked before signature

Questions.

What does the Health Score measure?

The state of a client relationship on a 0 to 100 scale, read from the communications and activity the firm connects: 80 to 100 healthy, 60 to 79 neutral, 40 to 59 at risk, below 40 critical. The trend matters as much as the number, so Luscent shows both, with the Signals that moved it.

Does a falling score mean the client is leaving?

No. It means the relationship has changed in ways worth a conversation: slower replies, a meeting moved, a new name in a thread. Luscent shows the messages behind the change; the relationship manager decides what it means and what to do. The score describes the client relationship, not the person managing it.

The pilotCohort open

Find out before you move a single client record.

Ninety days, at no cost, with three to five relationship managers, starting on invented clients. Success metrics are agreed in writing before anything is connected, so at the end there is a decision rather than an impression.

The terms

Length
90 days
Cost
Nothing
Seats
3 to 5 relationship managers
Data
Synthetic first
Client data
Stays in the EU
Writes back
Nothing
  1. DAY 0Metrics agreed, in writing
  2. ConnectWeek 1 to 4
  3. WEEK 4The book, ranked
  4. First Signal acted onWeek 5 to 8
  5. WEEK 8A decision on the record
  6. ReviewWeek 9 to 13
  7. DAY 90Written review
  • DataSynthetic first
  • ScopeRead-only
  • Writes backNothing
  • Seats3 to 5 RMs
  • EndsDay 90, written review