Reading the Room

The Problem

Most retail financial services journeys are designed around a standard customer in a standard situation. The digital flow assumes literacy, stability, and sufficient cognitive bandwidth to complete a transaction. The customer who is grieving, financially stressed, or cognitively impaired moves through the same journey and either completes it poorly, abandons it, or reaches a human who has no context for their circumstances. The signals were present in the journey data. The design was not built to use them.

The Pattern

The structural move is to design every customer touchpoint to detect circumstance signals and propagate them across the entire service delivery chain, so that every subsequent interaction responds consistently:

The structural moves

Behavioural signals

Present in the journey itself without requiring disclosure: extended session times, repeated form abandonment, unusual navigation patterns, accessibility feature usage, requests for callbacks, multiple failed attempts at the same task. These are observable in existing journey data and should trigger adapted responses automatically.

Contextual signals

Arising from the nature of the interaction: a bereavement notification, a request linked to redundancy, a query about payment difficulty, a request to access a pension early under unusual circumstances. These should automatically trigger a different response posture — not just a flag in the record but a change in how the next interaction is handled.

Cross-chain propagation

Detected signals must travel with the customer across channels, functions, and time. A vulnerability flag set on a call must be visible in digital, at renewal, and during claims or settlement. The propagation mechanism must be systematic — embedded in the customer record and surfaced to every touchpoint — not dependent on individual agents remembering to check notes.

Indicators It's Working

✓Vulnerability identification rates increase toward population-representative levels as designed disclosure pathways and behavioural detection take effect

✓The gap between internally reported vulnerability volumes and statistically expected population rates closes over time

✓Context flags persist accurately across channels without requiring customer re-disclosure — measurable by testing whether a flag set in one channel is visible in another

✓Post-interaction surveys show vulnerable customers feel their circumstances were understood and responded to, not just recorded

In Practice

Worked example 1

A retail bank analysed customers who had notified its collections team of financial difficulty and found they were subsequently contacted by the lending team within weeks with credit card and overdraft offers — the financial difficulty record sat in the operational system but never reached the marketing eligibility engine. The bank built a context propagation layer that flags all accounts associated with a financial difficulty notification, suppressing acquisition contact for a defined period and routing any product enquiry through specialist support. The technical change was modest; the experience change was significant. Customer complaints about insensitive contact during financial difficulty fell by over half within two quarters. The FCA's multi-firm review of retail banks' treatment of vulnerable customers cited the absence of cross-channel context propagation as a recurring weakness.

Worked example 2

A wealth platform analysed digital engagement data and found that customers spending more than three times the average session duration on a drawdown decision tool without progressing had a four-fold higher rate of subsequent complaints about retirement decisions. The platform introduced real-time behavioural monitoring: when session duration crosses a threshold without progression, the journey offers a callback to a retirement specialist and flags the case for a trained handler. The flag persists through the customer record, so every subsequent touchpoint — quarterly review, withdrawal request, advice meeting — is handled with awareness of the customer's likely circumstances. Complaint rates from the cohort fell by half and take-up of structured retirement guidance rose. The FCA's research on digital engagement practices in investment outcomes describes exactly this kind of behavioural detection and response design.

Watch-Outs

The most common failure mode is building detection capability that identifies vulnerability but does not change what happens next. Detection without response erodes trust faster than no detection at all — the customer disclosed something personal and nothing visibly changed. A second is over-engineering identification to the point where it feels intrusive: customers who feel surveilled rather than supported will disengage and disclose less. A third is propagating the flag without propagating the response: knowing a customer is vulnerable means nothing if every touchpoint still follows the standard process.

Evidence & AI Lens

E158% of vulnerable customers hide their circumstances due to trust issues and fear of worse treatment
E2The FCA expects firms to anticipate vulnerability, not just respond when customers explicitly declare it
E316% of vulnerable customers report firms are too slow to respond, versus 11% of the general population
E4Only 57% of vulnerable customers who do disclose feel their firm actually cares about their circumstances

AI & Automation Lens

Natural language processing applied to call transcripts and chat logs can identify vulnerability signals in real time, alerting agents or adjusting the digital journey before the customer reaches a breaking point. Machine learning models can identify behavioural patterns associated with vulnerability across digital journeys — banking, wealth, or insurance — enabling proactive intervention such as a callback offer, a simplified pathway, or a pause in automated processing before the customer disengages. Critically, AI can power the propagation layer: automatically tagging the customer record so that downstream systems respond consistently without manual handoff.

The problem, in one line

The move

First signal it's working