Skip to content
Neurally

AI in a regulated business.Built to survive the board, the regulator and the customer.

In financial services and insurance, the question is rarely whether AI can do the task. It is whether you can show who is accountable, that customers get good outcomes, and that the numbers are real.

Book a free discovery call (opens in a new tab) Where to start, and what it costs

If you lead a regulated firm, this may sound familiar.

  • “Every team has a pilot. None of them has been through risk and compliance.”
  • “The board wants an AI strategy. The risk committee wants to know who signs it off.”
  • “We could automate that decision tomorrow. We couldn't explain it to the FCA.”
  • “The vendor demo looked great. Nobody asked what happens when it's wrong.”

Why regulated firms get stuck

A regulated firm can usually find a use for AI in a week. Getting it live takes longer, because three audiences have to be satisfied at once: a board that wants a return, a risk function that wants accountability, and customers who are owed a good outcome.

Most programmes satisfy them in sequence. The business case is written, then the build starts, then compliance is asked for a view. Each step reopens the one before.

We work on all three together. The value case, the redesigned workflow and the accountability for each decision are produced as one piece of work, so the plan that reaches your board has already met the objections it will face.

The decisions that matter here.

Four choices shape almost everything else. Making them early is what keeps a regulated AI programme from being redesigned twice.

  1. 01

    Where AI may touch a customer outcome

    Pricing, eligibility, claims, complaints and anything close to advice carry Consumer Duty weight. Decide early which of them AI may touch, and on what terms.

  2. 02

    Which decisions stay with a named person

    For each workflow, set out what stays with a person, what AI assists and what AI takes on within agreed limits, with a way to review and reverse every decision.

  3. 03

    What the value really is

    Faster handling creates capacity. It is a cost saving only when spending falls. Boards in regulated firms tend to ask, so the business case keeps the two apart.

  4. 04

    What to build, buy or leave alone

    We sell no software, so advice on vendors and partners has nothing behind it. Third-party processing and data retention are covered in scoping.

What tends to go wrong.

  • Compliance arrives at the end

    A pilot that works gets redesigned, or shelved, once conduct risk is assessed. Testing the design against Consumer Duty outcomes first is far cheaper.

  • Accountability is assumed

    When AI is involved in a decision and nobody is named as accountable for it, the gap is found by a customer or a regulator.

  • The case rests on time saved

    Hours saved that never become lower spending or better outcomes do not survive a second board meeting.

How the four stages apply in a regulated firm.

  1. Stage 01

    Decide

    Opportunities are sized with revenue, cost, capacity and customer outcomes kept separate, and with conduct risk set beside the value case.

  2. Stage 02

    Design

    Workflows, roles and decision rights are redesigned, then tested against Consumer Duty outcomes before they reach a board.

  3. Stage 03

    Deliver

    Small, time-boxed pilots with acceptance criteria agreed in advance. Each ends in proceed, adapt or stop.

  4. Stage 04

    Embed

    A named owner for every result, training on the redesigned work, and tracking your own team runs.

How the four stages work

Where this comes from.

First
ChatGPT app for insurance, alongside the company's first consumer AI assistant, launched within FCA and Consumer Duty rulesCompare the Market, before Neurally
500,000
users added by an Open Banking credit product, with 60% monthly retentionClearScore, before Neurally

These are outcomes from roles Stuart held before founding Neurally. They are not Neurally client results, and the wider commercial figures reflect the work of whole teams, not AI alone.

Start with a plan your board and your risk committee can both approve.

The Deep Dive covers the whole business or several functions in two weeks. It ends with an operating-model outline, a board-ready 90-day mobilisation plan with owners and decision gates, and an AI governance and risk view. If you would rather test the water first, the one-day Workshop does that.

Talk it through first, free (opens in a new tab) Full scope on the services page

Engagement 03

AI Transformation Deep Dive

£12,5002 weeks

  • An operating-model outline: roles, teams, workflows and decision rights
  • The boundaries between people, AI assistance and automation
  • A board-ready 90-day mobilisation plan with owners and decision gates
  • An AI governance and risk view

Also worth a look: AI Opportunity Workshop (£2,500, 1 day).

Questions we hear.From this sector.

Do you give regulatory or legal advice?

No. We design products, workflows and operating models with Consumer Duty, conduct risk, UK GDPR and the EU AI Act in view, and work alongside your compliance and legal teams. Formal regulatory and legal advice stays with them and with your counsel.

Have you done this inside a regulated firm?

Stuart has, as an employee. At Compare the Market he wrote the consumer AI strategy, won board backing and launched the company's first consumer AI assistant and the first ChatGPT app for insurance, within an FCA and Consumer Duty environment. That is career experience from before Neurally, not a Neurally client result.

How do you treat Consumer Duty in practice?

Opportunities and designs are tested against Consumer Duty outcomes and conduct risk before they reach a board. Every workflow redesign names who is accountable when AI is involved, what stays with a person, and how a decision can be reviewed and reversed.

We already have a model risk framework. Does this replace it?

No. The work fits inside the governance you already have. Where that governance has a gap for AI, such as accountability for an AI-assisted decision, the gap is named and an owner proposed.

Related reading

Start with a conversation.Thirty minutes, free.

Tell us what you are trying to decide. If we can help, we will say how and what it would cost. If we can't, we will say that too.

Book a free discovery call (opens in a new tab) Or send a message