Why a Running Case?
In Level 1 you saw the difference between an AI-added product, where a chatbot is bolted onto an old workflow, and an AI-native product, which is designed around what AI makes newly possible. You can't make that distinction in the abstract. You need a real workflow, real people and real rules.
For the rest of this course, every design decision will be worked on one running case:
Case: Northway Advice, a fictional UK financial advice firm with 12 advisers serving retail clients on pensions, retirement income and investments. Northway wants to build an AI-native product for its own advice process. Later, it may sell that product to other firms.
Financial advice suits this course because the work involves a lot of documents, it is judged on outcomes, and it is closely regulated. If you work in law, insurance, healthcare or accounting, the same method carries over.
Step 1: Identify the User Roles
Start by naming everyone who touches the work, not only the person who clicks the buttons. Northway has four main roles:
- Client: a member of the public who wants to retire safely, combine old pensions, or invest a lump sum.
- Adviser: meets the client, gathers facts, judges suitability and makes the personal recommendation.
- Paraplanner: researches products, runs comparisons and drafts the suitability report for the adviser to approve.
- Compliance officer: checks files, samples advice for quality and answers to the regulator.
Do this: For your own product, list every role that creates, uses, approves or audits the work. A role you leave off the list is a role your design will fail.
Step 2: Describe the Job Each Role Needs Done
A job is the progress a person is trying to make. It is not the tool they use to make it. Each role has a different job and is judged on a different outcome.
| Role | Job to be done | Judged on |
|---|---|---|
| Client | Feel confident their money will last | Understanding the advice; a good long-term outcome |
| Adviser | Give advice that fits this client | Suitable advice, client retention, few complaints |
| Paraplanner | Turn facts into a clear, accurate report | Accuracy, turnaround time, rework rate |
| Compliance officer | Prove the firm's advice is sound | Clean file reviews, evidence ready for the regulator |
The jobs pull against each other. The paraplanner wants speed. The compliance officer wants evidence. The client wants plain language. A good design serves all three; a bad one pushes one role's problem onto another.
Do this: For each role, write one job and one measure of success.
Step 3: List the Constraints
In a regulated service, the rules shape the design as much as the user needs do. Northway works under these constraints:
- Consumer Duty. In force since 31 July 2023, the FCA's Consumer Duty requires firms to act to deliver good outcomes for retail customers. It sets out four outcomes: products and services, price and value, consumer understanding, and consumer support. Firms must also be able to evidence the outcomes customers actually get, including customers in vulnerable circumstances.
- Suitability. A personal recommendation must be suitable given the client's knowledge and experience, financial situation and objectives, including risk tolerance. In many cases the client must get a suitability report that explains why the recommendation is suitable.
- Record-keeping. The FCA sets minimum retention periods for suitability records: indefinitely for pension transfers, conversions and opt-outs; five years for life policies and personal or stakeholder pensions; three years in other cases. Anything the AI drafts or relies on may become part of that record.
- Accountability. The FCA has said it does not plan new AI-specific rules. It applies existing frameworks, such as the Consumer Duty and the Senior Managers and Certification Regime (SM&CR), to AI. A named senior manager stays accountable. "The model did it" is not a defence.
- Liability. The firm is responsible for the advice it gives, whatever tools produced it. A wrong figure in a report is the firm's error.
- Data protection. Client files hold sensitive personal and financial data. Under UK GDPR, as amended by the Data (Use and Access) Act 2025, people affected by significant decisions made solely by automated means must be told about them, can contest them, and can ask for a human to be involved.
Do this: Sort your constraints into three buckets: regulatory, liability and record-keeping. Then write down how each one would rule out a design. For example, "an AI that rewrites the report after sign-off breaks the record".
Step 4: Write Outcome-Based Job Statements
The most common mistake is to write a feature and call it a job. "Add an AI chat assistant to the client portal" is a feature. It says nothing about what success looks like.
Use this template:
When [situation], [role] needs to [outcome], measured by [metric], within [constraint].
Compare these two:
- Feature: "Use AI to generate suitability reports."
- Outcome: "When a fact-find is complete, the paraplanner needs to produce a suitability report the adviser can approve without rework, measured by time to approval and rework rate, within suitability and record-keeping rules."
The outcome version tells you who the product is for, what "done" means, how you will measure it and which rules it must respect. It also leaves room for an AI-native answer that nobody has imagined yet.
Do this: Rewrite any feature-style statement until it names a situation, a role, an outcome, a metric and a constraint.
Design Artefact: The Case Brief
Produce a one-page Case Brief containing:
- A role list (Step 1)
- A jobs-and-measures table (Step 2)
- A constraint list in three buckets (Step 3)
- Three outcome-based job statements (Step 4)
Checklist
- Every role that creates, approves or audits the work is listed
- Each role has a job and a measure of success
- Regulatory, liability and record-keeping constraints are named
- Each job statement names an outcome, not a feature
- Each job statement includes a metric and a constraint
Key Takeaways
- The running case is Northway Advice, a UK financial advice firm with clients, advisers, paraplanners and compliance officers.
- Each role has a different job and is judged on a different outcome, and these can conflict.
- The Consumer Duty, suitability rules, record retention, SM&CR accountability, liability and data protection all limit what you can design.
- Using AI does not move responsibility away from the firm.
- A good job statement names a situation, role, outcome, metric and constraint, never just a feature.