By Dimitri Limata, COO at MoiraCorp. Most UK advisory firms are buying AI tools before they understand how AI actually works. Five resources every leader should read first.
Most UK advisory firms I speak to are buying AI tools before they understand how AI actually works. That is the wrong order — and with the FCA’s expectations on AI evolving faster than most boards realise, getting the foundations right is no longer optional. It is the baseline.
Before you spend anything on AI tooling, start with the five resources below. They are free, widely respected in the field, and each addresses a different gap most firms have when they come to the AI conversation.
What it is: A practical guide from OpenAI on how to structure prompts so that you get consistent, useful outputs from large language models.
Why it matters: The single biggest determinant of whether an AI tool produces useful work in your firm is whether your team knows how to prompt it. Most do not, because nobody trained them.
Where to use it: Client emails, meeting summaries, report drafting, internal memo first drafts. → Read it here
What it is: A library of free, two-to-three-hour courses from Andrew Ng’s DeepLearning.AI that give a clear, practical introduction to how AI behaves under the hood.
Why it matters: You do not need to become a machine learning engineer. You do need to know enough about how these systems work to make sensible procurement decisions and to spot vendors who are overselling.
Where to use it: Upskilling non-technical teams quickly. The Generative AI for Everyone course is the right starting point for most adviser firm staff. → Browse the catalogue
What it is: A structured framework from Microsoft for implementing AI across a business — covering governance, integration, scaling, and the operational shifts adoption requires.
Why it matters: AI adoption is an organisational change problem, not just a tool selection problem. The framework gives you a map of the territory before you start moving through it.
Where to use it: Planning the operations, integration, and scaling stages of any AI programme. → Read the framework
What it is: The FCA’s public statement of where its thinking sits on AI use in financial services — covering Consumer Duty, SM&CR, operational resilience, and the regulatory expectations firms should be building towards.
Why it matters: This is the regulator’s view in their own words. If you are running a regulated firm and you have not read DP23/4, you are exposed. There is no AI policy your board can write that is more important than knowing what the FCA expects.
Where to use it: Compliance, governance, board-level readiness conversations. → Read DP23/4
What it is: The National Institute of Standards and Technology’s framework for managing risk, controls, and accountability in AI systems — the most widely-cited risk framework in the industry.
Why it matters: When the FCA or the ICO asks how your firm is managing AI risk, “we are following the NIST framework” is a defensible answer. “We have not really thought about it” is not.
Where to use it: Building AI safely and defensibly. Particularly important if you are deploying AI in workflows that touch client decisions, advice, or personal data. → Read the framework
Prompting is useful. Knowing how to structure a prompt to get a useful output from ChatGPT or Copilot will save your team time and reduce frustration. But prompting is not where firm-level AI value actually comes from.
The value comes from changing how the work flows. From mapping which tasks could be done differently, which workflows could be reordered, which decisions could be supported with better information at the moment they get made.
AI does not fix broken processes. It scales them.
The firms getting this right are not the ones buying more tools. They are the ones fixing workflows, data flow, and operational clarity first — and then introducing AI into specific places where it will pay back. The tool is the last decision, not the first one.
This is the foundation we build on with every client engagement. Before we recommend any AI tool, we map the operational layer underneath. Where is the time going? Where is the friction sitting? Which workflows are ready for automation, and which ones need to be redesigned first?
Only once that map exists does the AI conversation become precise — which specific workflows, which specific tools, which specific cost-benefit profile makes sense for this firm in its specific regulatory and operational context.
If your firm is navigating AI adoption and you want to get the order right — foundations first, tools second — that is the conversation we exist to have.
