You’ve almost certainly used one by now, even if you didn’t call it that. You typed a question into ChatGPT, asked Copilot to draft an email, or watched a colleague get AI to summarise a long document. Each time, you were using a large language model.
This guide explains what that means in plain terms, and what it does and doesn’t make sense to use one for if you run a financial advice firm.
What is a large language model?
A large language model (LLM) is an AI system trained on an enormous amount of text. Put simply, training means the system learns statistical patterns across billions of words: what tends to follow what, how ideas connect, how questions are usually answered. The result can write fluent, relevant text in response to a prompt.
It isn’t thinking, and it isn’t reasoning the way a person reasons. It’s doing something much narrower: predicting what a useful continuation of a given input looks like, based on the patterns it has absorbed. That distinction matters, and I’ll come back to it.
The “large” refers to scale. Modern LLMs contain hundreds of billions of parameters (the numbers that store the patterns learned in training), and they were trained on text from books, websites, academic papers, code and other sources running into the trillions of words.
What can an LLM do?
In practice, an LLM does five things well:
- Drafting. Given a brief, it can produce a first draft of an email, a letter, a report, a policy summary or almost any other document. It’ll usually need editing, but it gives you something to work from rather than a blank page.
- Summarising. Give it a long document and ask for a summary, and it’ll produce one quickly. In an advice firm this is one of the most reliably useful jobs it does.
- Answering questions about documents. When you limit an LLM to a specific set of documents, rather than letting it draw on its general training, it can answer questions about them with reasonable accuracy and much less invention.
- Reformatting and restructuring. Turning a table into a narrative, a set of bullet points into prose, or technical language into plain English.
- Generating options. Subject lines, alternative phrasings, agenda items or frameworks. The output often isn’t perfect, but it gets you unstuck.
What an LLM cannot reliably do
The limits matter at least as much as the list above.
An LLM doesn’t know what it doesn’t know. When it lacks the information to answer accurately, it may produce a plausible answer that’s wrong. This is called hallucination, and it’s a structural property of these systems, not a bug that will be patched away. For most consumer uses it’s a manageable nuisance. For a regulated firm producing suitability letters, file notes or compliance documents, it’s a material risk if a person doesn’t review the output before it goes anywhere near a client.
An LLM also has no current knowledge unless you give it some. Most models have a training cut-off, and they don’t know what’s happened since. Ask one about a regulatory change from last month without giving it the source text, and you may get an answer that’s out of date or invented.
And an LLM isn’t a compliance engine. It can help you draft, summarise and structure. It can’t tell you whether a recommendation is suitable for a particular client. That judgement belongs to a qualified professional, and nothing in this guide changes that.
The smart intern
A common way to think about an LLM is as a smart intern: sharp, keen, fast and capable, but knowing nothing about your business on day one. You wouldn’t let a new intern sign off a suitability letter unsupervised. You also wouldn’t refuse to give them any work because they weren’t a chartered planner. You’d brief them carefully, check their work, and give them more responsibility as you learned what they were good at.
That’s roughly the right relationship to have with an LLM. Use it where it saves you time or improves the result, and keep a person in the loop on anything regulated or client-facing.
An LLM is a capable first drafter and a useful summariser. It isn’t a decision-maker, and treating it as one in a regulated context is where firms get into trouble.
Where LLMs are practically useful for an advice firm
Most of the useful jobs in an advice firm are text-heavy and repeated. Five stand out.
Drafting client communications
An LLM can take bullet points about a client’s situation and produce a draft communication quickly. The adviser edits, personalises and approves. The time saved is between “I know what I need to say” and “there’s a document I can work with”.
Summarising long documents
Annual reports, fund factsheets, regulatory consultation papers. An LLM can turn a forty-page document into a one-page summary in under a minute. The summary needs checking, but for a first pass it saves a lot of time.
Answering questions about your own documents
This is the use that matters most to regulated firms right now. When you limit an LLM to a set of documents you control, such as your process documents, your file notes or a client’s policy terms, and ask it questions only about those, it invents far less. Google’s own NotebookLM documentation claims a hallucination rate below 1% when the model is restricted to uploaded documents. That’s a vendor figure rather than an independent test, but the direction is consistent: the AI is no longer drawing on its general training, it’s reading your material. The realistic limits guide covers what still goes wrong even then.
Internal knowledge bases
You can build a searchable, conversational front end to your internal procedures, templates and process documents. Instead of a new hire spending an hour hunting through a shared drive, they ask a question and get an answer with a link to the right document.
Preparing for meetings
Give an LLM a client’s file summary and ask it to suggest questions worth exploring, or gaps in the information gathered so far.
What this means for how you route work through AI
Start by being explicit about which tasks you use AI for, and what the human review step is for each.
- Map where text-heavy work happens in your firm. Where do advisers or admin staff spend time drafting, reformatting or summarising? Those are the candidate tasks.
- Sort them by risk. A first draft of a client newsletter is low risk. A first draft of a suitability letter that goes to compliance review before anyone sees it is medium risk. A final suitability letter going straight to a client with no further review is a task an LLM shouldn’t be doing on its own, ever.
- Limit the AI to your own material where accuracy matters. A setup that answers from documents you own and control (your process notes, file records and templates) suits regulated use better than relying on the model’s general training. The hallucination risk is lower, and the outputs can be audited.
- Document what you’ve deployed. The FCA and Bank of England issued a joint statement in 2025 signalling that AI governance for UK-regulated firms is moving from voluntary best practice towards regulatory expectation.[1] Knowing which AI tools your firm uses, for what, and what the human oversight is, is the baseline. If you can’t answer those questions now, start there.
A note on the technology moving quickly
The tools built on top of LLMs change fast, so be careful what you come to depend on. Some niche tools that were useful twelve months ago have been absorbed into bigger platforms, and some standalone tools being heavily marketed to professional firms now may not exist in their current form in two years. That isn’t a reason to avoid the technology. It’s a reason to choose AI tools with the same care as any other critical supplier.
The underlying capability (drafting, summarising, answering questions about documents) is real and useful. Which tool you use to get at it is a separate question, and one worth revisiting every so often rather than deciding once.
What you can do this week
You don’t need a formal AI strategy to start using these tools sensibly. You need one person to spend an afternoon experimenting, and then share what they found.
Pick one text-heavy task that happens often in your firm: drafting an email, summarising a document, or pulling key points from a piece of regulation. Run it through one of the mainstream tools (ChatGPT, Claude, or Microsoft Copilot if you’re on Microsoft 365). Judge the output honestly against the time it took.
That’s enough to start forming a view about where AI helps in your firm, and where it’s more interesting than useful.
This article is for informational purposes only and does not constitute regulated financial advice or a compliance opinion. Consult a qualified compliance professional for advice specific to your firm.
Sources
[1] Bank of England and Financial Conduct Authority, AI and machine learning in UK financial services: joint statement (2025). Available at https://www.fca.org.uk.