ChatGPT for financial advisers: direct use or purpose-built apps

Plenty of advisers and paraplanners already have a chat window open next to the back office. They use ChatGPT or Claude to tidy an email, explain a concept in plainer words or sketch the outline of a report. The obvious next question is whether that is enough, or whether a firm needs software built for advice work. ChatGPT for financial advisers can be a genuinely useful tool, provided the firm is clear about which jobs it suits and has done the data protection work first.

This article is a fair look at both routes. We should declare an interest: IFA Tools apps run on the firm's own Claude or ChatGPT account, so we think these models are very capable. The question is what surrounds them.

What a general chat assistant does well

A general assistant is flexible. You can ask it almost anything, and it responds in seconds. In advice firms, the jobs where it tends to help most are the ones where the adviser supplies the substance and wants help with the words:

• rewriting a draft email or letter so it is shorter and clearer;
• explaining a technical idea in plain English for a client who is not a specialist;
• producing a first outline for a presentation, a newsletter or an internal process note;
• summarising a long, non-confidential document such as a published consultation paper;
• brainstorming questions to ask in a meeting about a particular topic.

None of those needs client data, which is a useful test. Where a task can be done without personal information, a general assistant is often the quickest tool for it.

Where the chat window gets harder

Things change when the task is a regulated document built from client information. A suitability report, an annual review or a file check has a required structure, a firm house style, figures that must be right, and a record that someone may need to inspect later. A chat window can help with each of those, but the adviser has to supply the structure every time.

The work around the prompt

Doing a full annual review through a chat window typically means: finding the right valuation PDFs, deciding which pages to paste or upload, writing a long prompt that explains the firm's format, checking every figure against the source by hand, copying the result into the firm's template, applying the branding and saving a copy somewhere sensible. The model may do the writing well. The surrounding steps still take time, and they vary from one adviser to the next.

Consistency and evidence

Two advisers asking slightly different questions will get differently structured answers. For a compliance function, that makes outputs harder to review. Under SYSC 9.1.1R, "A firm (other than a common platform firm) must arrange for orderly records to be kept of its business and internal organisation… sufficient to enable the FCA to monitor the firm's compliance". A chat history is a record of sorts, but it is rarely organised around clients and files.

What purpose-built apps add

A purpose-built app wraps the model in the steps the job needs. Using our own apps as the example, because we can describe them accurately:

• Carbon turns a Fireflies meeting into a discovery document, scope of work, client value summary or attitude to risk report, in the firm's house style and word list, with branding applied, and drafts the follow-up email with the document attached.
• Strata reads the platform valuation PDFs, separates investment return from money paid in, tracks charges year on year and traces every figure to the report and page it came from.
• Scrutineer reads a whole client folder and ties each finding to a named FCA rule, quoting the passage from the file.

By our estimate, drafting one client document drops from about 2 hours by hand to about 15 minutes with Carbon. Much of that saving comes from the steps around the prompt, not the writing itself.

Other vendors build purpose-built tools too, mostly as hosted platforms. We compare several of them fairly in AI meeting assistants for UK financial advisers.

A quick comparison

What the plans cost

Anthropic publishes its Claude prices in US dollars. On claude.com/pricing, seen on 7 October 2026, Claude Pro was $20 monthly ($17 a month billed annually) and Team Standard was $25 per seat monthly ($20 billed annually). OpenAI publishes its own ChatGPT plans; check its current pricing page for the plan you are considering, as we have not confirmed those figures here. All prices may change.

IFA Tools apps are priced per adviser per month, excluding VAT: Scrutineer £99, Strata £89, Carbon £69, Reckoner £69, Asclto £59 and Freeboard £49, or £299 for all six. IFA Tools users also pay separately for their own Claude or ChatGPT account. That is a genuine extra cost, and we set out why we chose that design in why we use your own Claude or ChatGPT account.

Data protection before you paste anything

Whichever route you take, the same rules apply. The ICO's (now the Information Commission's) guidance on AI and data protection says that "In the vast majority of cases, the use of AI will involve a type of processing likely to result in a high risk… and will therefore trigger the legal requirement for you to undertake a DPIA." It also says "it is your responsibility to decide which lawful basis applies". Firms must decide and document whether an AI provider acts as a controller or a processor.

Before any client information goes into a chat window, check:

1. Whose account it is. Personal accounts used for work are hard to supervise. Use a firm account the firm controls.
2. The training setting. Anthropic's pricing page says its Team plan has "No model training on your content by default", and that model training is opt-out on consumer plans. Check and record the setting for whichever provider and plan you use.
3. The DPIA. Write it before you start, not after.
4. What staff may paste. A short written policy saves a lot of uncertainty.
5. Supervision. The FCA has said it will rely on existing frameworks, including the Consumer Duty and SM&CR, so outputs need the same review as any other client communication.

Our guide to AI data protection for advice firms goes through these steps in more detail.

Frequently asked questions

Can financial advisers use ChatGPT with client data?

The FCA has no AI-specific rules, but it relies on the Consumer Duty, SM&CR and SYSC, and the firm must still meet its data protection and supervision duties. That means a DPIA, a documented lawful basis, a firm-controlled account with the right training setting, and review of anything that reaches a client.

Is Claude or ChatGPT better for advice work?

Both are capable, and preferences differ. IFA Tools apps work with either, chosen in setup, so a firm can use the provider it already trusts.

Do purpose-built apps use different AI models?

Some vendors build their own; Aveni, for example, says it has built its own model for UK financial services, FinLLM. IFA Tools apps use the firm's own Claude or ChatGPT account.

To see the difference between a chat window and a finished client document, look at Carbon or book a demo: twenty minutes on Teams with Adam Booth, on a demo firm, with nothing to prepare.