The revolutionary Edinburgh AI tool tackling the financial advice gap

I asked it to write this column based on the transcript of my interview with the company you’re about to read about and I marvelled at how quickly it produced something pretty readable.

Rest assured, this isn’t the result. It didn’t hit the mark. In any case, I am old school and will never surrender my copy-writing to a bot. But imagine you are a financial adviser and you are able to massively reduce the time-consuming administration needed to on-board a client.

That’s where Aveni Assist comes in, an AI tool devised by a fintech spun out of Edinburgh University’s School of Informatics. The five-year-old company behind it, Aveni, embodies a trifecta of competitive strengths in Scotland: a vibrant fintech sector, world-leading AI expertise and historic strength in financial services, all centred on Edinburgh.

At the heart of Aveni Assist’s pitch is productivity. A typical UK adviser can spend up to 15 hours on follow-up tasks after two client meetings, such as writing an investment report, administering a customer relationship management system and checking compliance issues.

Aveni Assist can generate reports and carry out compliance checks within minutes, even generating detailed client follow-up emails. Since launch in 2022, the tool is now being used by 150 financial services firms, including Succession Wealth (part of Aviva), Rathbones and Octopus Money.

Joseph Twigg, Aveni’s chief executive and one of its three co-founders, argues this could help tackle the UK’s “financial advice gap”, which the Financial Conduct Authority (FCA) estimates affects up to 12 million adults.

These are people who need advice but are not wealthy enough to make it commercially viable for advisers to serve them. “The highest cost in advice is the adviser’s time,” Twigg says. “Automating the inputs and outputs means they can act more like financial coaches, opening up services to the mass market.”

If Aveni Assist sounds like another one of these AI-based tools that’s set to revolutionise life as we know it, you’d be right. Key to its functionality is use of large language models (LLMs) which analyse large amounts of data to learn the relationships between language and context.

It all started in 2019, when one of the other Aveni co-founders Lexi Birch, currently a professor at the School of Informatics, was working on an EU-funded research project helping media monitoring teams at the BBC come up with a way to turn news from outlets in Russia and elsewhere into usable English-language material. The project developed a speech-to-text machine translation model that inspired the creation of the Aveni business and the two tools, Aveni Assist and a second product, Aveni Detect which helps “scrub” compliance calls for accuracy.

At the time, Twigg had just left the former Aberdeen Standard Life and had been thinking of how to improve efficiency in the investment business.

“Financial services is high value but known to be quite inefficient,” he says. “We created our own language models so we’ve had a pretty good idea of what works and what doesn’t work from an LLM perspective in financial services for a number of years.”

The use of generative AI in financial services has been growing. A report by UK Finance and Accenture published in January estimated that the level of investment into generative AI in financial services will hit 16 per cent of all technology investment this year, up from 12 per cent last year.

However, Aveni embodies the rise of something more sophisticated: so-called agentic AI, or the use of AI agents that perform functions autonomously, with less human intervention than generative AI – like ChatGPT and Gemini. Twigg and his colleagues, including co-founder Jamie Hunter – are now in the vanguard of this new frontier.

In May, the company launched the first iteration of “FinLLM” an LLM designed to drive the new era of agentic AI in partnership with Lloyds and Nationwide.

Much of the funding for this, billed by Aveni as the UK’s first LLM specifically for financial services, has come from £11.5 million raised last year from the two financial institutions, along with Puma Private Equity.

Twigg was widely quoted at the time as saying the financial services industry “does not need AI models that can quote Shakespeare”, and that it instead needed AI models that “deliver transparency, trust and above all correctness”.

What this means in practical terms is that FinLLM uses specific data sets, including drawing on the FCA’s rule book, to come up with solutions for financial institutions that go well beyond the basic summarisation of generative AI, such as compliance checks, identifying vulnerable customers, and the creation of complex investment reports. The big idea is that the future of agentic AI in financial services will be delivered via small, highly tuned models for specific tasks.

Yet there are two questions looming. One is whether regulators can keep up with the rapid pace of technological innovation like this. Aveni is one of 130 firms that have applied to join an FCA scheme launching this month known as the “Supercharged Sandbox”, which will allow firms to experiment with AI with regulatory support.

Another is whether the business – especially tax - environment in Scotland is sufficiently enabling to keep fintechs like Aveni here. Twigg says other countries in Europe have been wooing the company with offers to relocate.

“When you get to this sort of stage, you become incredibly mobile,” Twigg says. Scottish government, take note.