This article is brought to you in collaboration with Gregory.
AI is rapidly reshaping the competitive landscape in banking, and for many institutions, the real challenge lies not in experimentation, but in implementation. Richard Davies, CEO of Allica Bank, has been focused on exactly that: how to successfully deploy AI across an organization and drive meaningful adoption at scale.
Founded in 2020, Allica is a digital bank focused on established small and medium-sized businesses. To date, it has lent over £3 billion and been twice named by Deloitte as the UK’s fastest growing technology company. In 2025 the Financial Times identified it as the second fastest growing company in Europe.
Richard delivered a fascinating keynote address at FinovateEurope, titled: “Successfully Implementing AI & Scaling Adoption: What Are the Challenges Around Rolling Out to Production?”. Afterwards, we sat down with him to talk about what it really takes to embed AI into a bank’s operating model.
Tell us a little more about your role as CEO of Allica Bank and what you’re focused on at the moment?
Richard Davies: Allica is a fintech bank focused on established small and medium-sized businesses. We typically define that as businesses with five or more employees or at least £500,000 in revenue. So we’re not talking about the very smallest microbusinesses, but those that are at a point where things start to get more complex and there are multiple staff to support.
We find these businesses fall into a gap between the corporate banking divisions and retail banking divisions of the major banks. That’s the space we focus on.
We have been building Allica for five or six years now and provide a full stack of services, including current accounts, cards and all types of lending. Increasingly, we are moving into financial operations areas such as spend management and cash flow forecasting. Alongside that, we have been thinking hard about how we can apply AI to power many elements of what we do across the organisation.
In your keynote, you spoke about successfully implementing AI and scaling adoption. What do you see as the biggest challenges for banks when it comes to rolling AI out in practice?
Davies: I would group it into three main areas:
First, ensuring that AI adoption happens across the whole company, rather than sitting in an innovation lab or small specialist team. A big focus for us has been getting people bought in, upskilled and confident, and encouraging teams to create their own simple, agentic use cases. I am a big believer that bottom-up adoption tends to win over purely top-down mandates.
Second is software engineering and product development. Around a third of our staff are in engineering, and that is probably the area that has seen the greatest progress in AI tooling. We have focused on helping people move towards more T shaped or full stack roles, and ensuring our tech stack is AI enabled to unlock significant productivity gains. Depending on what you measure, we are seeing productivity improvements of two to ten times.
Finally, there are more complex agentic use cases. We have specialized teams working on these, and we have been learning a lot over the past two years about what it takes to get them live in production. It’s exciting because beyond engineering, you start solving real world problems that consume a lot of human time and can be inconsistent when done manually.
A lot of banks are investing in AI at the moment. How should they decide where it makes the most sense to focus first?
Davies: My view is that you should not overly narrow your focus. If you pick two areas, you are neglecting ten others, and those areas will fall behind.
Perhaps I have the luxury of leading a fintech organization that is naturally inclined towards this, but I think AI needs to be embraced across the company. Where you do need focus is on infrastructure, including data quality, enabling access to different AI models and ensuring that is done company wide.
If I had to pick one area with immediate and certain benefit, it would be engineering. The productivity unlock in software development is huge. If teams are still working in traditional ways, they need to move quickly, not just for the company’s benefit, but for their own careers. The industry is shifting rapidly, and people need the skills and experience to keep up.
Beyond the technology itself, what changes do banks need to make internally for AI to really become part of how they operate?
Davies: Culture is a big part of it. People need to lean into it. You need the infrastructure in place, as well as training and upskilling so people feel confident using AI.
At the same time, organizations need to remain risk aware. Different AI use cases carry different risks, and teams need to understand those.
In many ways, it’s similar to previous organizational transformations, such as moving from traditional waterfall practices to agile. The enablers are not conceptually different, but it does require deliberate leadership and a clear view of how you enable the organization to change.
From what you’ve seen at FinovateEurope so far, what themes or conversations around AI in banking have stood out to you the most?
Davies: Some of the most interesting conversations have been happening off stage. Recently, we have seen software company valuations come under pressure following major AI model releases, with the view that people can now build their own software more easily.
At the same time, traditional banks have re-rated quite significantly over the past year. In the UK, share prices are up roughly 80 percent. It creates an interesting dynamic.
Fintech has at times in the past been viewed by investors as a poor relation to software, but in reality, building a fintech is much harder than building a pure software company. You have complex regulatory requirements and balance sheet considerations that software firms do not.
It feels like there may be a shift happening in the relative valuation of where companies with real assets versus asset light software companies. For many fintechs, particularly those with strong fundamentals, that could ultimately be a net positive.
Photo by Google DeepMind
Views: 8
