Joseph Twigg

Nationwide is the UK’s largest building society and is currently testing Aveni’s FinLLM across a range of financial services compliance use cases, with plans to leverage the technology across various applications once testing is complete. 

Developed over twelve months by the specialist team at Aveni Labs, FinLLM is a financial services specific large language model family designed to meet the financial industry’s stringent requirements for compliance, safety, and performance. 

Nationwide’s position as first commercial adopter is the natural progression of a deeper partnership: the building society has been an investor in Aveni since 2024 and played an active role in co-creating the model, helping shape both its technical build and Aveni’s approach to governance and risk. Its decision to deploy FinLLM in production underscores its confidence in the technology and its commitment to responsible AI adoption.

Currently used within Aveni’s Assist and Detect products FinLLM integrates structured and unstructured data, adheres to ethical AI principles, and is governed by rigorous oversight, all aligned with FCA guidance and the EU AI Act. Extensive benchmarking demonstrates that FinLLM consistently outperforms general-purpose LLMs on financial tasks, setting a new benchmark for responsible generative AI in one of the world’s most regulated industries.

Sri Kanisapakkam, chief data officer at Nationwide, said: “Since investing in Aveni and working together on co-creating FinLLM, we are delighted to see how it has developed, and even more so to be the first organisation to deploy it.

“The industry-specific family of models approach that Aveni has taken is exactly what we need: the flexibility to apply the right model for the right task, at scale. We’re excited by the early performance we are seeing during testing and the potential benefits FinLLM will bring both Nationwide and our customers, as we continually look to deliver better service and experience through the responsible adoption of new technologies.”

Joseph Twigg, CEO of Aveni, commented: “Nationwide’s decision to be both an investor in and the first commercial customer of FinLLM is a powerful endorsement of what we have built. Their early engagement has been instrumental in shaping a model that truly reflects the needs of a major UK financial institution.

“The family of models approach we have taken means that, in an era where AI sovereignty is becoming increasingly important, FinLLM is a fantastic example of UK AI innovation, a highly performant model, with data transparency, AI safety and ethics at its heart, that will start to deliver real automation for a range of use cases across UK financial services.”

Unlike general-purpose models that approximate regulatory context, FinLLM models are trained on the specific rules, workflows, and supervisory expectations that apply to regulated firms.

The family of models approach allows organisations to select and deploy the right model for each use case and scale across their business as needs evolve. Every model is optimised for its specific use case, from agent assurance to compliance detection, and gets sharper with every deployment.