Research & products

Curiosity with a practical purpose.

Our long-term ambition is to bring business expertise into AI research: exploring foundation models, developing useful products of our own and contributing original model capabilities over time.

Research agenda

Questions worth investigating.

These themes guide the research we intend to develop alongside client services and engineering. They are an agenda, rather than claims of completed findings.

Models & specialist knowledge

How can AI understand business work?

Explore how foundation models interpret financial, procurement and operational information, and where retrieval, task design or model adaptation improves results.

Evidence & reliability

Can the result be checked?

Investigate source references, calculation checks, uncertainty and evaluation methods that help people inspect generated outputs.

Agents & human judgement

When should a system ask for help?

Explore bounded tool use, exception handling, recovery from errors and the effort required for meaningful human review.

Original capability

What should we develop ourselves?

Identify opportunities for reusable software, domain-specific model capabilities and, over time, research towards new foundation models.

Demonstration prototype

Finance Operations Lab

A demonstration of how people and AI can share a finance workflow. Defined tasks prepare evidence and proposals; a human reviewer approves, returns or holds the work.

01

Prepare

A defined task gathers evidence and proposes an action.

02

Review

A person inspects the evidence and decides what happens next.

03

Record

The decision stays visible in the demonstration history.

The Lab uses illustrative data and a simulated accounting connection. It does not change a live ledger.

Ask about a demonstration

A longer horizon

From useful experiments to original products.

Our ambition includes products built with existing foundation models and research towards models of our own. Each step needs a clear purpose, appropriate data, technical evaluation and practical feedback.

Explore

Research questions

Define a task, a baseline and a way to measure improvement.

Develop

Experiments & prototypes

Compare approaches, document limitations and test ideas with people.

Progress

Products & model capabilities

Develop promising work into supported products and deeper research as evidence and resources allow.

Use of client information for research or model training requires a separate agreement. Access for service delivery does not imply permission for research.

A practical first conversation

A business question. A research possibility.

Interested in a prototype, a shared research question or a potential collaboration? Tell us what you would like to explore.