How Toluma built AI into the day-to-day, team by team.
- Client
- Toluma AS
- Sector
- Fund management
- Where
- Norway
- Format
- Six-week bootcamp
- Status
- In production
The situation
Toluma is a licensed Norwegian alternative investment fund manager. The licence brings the obligations of a much larger institution across its portfolio, CEO-office and accounting work, and a lean team carries all of it. Fund operations run on investor and transaction data, and a licensed manager answers for how that information is handled.
The team knows that work in depth. The opening was not to add another system, but to give each part of the team a way to put AI to work on the routine parts of its own day, without that data leaving the firm's control.
What we did
We ran a bootcamp with Toluma, sitting down with each team in turn — portfolio management, the CEO office and accounting — to show them how AI works on the tasks they do every day.
From there, and with their own domain knowledge, we built agents that fit each team's day-to-day. They run inside Toluma's own environment: investor and transaction data stays in the firm's tenant and is never used to train a model. The routine work moves faster, and the judgement stays with the people: every result is reviewed and signed off by the team before it reaches the board or an investor, as a licensed manager's work has to be.
What changed
- Agents in daily use. Portfolio management, the CEO office and accounting each have agents built around their own work.
- The team can build with AI. Toluma came out of the bootcamp putting AI to work on its own tasks.
- Investor data stays in the firm. The agents run in Toluma's own tenant. Investor and transaction data never leaves it and is never used to train a model.
- Judgement stays with people. Every result is reviewed and signed off by the team before it reaches the board or an investor, as the licence requires.
Other projects
