Sponsored content: Vytenis Sakalas, chief executive and co-founder at Moterra, discusses how insurance companies can best approach implementing AI that their staff find useful
Most broking teams I speak to are already using AI. Account handlers summarise renewal packs, placement teams tidy market submissions and claims staff draft updates with it.
In insurance the tool people settle on is often Claude, because it copes with long policy wordings, awkward spreadsheets and routine work.
The Bank of England and FCA survey published in October found that 86% of UK financial services firms now use AI, up from 75% in 2024.
But sooner or later, someone will ask where client data goes. For an MGA, it may be a capacity provider during a delegated authority audit. For a broker, a commercial client’s procurement team, a PI insurer at renewal or the FCA.
The second question, usually from legal, is what happens if the supplier stops trading.
The Senior Managers and Certification Regime puts responsibility for AI with a named senior manager and under UK GDPR a prompt containing a policyholder’s details is personal data.
That means ’We have a policy’ is not an answer if the office does something else.
Why banning does not work
In my experience, blocking public AI tools lasts about a month before the use moves to personal phones.
The pattern is the same in most firms I see – someone tries Claude, gets a day’s document work done before lunch, then runs into IT or compliance.
Most companies that come to us ask one question – how do we keep the Claude our people already like, with the data protection and regulatory cover we need?
The answer is an approved tool that is better than the free one – it sees the firm’s own documents and systems and runs somewhere the firm controls.
A defensible set-up
A firm should be able to describe its AI use in four sentences:
- It runs inside our own cloud, in a UK region, and client data is not used to train anyone’s model.
- Access follows the permissions we already have, so a handler’s assistant sees only that handler’s clients.
- Every request is logged somewhere the supplier cannot edit, and a named person signs off anything that reaches a client.
- If the supplier disappears tomorrow, the tool keeps running, because the environment, the keys and the logs were ours all along.
This is the approach we take at Moterra, deploying Claude inside a firm’s own AWS or Azure account. The model itself does none of the second and third points – they come from the management layer around it, the part firms underestimate when they build it themselves.
What comes next
More professionals now open Claude first thing and keep it open – it reads the inbox, pulls the client record and drafts the submission, while they check, decide and sign off.
The email, CRM and policy systems become the record, but nobody spends their day in them. Anthropic has already released ready-made Claude tools for legal teams and insurance will have its own soon.
Ask your teams honestly which tools they use. Then decide where an approved tool should run, who sees what and whether you pay per seat or per use, before the first invoice.
For a firm of thirty people this is no longer a six-month IT project. Firms that do this now will have a short, confident answer when the question arrives.
- For more information about Moterra, click here
