Embedded AI for an investment bank
Boutique investment bank, London · AI implementation
Since January 2026 an OpenKit AI lead has worked inside a London investment bank, building a CRM that updates itself from email and meetings, origination tooling that screens the market against live mandates and drafting support for deal material. Every piece runs inside the firm’s governance and nothing leaves the building without a person signing it off.
Overview
At this boutique investment bank in London, people in coverage, origination and execution all need to know who has seen a company, what was said, what remains outstanding and what an investor wants. That information arrives in conversations, emails, meeting recordings and attachments.
The information exists, but it sits in individual inboxes and notebooks, where it is useful to the rest of the firm only once someone records it. The people best placed to record it are the ones with the least time to do so.
Origination has a different problem. The market holds far more companies than a team can screen by hand, so finding and qualifying targets takes time away from talking to them.
The firm wanted AI built into the way it works and its team fully trained to use it. It brought in an OpenKit AI lead to work inside the team, learning how each function works and building with the people in it.
200,000+
Company records automatically screened against mandate criteria
Every inbox, every meeting
Absorbed into one shared record of companies, investors and deals
Since Jan 2026
An OpenKit AI lead working inside the team every week
Approach
What OpenKit built
The lead works alongside the team week to week. Each tool started from a specific person’s workflow and went live on the firm’s mandates. The lead refined it on their feedback and changed or removed anything that did not earn its place in the day.
A CRM that keeps itself current
Built with the coverage team, it matches every email, calendar invite and recorded meeting to the deal, company or investor it concerns. Deal stages move only when a quoted line from the source supports the change. Anything uncertain goes to a person to approve, edit or reject. Once rejected, a suggestion is not raised again.
Agents that follow up
Agents read meeting notes and emails as they arrive, record introductions, capture commitments made on calls and chase what remains outstanding. The whole team sees the same picture of each deal and each investor relationship without anyone typing it up.
Origination that arrives screened
Live searches run against each mandate’s criteria across company registries, financial data and the open web. Only the companies that fit reach the team, with the evidence for why. Each one comes with a cited narrative of the business and the right people to speak to. The tooling never contacts anyone on its own.
Drafting support for deal material
Reusable skills produce first drafts of information memoranda, information request lists, investor notes and buyer lists from the firm’s own material. Guardrails make sure the people accountable for each document review and edit the draft before it goes anywhere, so the judgement and the standard stay with the team.
One connected workspace
Claude is connected to the firm’s CRM, mailboxes, meeting transcripts and the data sources the team already licenses, so questions are answered from live firm data.
Governance first
Corporate finance runs on confidentiality, so OpenKit built the controls in from the start.
- Nothing sends automatically. Drafts are prepared on request and a person sends them.
- Every change an agent makes to the CRM carries its source and an audit trail.
- Usage is metered with ceilings and alerts, so spend stays predictable.
Result
A record the firm can trust
The CRM now holds a current picture the whole firm shares. A meeting is on the record shortly after it ends and partners can see where any relationship stands without asking around.
Everyone signs in with their own company account and access follows each person’s role. Claude reaches only the mail, attachments and records that person could already open.
Origination starts from a screened list
Every live mandate has its own list of companies that fit, so the team’s origination work begins with deciding who to approach and what to say.
The team builds too
The lead trained the team to build their own Claude skills and artifacts. An origination associate built a prototype tool for screening a new market, which OpenKit later rebuilt into the firm’s platform. Others on the team proposed tools that OpenKit then built.
An ongoing role
The firm has since increased the lead’s time and extended the role.
Other engagements
What could your business do with AI?
Find out with experts who become part of your team. We uncover opportunities, get ideas working, and help your people build on the results. We reply within one working day.
Get in touch AI Implementation