PE fund, partner-led
We put AI agents into production across a portfolio
A partner-led PE fund wanted to turn scattered AI ideas from its portfolio companies into actionable initiatives with shared operational responsibilities.
| Industry | Financial Services |
|---|---|
| Services | AI Transformation, AI Agents |
Portfolio companies submitted hard-to-compare ideas
The fund had proposals from sales, customer service and administration across its portfolio companies. Some teams had already tested tools. Others described their wishes in presentations. The operating partner could barely tell which initiatives would reduce recurring work and which first needed foundational work.
The portfolio companies used different systems and had their own approaches to approvals. A central solution for everyone would have required extensive integrations. At the same time, each company risked answering the same questions about data access, quality and support from scratch. The fund lacked a shared basis for assessment.
Every initiative needed a named process owner
Our task was to assess opportunities for AI across the portfolio against real workflows and support suitable agents through to production. The fund set the selection framework. Management at each company had to confirm who owned the affected process and what capacity was available for data maintenance, testing and ongoing support.
We developed agents and operating rules together
Process mapping made effort visible
We described inputs, processing steps, exceptions and handovers. The assessment connected expected workload reduction with data access, integration effort and responsibility.
A service agent used approved knowledge
It answered recurring requests and handed unclear cases to the portfolio company’s customer service team with sources and processing status.
A proposal agent prepared drafts
It combined relevant service components and project knowledge. Sales reviewed scope, assumptions and terms before sharing the proposal.
The operating model governed changes
A shared template described business acceptance, technical support, incident reporting and shutdown. Each company added its contacts and specific system boundaries.
The fund prioritized with the business teams
We discussed candidates with the operating partner, management teams and process owners. With the business teams, we reviewed completed cases to check whether the necessary knowledge was available. We deferred initiatives with unresolved data rights or missing support, even when the initial demonstration looked convincing.
Implementation ran in short working cycles with the future users. Customer service checked answers against past inquiries. Sales compared proposal drafts with accepted service descriptions. Joint reviews focused on specific failures and the changes needed before approval.
We standardized test records and responsibilities while keeping integrations at the individual companies. This required additional adaptation for each system. In return, data access and business approvals remained under the control of the people responsible for the processes. Across the portfolio, we shared methods and sanitized test examples without collecting customer documents centrally.
Portfolio companies adopted agents in daily operations
The agents became part of the respective workflows. Users could report errors, owners could approve changes and technical support staff could monitor operations. The fund gained a comparable view of ongoing initiatives and their outstanding prerequisites. Further ideas remained under assessment until data access and process ownership were clear.
Shared rules support decentralized implementation
Compare workflows before tools
A documented process shows which work disappears and which checks remain with the business team.
Assign support before building
An agent in production needs someone to maintain knowledge sources and handle errors reported by the business team.
Share proven testing procedures
Portfolio companies can use the same acceptance criteria while keeping customer data and integrations separate.
A first conversation takes 30 minutes.
We discuss which AI initiatives in your portfolio already meet the prerequisites for production use.