I’ve been on a lot of calls lately with private equity firms trying to figure out AI.

Not “should we use AI” calls.

That ship has sailed.

These are the harder conversations: where do we actually start, what’s worth paying for, and how do we get our teams to use this stuff consistently? Three calls in this week. Three very different firms.

And yet the same five themes kept surfacing.

I think they apply well beyond PE, to enterprise and to non-profits.

Build vs. Buy is the wrong question (until it isn’t)

Every firm is wrestling with whether to develop AI capabilities in-house or rely on external platforms.

One firm is actively evaluating a PE-specific AI product.

Another tried several expensive specialized platforms built for investment workflows and concluded that enterprise-grade Claude outperformed them at a fraction of the cost.

A third went broad with a consultant-led rollout and eventually pivoted toward fewer, deeper engagements.

Most PE-specific AI products are charging a premium for wrappers around the same underlying models. The differentiation is workflow design, not the model. That’s worth knowing before you sign an annual contract.

Point solutions have a short shelf life

One firm made this explicit: they shifted focus away from single-use AI tools toward automation platforms like Zapier and N8N after finding that point solutions delivered limited ROI.

Another firm identified back-office automation as their next big focus area. The same tools, different application. The pattern I keep seeing: firms buy a tool for one job, it works okay, and then it just… sits there.

Automation infrastructure, by contrast, compounds. You build one workflow and it unlocks five others.

The real prize is the investment workflow

When I ask PE firms what they most want AI to do, the answers are remarkably consistent.

• Deal intake: monitoring inbound emails, triaging opportunities, generating pass/meet recommendations.
• Investment papers: first drafts that are good enough that analysts spend time refining rather than starting from scratch.
• Quarterly reporting: pulling from portfolio data without three days of manual work.

They’re core to how a fund creates value. Which means AI ROI is measured in hours saved but, more importantly, in decision quality and speed.

Uneven adoption is the norm, not the exception

Every firm acknowledged the same thing: a handful of power users, a middle group who dabble, and people who are still treating AI as glorified Google search.

This is driving demand for structured training and internal champions programs, not just tool access.

Here’s what I’ve learned: tool access without workflow context produces tool abandonment.

The firms making progress aren’t the ones with the best subscriptions.

They’re the ones where people are stepping back, analyzing their work and then getting AI to work for them, how they already work.

Small cohorts, ongoing contact, real change

Across all three conversations, the engagement model that resonated was the same: small groups of three to four people, ongoing retainer structure, real-time support between sessions.

Not a one-day workshop. Not a license and a help doc.

This isn’t surprising. Behavior change doesn’t happen in a single session. It happens when someone is there when you get stuck on a Tuesday afternoon.

Alex

P.S. If you’re a GP, operating partner, or COO at a fund thinking through any of this, I’d love to hear where you’re stuck. Reply and tell me.