A practical guide · for funds and boards

The AI operating partner role, explained.

Private equity created the operating partner to convert deal theses into EBITDA. The AI operating partner is the same seat for the assumption now written into almost every model: that AI will move margin, growth, or multiple during the hold. Search firms including Korn Ferry and Heidrick & Struggles have named it the newest portfolio value-creation role. This page is a working guide to the seat, from someone who does the work.

Why funds are creating the seat now

Because the assumption moved into underwriting. When AI was an experiment line-item, a vendor or a consultancy could carry it. Once it prices deals, someone inside the fund has to be accountable for it the way an operating partner is accountable for pricing or procurement: with a number, a baseline, and a hold-period deadline.

The three failure modes

• The narrator: sizes prizes beautifully, ships nothing. Eighteen months in, the thesis is still a slide.
• The technologist: ships systems that never touch the P&L because nobody sized where the value was or owned the adoption.
• The tourist: runs pilots for the annual meeting. Pilots that neither ship nor die are the most expensive outcome in this category.

What good looks like

One person who can do three jobs that are usually three hires: diagnose where the value is to investment-committee standard, build the systems personally rather than manage a vendor to a spec, and own the result like an operator so the capability outlives them. Each target portfolio company gets the same 90-day arc: two or three opportunities sized in dollar terms, one workflow shipped to production with an owner and evaluation gates, and a scoreboard the CFO signs.

Hiring criteria that separate operators from narrators

• A number they realized, not sized, with a baseline they can walk you to.
• A system they built themselves that still runs without them.
• A P&L they owned, where a bad call cost them their own money.
• Diligence reps: they can pressure-test the AI line in a model before you pay for it.
• Governance instincts: evaluation gates, kill criteria, and decision rights, not tool enthusiasm.

Common questions

Full-time or fractional?

Both models work. Fractional fits funds testing the seat across two or three portfolio companies; full-time fits a fund with an AI assumption in most of its models. The failure mode is not hours, it is authority: the seat needs a mandate to ship, whatever its size.

Where should the seat report?

To whoever owns value creation, not to IT. The moment the seat reports into technology it becomes a platform-selection function and stops moving the earnings bridge.

How is the seat measured?

The same way any operating partner is measured: realized EBITDA against the entry thesis, with baselines a CFO signed. Activity metrics (pilots launched, tools evaluated) are the tell that the seat is failing.

What does it cost to get wrong?

A mis-hire costs a year of hold period, which for most funds is worth far more than the salary. The common mis-hires are the technologist without P&L judgment and the strategist who cannot build.