Free · 3 minutes · one portfolio company

The AI Pilot-to-P&L Scorecard

Six ways portfolio AI stalls before it reaches the P&L. Score one portfolio company 0 to 3 on each. The total tells you how close the board's AI mandate is to an actual number, and which failure mode to fix first.

Score a portfolio company See the Portco AI Diagnostic

95%

of enterprise generative-AI pilots produce no measurable P&L impact. MIT NANDA, 2025

of operating partners have seen a demonstrable AI premium in a completed transaction. Accordion / Wakefield, 2026

68%

of PE-backed CFOs told to prioritize AI do not know where to begin. Accordion, August 2025

Score the six failure modes

0 means the mode is not present. 3 means it is acute and costing you now. Score one portfolio company, not the fund.

The failure mode 0 not present · 3 acute

The License Trap

You pay for AI seats and cannot show usage or output that maps to a dollar.

The Strategy Shelf

There is an AI strategy deck. Months later nothing has shipped from it into the business.

The Demo Graveyard

Pilots demo well on curated data, then die before production and the real system of record.

The Workshop Certificate

The team completed AI training. Day-to-day behavior and output did not change.

Vendor Lock-in

The AI lives inside a tool you do not control. The prompts, the data and the leverage sit outside the company.

No Adoption Owner

No named owner, no incentives and no change plan for getting the workforce to use it.

0 / 18

6 modes left to score

Your score

0 / 18

Priority failure mode:

Your first fix, mode by mode

Ordered by your score. Start at the top. Each of these is a first move, not a program: it takes days, it produces a number, and it makes the next move obvious.

01The License Trap

Pull usage by seat for the last 90 days. Tie the renewal to three named workflows with a named owner each, and cut every seat that does not map to one. Redirect the recovered spend into building the one workflow that does map to a dollar.

02The Strategy Shelf

Pick one function, one owner and one 100-day number out of the deck. Ship something into that function's real work within 30 days, even if it is narrow. A narrow shipped thing beats a complete strategy, and it turns the deck into evidence the board can read.

03The Demo Graveyard

Move the pilot into the real system of record, with real data, real permissions and a named maintainer. Give it a weekly usage number before it gets a single new feature. If it cannot survive production data, kill it this quarter rather than next year.

04The Workshop Certificate

Replace certificates with a live registry of shipped workflows per team: what was automated, by whom, and what it saved. Make that registry the training record. Teams with nothing in it get help, and teams with something in it get the next problem.

05Vendor Lock-in

Own the prompts, the data and the integration layer in your own repository and your own accounts, and keep the model swappable behind one interface. Buy vendors for speed, never for the layer that holds your workflow logic. Put data export and prompt ownership into the contract at this renewal.

06No Adoption Owner

Name one accountable owner with the authority to change the workflow, not just to recommend it. Put adoption in that person's incentives and publish one weekly number everyone can see. Report the number in the same pack as the value-creation plan.

Request the Portco AI Diagnostic Book a portfolio call

Portco AI Diagnostic: $7,500 for two weeks on one portfolio company, credited against a sprint or retainer within 90 days.

Context

Where portfolio companies typically land

No score distribution is published from this Scorecard yet. The best available picture comes from the fund-level surveys.

36% of PE-backed portfolio companies use AI in day-to-day operations. FTI Consulting, 2026
7% call AI fully integrated across the portfolio. FTI Consulting, 2026
9% of operating partners have seen a demonstrable AI premium in a completed transaction. Accordion / Wakefield, May 2026

The Scorecard's aggregate bands are published in the Portco AI Benchmark once the sample passes 50 companies.

Add this company to the benchmark

Email me the scored breakdown and the fix for my top failure mode

Your six scores, your band and the first move on your priority mode, written out. No sequence you cannot leave in one click.

The six failure modes, defined

These are not technology problems. Every one of them is a decision that was never made, or an owner who was never named. More than 80% of AI programs still fail, usually on use-case misalignment, user adoption and unclear success measures - AlixPartners. The six below are what that looks like inside one portfolio company. The long version is in Six ways portfolio AI stalls before the P&L.

The License Trap

The portfolio company buys AI seats - Copilot, ChatGPT Enterprise, a vertical tool - and books the cost in month one. Usage is never tied to a named workflow, so the only number anyone can produce at renewal is spend. The license has become the AI program.

The Strategy Shelf

An AI strategy deck exists and it is usually good. Nothing in it carries an owner, a date or a number, so nothing ships. Two quarters later the deck is on SharePoint and the mandate has not moved an inch.

The Demo Graveyard

Pilots are built on exported data in a sandbox and they demo well. Production means the real system of record, real permissions, real edge cases and a maintainer, and the pilot was never designed for any of it. The demo wins the meeting and dies before the P&L.

The Workshop Certificate

Everyone attended the training and the completion rate goes into the board pack. Training changes what people know; it does not change what the work asks of them. Without a changed workflow and a changed standard, Monday looks exactly like the Monday before.

Vendor Lock-in

The working AI sits inside a vendor product, with the prompts, the data and the integrations on the vendor side. Switching costs rise every quarter and the margin sits with the vendor. At exit the buyer sees a subscription, not an asset.

No Adoption Owner

AI is assigned to a committee, a workstream or IT plus the consultants. Nobody's compensation, review or weekly number depends on the workforce using the system. Adoption is the whole job and it is the one job nobody owns.

How to read your score

0 - 5

On Track

AI is reaching real work and the discipline is in place. One mode is usually still soft. Watch it before it compounds, and keep publishing the number.

6 - 11

Pilot Purgatory

One or two failure modes are quietly draining the mandate. The EBITDA the board is counting on has not appeared. Fixable in one focused quarter, but not on its own.

12 - 18

Acute

The mandate is not reaching the P&L and the cost is EBITDA already promised in the value-creation plan. This does not self-correct. It needs an accountable owner who fixes adoption and ships the system.

The band matters less than the shape. Two portfolio companies can both score 10 and need opposite work: one has bought licenses nobody uses, the other has shipped something good that nobody owns.

Where the six come from

They come from doing both halves of the job. Three tours as a Chief Human Resources Officer, across Fortune 500, VC-backed and global organizations, is where the adoption, incentive and workforce failures were learned: the training that changed nothing, the tool nobody owned, the initiative that lived in a deck. Building and shipping AI systems that run in production - due diligence, HR policy, compliance, executive reporting - is where the technical failures were learned: the pilot that worked on exported data, the vendor that owned the prompts. Working inside PE-backed portfolio companies is where the two meet, on a hold-period clock, in front of a board that wants a number. The six modes are simply the failures that showed up in every engagement, in that order of frequency.

Questions

What does the AI Pilot-to-P&L Scorecard measure?

It measures six failure modes that keep AI from reaching the P&L in a portfolio company: the License Trap, the Strategy Shelf, the Demo Graveyard, the Workshop Certificate, Vendor Lock-in and No Adoption Owner. You score one portfolio company 0 to 3 on each, where 0 means the mode is not present and 3 means it is acute. The total out of 18 places the company in one of three bands and the highest-scoring mode is the one to fix first.

Who should complete it - the operating partner or the portfolio company team?

Either, and the comparison is useful. Operating partners usually score each portfolio company themselves to rank where the AI mandate is furthest from a number. Portfolio company CEOs, COOs, CFOs and CHROs score their own company. Where the two scores disagree, the gap is normally the Demo Graveyard or No Adoption Owner, and that gap is worth the conversation on its own.

What do the three score bands mean?

0 to 5 is On Track: AI is reaching real work and the discipline is in place, so watch the one soft spot. 6 to 11 is Pilot Purgatory: one or two failure modes are draining the mandate and the EBITDA the board is counting on has not appeared, which is fixable in a focused quarter. 12 to 18 is Acute: the mandate is not reaching the P&L at all and it does not self-correct without an accountable owner who fixes adoption and ships the system.

How is this different from a generic AI readiness assessment?

A generic AI readiness assessment scores technology maturity: data, infrastructure, governance, skills. This one scores the six ways money and mandate actually get lost between a pilot and the P&L in a private-equity-backed company, which is where more than 80% of AI programs fail, usually on use-case misalignment, user adoption and unclear success measures. It is written for a hold period and a value-creation plan, not for a technology roadmap.

What happens after I score a portfolio company?

You see the total, the band, the priority failure mode and a specific first fix for each of the six modes on the page. If you ask for it, the scored breakdown and the fix for your top failure mode are emailed to you. If you want the fix run properly, the Portco AI Diagnostic is a two-week engagement on one portfolio company at $7,500, credited against a sprint or retainer within 90 days.

Score it, then fix the top one

Two weeks on one portfolio company produces the workflow and adoption map, the EBITDA-impact model, the 100-day plan and the board one-pager. Or start with 30 minutes.

Book a portfolio call Portco AI Diagnostic