Skip to content

The definitive guide · Version 1.0 · September 2026 · updated every six months

The Pilot-to-P&L Standard

The operating standard a PE-backed company must meet before an AI mandate shows up as a number the CFO signs. Ten criteria, six failure modes, one free assessment, a 100-day playbook and an annual benchmark. Written by the operator who holds the seat, not by a vendor with a product to sell.

The standard in one sentence

A portfolio company meets the standard when AI runs inside its system of record, with a named adoption owner, a baselined number in the CFO's own loaded labor cost, weekly usage measured, and the result reported to the board in basis points of EBITDA.

Definition

Pilot purgatory has an exit. It is a standard, not a strategy.

Most portfolio AI work ends in one of two states. A pilot that demoed well and never reached production. Or a licence that everyone has and nobody can tie to a dollar. Sponsors call the whole condition pilot purgatory. Only 36% of PE-backed portfolio companies use AI in day-to-day operations and 7% call it fully integrated across the portfolio (FTI Consulting, 2026).

The Pilot-to-P&L Standard names the conditions under which the number arrives. It is deliberately not a maturity model. Maturity models grade activity: seats, certificates, proofs of concept. The standard grades ten conditions a CFO can check in a day. A company either meets each one or it does not.

Advisors advise. Vendors build. The standard asks one question: who owns the number, and can the board see it this month.

Why now

The hold period is the deadline.

0

years, the record average holding period. McKinsey, 2026

0

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

0

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

0

of operating partners have seen an AI premium in a completed exit. Accordion, 2026

Buyers now ask about AI in diligence. A company that meets the standard walks into an exit with an adoption record, a named owner and an EBITDA attribution. A company that does not walks in with a deck.

The ten criteria

Pass or fail. Checkable in a day.

Each criterion has a pass condition and a way to check it without a consultant. A company meets the standard in a function when all ten hold for that function.

  1. 01

    A line in the value-creation plan with a number

    Pass: AI appears in the VCP with a target in dollars or basis points, not as an initiative. Check: read the last board deck.

  2. 02

    One named adoption owner

    Pass: a single senior person inside the company owns adoption and the number sits in their incentives. Check: ask who loses their bonus if usage stays flat.

  3. 03

    A baseline in the company's own loaded labor cost

    Pass: the function's hours and cost were measured before anything shipped. Check: the baseline uses the CFO's numbers, not a vendor's benchmark.

  4. 04

    Runs in the system of record

    Pass: the system lives where the work already happens (ERP, CRM, HRIS, ticketing), not in a sandbox or a separate tab. Check: watch one person do the task.

  5. 05

    Real data, named maintainer

    Pass: production data flows through it and one person is responsible when it breaks. Check: ask what happened the last time it failed.

  6. 06

    Weekly usage measured and published

    Pass: usage by person is counted weekly and visible to the team. Check: ask for last week's number.

  7. 07

    Three to five trained operators

    Pass: people inside the company can run and extend the system without the builder. Check: ask them to change one prompt and one rule.

  8. 08

    The company owns the IP

    Pass: prompts, data and the integration layer are the company's; the model can be swapped. Check: read the vendor contract's termination clause.

  9. 09

    Monthly board report in basis points of EBITDA

    Pass: the board sees recoverable cost and adoption rate against the baseline every month. Check: find the line in the last board pack.

  10. 10

    Packaged as a playbook

    Pass: the rollout is documented so the next function, and the next portfolio company, costs less. Check: ask for the owner brief, training pack and board template.

Criteria 2, 6 and 9 decide the others. A company that names the owner, publishes the usage number and reports monthly tends to close the remaining gaps inside one quarter.

The six failure modes

How companies miss the standard

Every miss shows up as one of six patterns. They are what the free Scorecard measures, 0 to 3 each.

  1. 01

    The License Trap

    Seats paid for, no usage that maps to a dollar. Misses criteria 3, 6, 9.

  2. 02

    The Strategy Shelf

    A deck, months old, nothing shipped. Misses 1, 4, 5.

  3. 03

    The Demo Graveyard

    Pilots that die before production. Misses 4, 5, 7.

  4. 04

    The Workshop Certificate

    Training completed, output unchanged. Misses 6, 7.

  5. 05

    Vendor Lock-in

    The leverage sits outside the company. Misses 8.

  6. 06

    No Adoption Owner

    Nobody's bonus depends on usage. Misses 2, and then everything else.

Playbook · updated every six months

The 100-day plan to meet the standard

One function, one owner, one number, one system. Day 0 to 30 diagnose and baseline; 31 to 60 ship into the system of record and train; 61 to 100 scale, report to the board and package the playbook.

Read the playbook

Benchmark · annual · first edition 2026

The Portco AI Benchmark

Where lower-middle-market portfolio companies actually stand against the ten criteria, built from Scorecard results and a 12-question survey of operating partners and portco leadership. Respondents get the report first.

Add your company

Getting there

Meet the standard in one hold-period quarter.

PortLev's engagements are built around the ten criteria. Each rung credits into the next; nothing is sold that a board cannot read on one page.

  1. Portco AI Diagnostic · two weeks · $7,500, credited

    Scores the company against all ten criteria in the functions that matter, builds the EBITDA-impact model in its own loaded labor cost and names the adoption owner.

  2. 100-Day Pilot-to-P&L Sprint · from $45,000

    Meets criteria 1 to 10 in one function: shipped system, trained operators, weekly usage, board number, packaged playbook.

  3. Fractional AI Operating Partner · $19,650 per month

    Holds the standard across functions and portfolio companies; monthly board reporting; playbook reuse so each rollout costs less.

Questions

About the standard

What is the Pilot-to-P&L Standard?
An operating standard of ten criteria that a PE-backed portfolio company meets when AI has stopped being a pilot and has become a number in the P&L: a named adoption owner, a baselined number in the company's own loaded labor cost, a system running in the system of record with real data, weekly usage measured, and monthly board reporting in basis points of EBITDA. PortLev coined it in 2026 and updates it every six months.
How do I assess a portfolio company against the standard?
Start with the free AI Pilot-to-P&L Scorecard: six failure modes scored 0 to 3 in three minutes, which tells you how far the company is from the standard and which failure mode to fix first. The Portco AI Diagnostic then produces the costed plan to meet all ten criteria in one function within 100 days.
Why ten criteria and not a maturity model?
Maturity models grade activity: seats licensed, people trained, pilots run. The standard grades the conditions under which a number reaches the P&L. Each criterion is a pass or fail a CFO can check in a day, which is what a board needs during a hold period.
Does the standard apply outside private equity?
Yes. It was written for PE-backed companies because the hold period forces a deadline, but any founder-led or mid-market company with a board-level AI mandate can apply the same ten criteria. Owner-operators with 5 to 200 people are served by the separate AI Org Audit.
How often is the standard updated?
Every six months, with a dated version number on this page. Changes come from engagements, from the Portco AI Benchmark data and from the published surveys cited here. Version 1.0 is dated September 2026.

Find out how far one portfolio company is from the standard.

Three minutes for the Scorecard. Twenty minutes for a conversation about the function you would fix first.

Score a portcoBook a call