Assessment · free · 3 minutes
AI Pilot-to-P&L Scorecard
Score one portfolio company on the six failure modes. The total says how far it is from the standard; the top mode says what to fix first. Emailed with the first fix.
Score a portcoThe definitive guide · Version 1.0 · September 2026 · updated every six months
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
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
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
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.
Pass: AI appears in the VCP with a target in dollars or basis points, not as an initiative. Check: read the last board deck.
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.
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.
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.
Pass: production data flows through it and one person is responsible when it breaks. Check: ask what happened the last time it failed.
Pass: usage by person is counted weekly and visible to the team. Check: ask for last week's number.
Pass: people inside the company can run and extend the system without the builder. Check: ask them to change one prompt and one rule.
Pass: prompts, data and the integration layer are the company's; the model can be swapped. Check: read the vendor contract's termination clause.
Pass: the board sees recoverable cost and adoption rate against the baseline every month. Check: find the line in the last board pack.
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
Every miss shows up as one of six patterns. They are what the free Scorecard measures, 0 to 3 each.
The License Trap
Seats paid for, no usage that maps to a dollar. Misses criteria 3, 6, 9.
The Strategy Shelf
A deck, months old, nothing shipped. Misses 1, 4, 5.
The Demo Graveyard
Pilots that die before production. Misses 4, 5, 7.
The Workshop Certificate
Training completed, output unchanged. Misses 6, 7.
Vendor Lock-in
The leverage sits outside the company. Misses 8.
No Adoption Owner
Nobody's bonus depends on usage. Misses 2, and then everything else.
Assessment · free · 3 minutes
Score one portfolio company on the six failure modes. The total says how far it is from the standard; the top mode says what to fix first. Emailed with the first fix.
Score a portcoPlaybook · updated every six months
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 playbookBenchmark · annual · first edition 2026
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 companyGetting there
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.
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.
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.
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
Three minutes for the Scorecard. Twenty minutes for a conversation about the function you would fix first.