A portfolio review at a mid-market private equity firm surfaces six portfolio companies running six different AI subscriptions doing largely overlapping work. None of the AI licenses are instrumented for usage. The combined bill is larger than any single portco's engineering budget, and nobody at the fund can say what the AI license spend is actually returning.
This is not an AI adoption problem. Adoption already happened, independently, six times over. The gap is that nobody at the fund level has ever looked at what the portfolio pays for AI as one number, decided which licenses earn their renewal, and negotiated the portfolio's combined volume the way it already negotiates ERP and CRM contracts after an acquisition.
Every portco procures AI software the way any standalone business does, one team at a time. A 2026 study from Larridin covering AI tool usage across enterprises found the average organization runs 23 distinct AI tools, and only 38% maintain a complete inventory of what is actually active. Research from worqlo narrows that to 14 to 18 distinct AI tools specifically at organizations with 1,000 or more employees, a range that has roughly doubled since 2023.
Portcos below that headcount are not exempt from the pattern, just smaller in scale:
A fund with five or six portfolio companies is not looking at one instance of this pattern. It is looking at five or six, each accumulating its own uninventoried AI stack in parallel, on its own renewal calendar, with no one at the fund level holding the combined list.
Add-ons made up close to three-quarters of North American PE deal activity in 2025, according to PitchBook's Global PE Report, as cited by CohnReznick. Every add-on that closes brings a few things with it that rarely make it onto the integration checklist:
Post-merger integration work already covers ERP consolidation, duplicate CRM licenses, and data unification as standard practice. AI subscriptions sit outside that review because they read as a productivity tool rather than a recurring financial commitment. At close to three-quarters of deal volume running through this pattern, the AI license pile grows with every acquisition, and most integration checklists have not caught up to it yet.
Seat-based AI pricing punishes exactly the pattern most portfolios fall into: paying for assigned seats rather than active ones.
None of these numbers are unusual on their own. What changes the math is six portcos each carrying that same 18 to 25% seat-waste rate independently, on different tiers, at different price points, with no one comparing the totals. A single portco absorbing that waste on a $19 tool is a rounding error. Six portcos absorbing it simultaneously, some on Enterprise-tier pricing several times higher per seat, is a portfolio-level number the fund has never seen written down in one place.
SaaS and application rationalization practice generally sorts every tool a company runs into one of four buckets, a pattern that mirrors Gartner's TIME framework for application portfolios. Applied to AI licenses specifically, the same four buckets work:
This sorting exercise is where most of the return shows up. SaaS stack rationalization programs commonly recover savings in the range of 30% of software spend once unused seats, duplicate platforms, and downgraded tiers are counted together, according to industry rationalization research. AI licenses tend to have more room to run in that range than older software categories, since most AI tools were adopted in the last two years without any renewal discipline built in yet.
Consolidation is not always the right call, and treating every overlapping tool as a candidate for elimination misses cases where the overlap is only surface-level. A general-purpose assistant and a specialized coding assistant both count as "AI tools" in an inventory, but they rarely compete for the same job. Two reasons to keep more than one AI vendor in the portfolio even after a rationalization pass:
The goal of a rationalization pass is not the smallest possible vendor list. It is matching what's licensed to what's actually earning its cost, and only cutting overlap when the overlap is real.
Once a portfolio-wide inventory exists, mapped against real usage and billing data rather than self-reported tool lists, three things become possible that were not possible before:
Several vendors, including Microsoft and Adobe, already support enterprise license agreements that consolidate many individual licenses into one organization-wide contract with simplified budgeting and reduced per-seat cost. A fund that has built a portfolio-wide AI license inventory is positioned to negotiate exactly that kind of agreement across its portcos, instead of six separate portcos each negotiating their own AI contracts from a weaker position.
The standardization also travels. Once one portco's AI license usage is mapped and rightsized, the same inventory process applies to the next add-on that closes, instead of every acquisition starting the AI spend question from zero.
Ideas2IT works with private equity firms and portfolio companies the same way it approaches any post-acquisition execution gap: with a scoped assessment before any recommendation, not a platform pitch. The audit runs in three parts:
Ideas2IT holds SOC 2 Type II and ISO 27001 certifications and is an AWS GenAI Specialist Partner, which matters here because the audit touches billing and usage data across every portco in the portfolio, and the fund needs that handled to the same standard the rest of its technology engagements require.
If the fund has more than one portco running AI tools and no single view of what the portfolio pays for all of them combined, that is the gap to close before the next AI budget request reaches the IC.
Week 1: Full inventory of AI subscriptions across every portco, with seat counts mapped against actual usage
Week 2: Keep, consolidate, renegotiate, retire scoring for every license, plus a consolidation and negotiating recommendation across the portfolio
What the fund walks away with:
Didn't find what you were looking for?
