The $18 Trillion Number That Should Reframe How PE Firms Think About AI Spend

AI isn't creating less value than expected because the models aren't good enough.

Most companies are asking AI to compensate for years of accumulated enterprise debt. New research from Genpact and HFS Research puts a hard number on the problem: roughly $18 trillion in enterprise value remains trapped behind four forms of debt that existed long before generative AI arrived. The study surveyed more than 2,000 enterprise executives across 16 industries.

One sentence from the report captures it: "Every dollar spent on AI atop a broken foundation is a dollar working against itself."

That's the research summary. The rest of this issue is about what it means if you're operating inside a PE hold period rather than a Global 2000 transformation program.

The research identifies four interconnected enterprise debts. Each one amplifies the others, making AI harder to deploy and even harder to scale.

Data debt is the single biggest AI blocker, cited by 33% of respondents. Without trusted, integrated, AI-ready data, use cases stay in proof-of-concept permanently.

Technology debt (28%) inflates AI unit costs and makes integration into core workflows difficult.

Process debt (23%) makes AI agents unreliable in production. They operate inside broken workflows, automating inconsistency rather than removing it.

Talent debt (16%) slows adoption and limits the human judgment that agentic operating models depend on at the last mile.

Process and data debt each represent nearly $7.7 trillion of the $18 trillion total. Resolving them unlocks approximately 8% faster annual revenue growth and 16% annual cost reduction.

Yet 85% of leaders surveyed say these debts are actively limiting their AI value. Over half have no funded plan to address them. Only 6% have built, measured, and scaled a resolution program. Those are the companies seeing measurable results.

For an operator or investor in a PE-backed company, the plan is different from that of a Global 2000 CIO.

The Global 2000 companies in this study have full-time AI governance functions, dedicated data engineering teams, and the luxury of multi-year transformation timelines. Portfolio companies operating inside a 3-to-5-year hold period have none of those.

What they have is a clock.

Portfolio companies running multiple ungoverned AI initiatives can be spending $10,000 to $25,000 a month. The question isn't whether enterprise debt exists. It almost certainly does. The question is which debt is the binding constraint on value creation in the next 12 to 18 months, and whether it can be resolved within the hold period at a cost that improves the multiple.

The Genpact report is written for enterprises thinking about long-term structural transformation. PE-backed companies need to think about it as a triage problem with a return deadline.

In my experience across PE-backed software companies, the debt mix is predictable before you look at a single system. It follows how the deal itself was structured.

Founder-built companies that scaled fast tend to carry heavy process debt and technology debt. They built what worked at 50 customers and never had time to rebuild it for 500. The AI they're deploying now operates inside processes that were already inconsistent before automation accelerated them. The result is what the Genpact report calls "encoding existing inefficiencies into automated systems and running them at speed." Then presenting that acceleration to the board as AI progress.

Acquired or PE-assembled companies, particularly those that have gone through multiple mergers, tend to carry the worst data debt. When three systems of record become one company without proper data consolidation, the AI doesn't know which customer record to trust.

The AI isn't making bad decisions. It's receiving multiple conflicting versions of the truth, one from each system of record, that were never reconciled. What looks like an AI cost-reduction initiative on paper quietly becomes a data integration effort in practice, and it's exactly the kind of problem pre-close financials are not built to catch.

Companies that hired aggressively in 2021 and 2022 and then cut in 2023 tend to carry serious talent debt. The institutional knowledge of how processes actually work left with the people who were let go. What remained was a team that knows what the process is supposed to do but not why it was built that way. That distinction matters enormously when you're determining which steps are safe to automate.

*Some of what looks like talent debt in a secondary buyout is something else: a prior sponsor's spend, deferred on purpose to protect their exit multiple. Different diligence question. More on that in a future issue.

None of these patterns show up cleanly in financial diligence. They surface in engineering interviews, architecture reviews, and the first 90 days after close.

1. Which debt is actually blocking AI value right now?

All four types of debt exist in most companies. The question is which one is the binding constraint on the specific use case being funded. A $200,000 AI budget allocated to a customer-facing initiative within a company with severe data debt will spend most of that money on discovering the data problem, not solving the use case.

2. Can the binding constraint be resolved within the hold period at a cost that improves the multiple?

Some debts are 90-day problems. Some are two-year migrations. A process debt problem that can be resolved with workflow redesign in 60 days and then automated is worth funding aggressively. A technology debt problem that requires a platform rebuild before AI can be deployed is a sequencing conversation, not a green light to spend on AI now.

3. Is success measurable in terms the board will recognize?

The 6% of companies with measurable results are the ones that built, measured, and scaled, in that order. Portcos that measure AI success in output metrics (cost per task, cycle time, error rate, support ticket volume) rather than input metrics (dollars spent, models deployed, features built) are the ones that convert AI investment into defensible EBITDA.

According to the research, process debt consumes roughly 40% of employee time through manual, ungoverned work. Before the next AI investment conversation in a portfolio company, ask the engineering or operations team to show you the three most manual processes in their workflow. If those processes are well-documented and consistently executed, AI can probably help. If the team struggles to describe them clearly, that's the debt that needs to be resolved first.

Portco Brief is a weekly briefing for PE operating partners and portfolio company executives focused on technology, AI, and value creation.

Subscribe at portcobrief.com.

Source: Genpact and HFS Research, "The $18 Trillion Opportunity: How Four Enterprise Debts Will Make or Break Your AI Future," June 2026. Survey of 2,002 enterprise executives across 16 industries and 14 functions.