PE's AI ambitions are hitting a wall - and franchise portfolios are exhibit A

Private equity's enthusiasm for AI is not in question. AI and machine-learning global PE deal value more than tripled from $41.7 billion in 2023 to $140.5 billion in 2024, according to Accenture. EY data shows 38% of PE firms expect to spend more than half of their total budget on AI in 2026. The investment thesis is clear, the capital is flowing, and board-level commitment is real. The execution story is another matter.

According to FTI Consulting's 2026 Private Equity AI Radar, only 36% of portfolio companies have deployed AI across use cases, and just 7% have achieved enterprise-scale deployment. Grant Thornton's 2026 Private Equity Insights AI Impact Survey found that 45% of PE firms are still in the piloting phase, with only 5% having fully integrated AI into operations - compared to 14% across all industries. The gap between what PE firms are spending on AI and what they are getting from it is not a technology problem. It is a data problem.

The constraint nobody wants to name

PE Professional's April 2026 investigation into why private market AI ambitions are stalling found the same answer surfacing from firm after firm: data quality, not the sophistication of AI models, is the primary constraint on adoption. PE portfolios contain large amounts of unstructured information from spreadsheets, emails, PDFs, and data files built on undefined schemas, inconsistent identifiers, and unpredictable formats. The models are ready; the data isn't.

AI can be a multiplier of operational weaknesses and risk exposure if implemented on inadequate, old data infrastructure.

AI doesn't neutralize bad data; it amplifies whatever is underneath it. A model trained on inconsistent location-level records doesn't produce mediocre outputs - it produces confidently wrong ones. For PE operating partners, the risk is decisions made at portfolio scale on information that looks authoritative but isn't.

Why franchise portfolios face a harder version of this problem

The data foundation challenge affects every PE portfolio company. Franchise portfolios face a version of it that is harder to solve.

When a PE firm acquires a franchise brand, it inherits that brand's technology stack, reporting conventions, and data architecture - or lack thereof. PwC's analysis of AI and data readiness in PE portfolio companies notes that middle-market portfolio companies lag larger enterprises in technology and data maturity due to historical underinvestment, producing fragmented systems, limited automation, and reliance on manual workarounds.

For a single acquisition, that's a known integration challenge. For a portfolio operator managing 5, 15, or 30 brands, the work repeats at every deal.

Each brand in a multi-brand portfolio uses different formats, definitions, and reporting standards, according to PwC. Portfolio companies routinely collect and submit financial data by email, with associates copying figures into spreadsheets that then feed manual reporting chains. That isn't a legacy problem at a single brand; it's the standard operating mode across a majority of acquired companies.

Portfolio-level reporting then requires manual reconciliation at every step. Operating partners trying to surface AI-ready insights across a portfolio of franchise brands are not one model away from that capability. They are multiple data consolidation projects away.

At scale: Apex Service Partners manages 107 acquired HVAC brands generating $1.3 billion in annual revenue across the top 50 U.S. markets. Authority Brands operates 15 home-service franchise brands, adding 246 new franchise owners and 340 new territories in 2025 alone. Blackstone's Champions Group Holdings operates HVAC and electrical brands across multiple regions with disparate legacy systems. The same structural challenge repeats at each platform: the portfolio is growing faster than the data infrastructure supporting it.

The timeline math that operating partners are ignoring

Gartner has predicted that 60% of AI projects unsupported by AI-ready data will be abandoned through 2026. S&P Global Market Intelligence found that 42% of companies abandoned at least one AI initiative in 2025, with the average sunk cost per abandoned initiative reaching $7.2 million (S&P Global, Voice of the Enterprise, 2025). Pertama Partners' analysis of AI project failure rates found that data quality issues affect 71% of failed projects, with data preparation consuming an average of 61% of the total project timeline.

If data preparation is consuming 61% of a project timeline, and integration timelines are running 2.4 times the original estimate, then the real cost of AI deployment in underprepared portfolio companies is not the model cost or the vendor fee. It's the months spent trying to make ingestion work on data that was never built to be ingested.

PE hold periods currently average 5.8 years - down from a peak of 7.1 years in 2023. An AI implementation that requires 12 to 18 months to reach production - if data preparation goes smoothly - consumes a meaningful portion of the value creation window. Every month spent cleaning and normalizing data at one brand is a month not spent generating returns across the portfolio.

Firms moving fastest aren't spending less time on data. They're spending that time earlier, before the AI projects start, treating data foundation work as infrastructure rather than project overhead.

What the deployment gap actually reveals

The 7% enterprise-scale AI deployment figure from FTI Consulting is often read as evidence of slow adoption. It more accurately reflects how few portfolio companies have solved the prerequisite problem.

Enterprise-scale deployment requires the AI system to have consistent, reliable data across the full scope of the deployment. For a portfolio operator, that means consistent data across every brand, every location, and every data category the AI needs to inform. A platform that has solved this at one brand has not solved it at the portfolio level. The data work has to happen at each layer.

PE Professional describes the firms that have cracked this as those built on "flexible, centralized, automated data foundations with proper data ontology" - meaning they've mapped the relationships between all datasets across the portfolio, not just standardized formats within a single brand. At that point, agentic AI can execute reconciliations, process operational notices, and draft reporting with confidence. Below that threshold, the same tasks require manual validation that negates most of the efficiency gain.

Why early movers pull ahead - and stay ahead

Franchise portfolio operators who solve the data foundation problem before their competitors build AI returns across brands in ways that single-brand operators and late-moving portfolio companies cannot.

When a model is trained or fine-tuned on consistent data from one brand in a portfolio, the learnings transfer to other brands more efficiently than starting from scratch. When agentic systems are deployed at portfolio scale, oversight and management overhead is shared. When location-level data is standardized across brands, the portfolio-level reporting that operating partners need to allocate resources and identify underperformers becomes automated rather than manual.

The inverse holds too. AI search visibility for franchise systems depends on data consistency at the location level. According to 5W PR's HVAC & Plumbing AI Visibility Index, reported in Plumbing & Mechanical, about 87% of independent HVAC and plumbing contractors have near-zero AI citation share in their metro and category. PE-backed portfolio brands have a structural advantage through cross-brand citation reinforcement - but only if underlying location data is consistent and accessible.

The data foundation problem is not a technical afterthought to AI deployment. It determines whether portfolio-scale AI is possible at all.