The AI proof gap: why most franchise networks can't show what their AI is doing

There is a specific kind of problem that emerges when an organization deploys AI at scale without first building the infrastructure to measure it. The AI runs, reports get generated, and someone presents a slide deck at the quarterly board meeting. But when a board member or PE operating partner asks "can you show us where this created value?" the answer becomes complicated.

This is the AI proof gap. And for franchise networks, it is more exposed than almost any other business structure.

What the data actually says

Grant Thornton's April 2026 AI Impact Survey, which drew on 950 C-suite and senior business leaders across ten industries, found something that should make any franchise executive pay attention: organizations with fully integrated AI are nearly four times more likely to report revenue growth than those still running pilots.

That finding gets more complicated when you look at what "fully integrated" actually requires. The same survey found that 78% of business executives lack strong confidence they could pass an independent AI governance audit within 90 days. Among organizations still in the piloting phase, only 7% feel very confident about their governance posture. Among fully integrated organizations, that number rises to 74%.

The gap between those two numbers is not a coincidence. It reflects a relationship that most organizations have not yet internalized: governance is not compliance overhead. It is the mechanism that turns AI activity into provable business results.

The 95% problem

MIT's NANDA Initiative published its State of AI in Business 2025 report after analyzing 150 leader interviews, a survey of 350 employees, and 300 public AI deployments. The headline finding was stark: 95% of enterprise generative AI pilots fail to deliver measurable ROI.

What the MIT research identified as the core failure is worth examining carefully. Rather than the quality of the AI models, the problem was what the researchers called the "learning gap": most generative AI systems do not retain feedback from use, do not adapt to organizational context over time, and do not improve. They are deployed as static tools into dynamic organizations and then measured against expectations that assumed the tool would keep learning.

The research also found a meaningful difference in success rates depending on how organizations acquire AI capabilities. Purchasing tools from specialized vendors succeeds roughly 67% of the time. Internal builds succeed only about one-third as often.

What vendors promise versus what research shows

Deloitte's research on AI ROI, published across 2024 and 2025, found that most organizations achieve satisfactory returns on typical AI use cases within two to four years. Only 6% report payback in under a year. Among the most successful projects, only 13% see returns within 12 months.

This is a significant problem for any franchise network that has been sold on AI by vendors promising three-to-six-month payback periods. The vendor promise is not baseless: narrow, well-scoped implementations like call routing, basic chatbots, and rule-based scheduling automation can generate measurable savings quickly. But those are not the implementations franchise boards and PE partners are asking about. They are asking about AI that changes revenue trajectories, improves franchisee performance, or creates durable competitive advantage. That kind of AI takes time to prove.

Organizations with formalized AI governance are 2.2x more likely to demonstrate ROI than those without.

McKinsey's five-layer AI measurement framework identifies strategic outcomes (revenue, operational efficiency, labor productivity, quality and compliance, and strategic value) as the dimensions that matter most. The finding buried inside that framework is the same one Grant Thornton surfaces from a different angle: the organizations that can prove ROI are the ones that built the measurement infrastructure before they deployed it across hundreds of locations.

Why franchise networks face this differently

The proof gap is not unique to franchising. But franchise networks have structural characteristics that make the problem substantially harder to solve than it is for a single-entity enterprise.

Adoption variance across locations. A traditional enterprise can mandate a technology rollout. Franchise agreements create a more complicated dynamic. Research analyzing 29 franchisees across seven franchise systems found consistent perceptual gaps between franchisors and franchisees on what drives adoption, with intracommercial competition acting as a barrier and low levels of knowledge-sharing among franchisees. When a franchisor deploys AI and twenty percent of locations embrace it while forty percent modify it and the remaining forty percent ignore it, measuring network-wide ROI becomes nearly impossible.

Data fragmentation across sites. A franchise network operating fifty to three hundred locations typically has locations running different point-of-sale systems, different CRM implementations, different scheduling tools, and different reporting conventions. Corporate teams lose visibility into location-level performance. Regional managers fill the gap with institutional knowledge. The result is that when AI is deployed, the baseline data needed to measure its impact often does not exist at the network level.

The enthusiasm-deployment gap. The IFA's State of Franchise Marketing Report found that 87% of franchisees believe more extensive use of AI tools would improve their marketing performance. The IFA's 2025 Franchisor Survey found that only 28% of franchisors have actually incorporated AI and increased automation to address their operational challenges. That gap between what franchisees believe AI can do and what franchisors have deployed is not a technology problem. It is a measurement and proof problem. Franchisors are not deploying AI across their networks because they cannot yet show that it works and cannot design measurement systems that would let them prove it.

The governance pressure is arriving

Private equity is beginning to formalize what was previously an informal expectation. Vista Equity Partners, according to EY's AI in Private Equity research, now requires each of its portfolio companies to submit quantified benefits from generative AI initiatives as part of annual operational planning. The specifics vary by fund and portfolio company, but the direction is consistent: PE operating partners are moving from asking "what AI are you using?" to asking "what did it produce?"

For franchise networks backed by PE, or those with PE-backed operators among their franchisee base, this creates an accountability structure that the existing AI deployment playbook does not support. Deploying AI without building measurement into the architecture from the beginning means arriving at annual planning without the evidence the LP-facing reports require.

The measurement gap is architectural, not analytical

Read together, the Grant Thornton, MIT, Deloitte, and McKinsey research points to the same conclusion: the measurement problem is not something organizations can retrofit after deployment. Organizations that demonstrate ROI built measurement infrastructure before they scaled. Those that cannot demonstrate ROI deployed first and tried to build measurement second.

For franchise networks, this has specific implications. Network-level vanity metrics (total AI interactions, licenses deployed, locations onboarded) will not satisfy a governance audit or a PE operating partner's annual review. What satisfies those standards is location-level outcome data: did AI change conversion rates at this location, did it reduce response time to inbound leads, did it improve technician scheduling efficiency in a way that shows up in revenue per location?

Building that kind of measurement requires establishing baselines before deployment rather than after. It requires governance frameworks designed to work within the constraints of franchisee autonomy, rather than frameworks built for centralized enterprises that treat every location as a fully controlled business unit. And it requires realistic timelines: two to four years for strategic AI implementations to show full returns, with clear intermediate milestones that demonstrate progress before the full payback period closes.

What separates the 5%

MIT research identified the 5% of enterprise AI deployments that do deliver measurable ROI. Organizational readiness is the common thread - not superior AI technology: feedback loops that actually close, governance that creates accountability without creating bureaucracy, and measurement that connects AI activity to outcomes the business already tracks.

The core issue is not the quality of the AI models, but the learning gap for both the tools and the organizations.

Franchise networks that are about to scale AI deployments face a choice. They can move quickly with the tools available, optimize for speed of deployment, and build measurement later. That is the path most organizations have taken. The Grant Thornton data shows where it leads: 78% unable to pass a governance audit, nearly half reporting that AI underperforms because controls and compliance are not working.

The alternative is to build measurement infrastructure into the architecture before scaling. Establish location-level baselines. Define what governance looks like at the franchisee autonomy level, not just the corporate level. Set timelines based on Deloitte's two-to-four-year reality, not vendor promises of six months. Create feedback loops that allow the network to learn from early locations before mandating adoption across hundreds of sites.

The 4x revenue growth advantage belongs to organizations that chose the second path. The proof gap belongs to the ones who chose the first.

BeForm works with service-based franchise networks and PE-backed portfolio companies to identify where AI creates measurable impact and build measurement infrastructure from the first engagement. If your network is deploying AI without a clear proof structure, the Discovery Workshop is where that changes.