Spoiler alert: It’s not because they want to move slowly.

At IFA in Las Vegas, we sat down with four of the country’s largest franchisors. They were all using AI. On the surface, they are ahead of the curve. They’ve allocated a budget for AI, launched enterprise-level tools, and secured organizational buy-in.

But beneath that promising surface, I noticed significant issues that were undermining widespread adoption. These organizations are taking action, but they are struggling to convert access into adoption. They obviously want AI transformation for their business, but they’re not getting lift-off. In an industry where scalability is everything, these “early adopters” are finding that simply having the tools isn’t the same as winning the race. If your franchise systems are “doing AI” but not seeing the results, read on to see if the four pitfalls describe your situation.

1. Choosing Subpar Models Slows Adoption

Two of the four organizations I spoke with deployed AI solutions that were, frankly, not best-in-class. The top overall models right now, per consensus and our humble view, are OpenAI’s ChatGPT, Google’s Gemini, and Anthropic’s Claude. And we saw “top” not necessarily based on a scoreboard showing performance on various tests, but also on general usability and enterprise suitability.

The result of choosing an alternative model was a lack of employee adoption. It was clear that people wanted to adopt it, but there was something about the limitations and weak usability of the selected model that prevented that adoption. The trouble is, if a tool doesn’t immediately change behavior or make work easier, it becomes shelfware. There’s a limited window where people pay attention to something new, taking time out of their busy schedule, and the team’s enthusiasm must be engaged in that limited window.

2. Restrictive Data Policies Stifle Innovation

Data privacy is critical. Absolutely critical.

That said, above-and-beyond strict internal rules create a “chilling effect”.

Two individuals shared that their company implemented governance so strict that employees were hesitant to actually use the platforms. Leading AI providers already offer enterprise-grade security and confidentiality. These leading providers are worth half-a-trillion dollars; valuations that are not possible if their suitability for enterprise-grade privacy and confidentiality weren’t up to standard. Famously, or infamously, Anthropic is embroiled in a tiff (that may have been resolved on February 25th when they backed down from their legacy policies) with the Department of War. The DoW has been using Anthropic’s Claude LLM and wants to expand use. These platforms are being used in the most sensitive environments possible.

Effective AI transformation for franchising requires smart guardrails, not locked doors that prevent your team from building AI intuition. IT departments and senior leaders should be reading the privacy and confidentiality policies of their LLMs. However, the sensitivity being displayed does not match the privacy and confidentiality policies that their LLMs have. These companies are being more careful than what is warranted, and that is limiting their organization’s use of AI and slowing their broader AI transformation.

3. The Narrow Win Trap

Businesses invest in specialized, “fit-for-purpose” tools for a single workflow. That’s been the common approach to software over the last ~20-30 years. These tools certainly offer robust capabilities.

However, what isn’t happening is a careful review of the new paradigm. With platforms like Anthropic’s Claude or Microsoft’s Co-Pilot, so much capability is being collapsed into their general purpose LLM platforms. These platforms, and cost effective multi-purpose agent platforms like Lindy and Zapier, offer the ability to execute a wide wide variety of tasks. There’s an opportunity to no longer buy a dozen niche tools when one powerful engine can handle a vast majority of tasks for a fraction of the price.

Additionally, tool proliferation also carries a hidden cost of added administration and maintenance.

The shift with the uber powerful general purpose platforms like a ChatGPT is this: use them really well because they can do the work that previously required multiple specialized tools.

4. The Internal Build Hidden Cost

One organization tasked their IT team with building a proprietary AI solution. On one hand, building AI applications has gotten hyper cost effective today. That’s why software stocks like Intuit and Salesforce are down 30-50%. Someone in their basement can vibe code a replacement in a few days. So, an IT professional should be able to do even better.

However, building internally comes with real costs and opportunity costs. The real costs are: time taken to build, ongoing cost and effort to maintain, staying on top of security issues, break-fixes, upgrading to keep up with innovations, and more. The opportunity costs include foregoing the same use, potentially, on a general purpose platform like a ChatGPT or a Claude.

The Strategic Takeaway

AI adoption at your franchise business is a fantastic strategic objective. And we applaud companies for getting going; in fact, we say the same thing “just get started!” However, as we noted talking to some of the largest players, there can be ways to get going that actually hinder adoption. We suggest:

  • Consider including access to one of the three major LLMs – ChatGPT, Claude, or Gemini – so they experience best-in-class AI use that encourages adoption. It’s like the iPod versus other MP3 players from yesteryear – the right form factor and/or user experience can make a world of difference.
  • Carefully read the security, privacy, and confidentiality policies of your chosen provider. And we recommend not creating a too-restrictive policy for internal use that ignores the strong guardrails of your LLM provider and consequently limits your team’s adoption.
  • Think carefully about fit-for-purpose solutions. We are coming off of a two-to-three decade run where these solutions were the norm. It’s hard to change. However, the general purpose platforms are so exceedingly capable and also cost effective that it’s absolutely worthwhile starting with them – and understanding how to use them well via AI training and intuition – instead of one-by-one fit-for-purpose solutions.
  • Think carefully about building in-house. The hidden costs of doing so are real. And the robustness of the general purpose platforms merits their use instead of in-house.