Everyone is talking about agentic AI.
Not the chatbot-answers-your-question kind. The new kind. AI that does not just advise — it acts. It processes a refund. It chases a late invoice. It triages a customer complaint, updates the CRM, and escalates to the right person. It reads the operations manual your franchisees never open and applies it to the problem in front of them — at 2am on a Sunday, without calling head office.
This is genuinely different from what came before. And it is coming to franchise operations fast.
But before you rush to deploy your first digital agent, I would like you to slow down and answer two questions. Get these right, and agentic AI could transform your network. Get them wrong, and you will just be doing the wrong things faster.
Question one: does this process deserve to survive?
Every franchise network has processes that exist because they have always existed. The five-step approval chain that was created after someone made an error in 2019. The weekly report that takes half a day to compile and that three people glance at before filing. The reconciliation between two systems that should have been integrated years ago but never were, because it was never quite painful enough to fix.
These processes are not just inefficient. They are organisational scar tissue — the residue of problems that were patched rather than solved.
And here is the danger with agentic AI: it will happily automate all of them.
An AI agent does not ask whether a process makes sense. It does not wonder why a five-step approval exists for a £50 purchase. It does not question why the same data is being entered into three systems. It just does it — faster, more consistently, and without complaint.
Which sounds like a good thing, until you realise that you have just invested time and money in making your business efficiently wasteful.
Before you hand a workflow to an agent, ask the question I suggested in a previous post: if this task did not exist, would we invent it? Not “is this task necessary” — of course it feels necessary, it has always been done. But if you were designing your operation from scratch today, with today’s tools, would you design it this way?
If the answer is no, do not automate it. Retire it. Then think about what should replace it.
The best use of agentic AI is not to do your existing work faster. It is to do the right work — the work that actually creates value for your franchisees and their customers — and to do it consistently, at scale, across every location in your network.
The accuracy trap
So you have cleaned up your processes. You are ready to deploy. And then someone on the board asks the question that stops every AI project in its tracks:
“What if it gets it wrong?”
Fair question. Important question. But it is almost always asked in a vacuum — as though the alternative to AI making errors is humans making none.
Let me tell you a story.
I was working with a business that had deployed an AI-driven process for handling customer emails. It was categorising enquiries, extracting key information, and routing them to the right team. The system was running at 95 percent accuracy.
The client was not happy. “Ninety-five percent is not good enough,” he said. “We cannot have one in twenty emails handled incorrectly.”
I understood his concern. So we ran an experiment. We took the same sample of emails and gave them to his five best call centre staff — the people he considered his gold standard. We measured their accuracy on the same task.
The result? Ninety percent.
His best people were less accurate than the AI he wanted to switch off.
This is not an unusual finding. Research consistently shows that human accuracy on routine processing tasks sits between 95 and 99 percent for simple work, dropping sharply as complexity increases. Contact centre disposition — the act of categorising what a customer called about — has been measured at as low as 60 percent accuracy in studies of manual processes. Data entry without verification has an error rate of up to 4 percent, meaning 400 errors in every 10,000 entries.
We do not notice this because we have never measured it. Human error is absorbed quietly. It becomes part of the background noise of operations. The misrouted email gets forwarded eventually. The miscategorised call gets corrected by someone downstream. The data entry error gets caught in the monthly reconciliation — or it does not, and nobody knows.
AI makes errors visible because AI can be measured precisely. And that precision creates an unfair comparison: AI at 95 percent accuracy versus an imagined human at 100 percent.
The imagined human does not exist.
Building in the right checks
None of this means you should deploy AI blindly and hope for the best. The right approach — and this is not new, it is just good operational practice — is to design your processes with appropriate verification.
For high-volume, lower-stakes tasks, let the agent run and flag exceptions. If your AI is processing 10,000 invoices a month at 98 percent accuracy, you are not reviewing 10,000 invoices. You are reviewing 200. That is a very different job from the one you had before, and a much better use of your team’s expertise.
For higher-stakes decisions, build in human checkpoints at the moments that matter. The agent does the preparation, the analysis, the routing. A person makes the final call on the cases that require judgment. This is not a compromise — it is the optimal design. Research from UNLV found that humans empowered by automation are over twenty times more accurate than humans working alone.
For franchise networks specifically, this matters at scale. A 5 percent error rate on a process might feel manageable when you are looking at a single location. Across 200 locations, that is not a rounding error — it is a material cost, a compliance risk, and a customer experience problem. Consistent, measurable, auditable processes — with humans in the loop where it counts — are not just more efficient. They are more reliable than what you have today, even if what you have today feels like it works.
The point is not that AI is perfect. It is that the honest comparison — AI with proper checks versus humans without them — almost always favours the AI. And the conversation changes entirely when you stop comparing AI to perfection and start comparing it to reality.
Question two: what do your people do with the time?
Let us say you have done the hard thinking. You have retired the processes that did not deserve to survive. You have deployed agents on the ones that do, with sensible checks in place. And it is working. Your team is spending less time on administration, data entry, reconciliation, and routine customer handling.
The stat doing the rounds is that AI automation can reduce staff workload by 40 percent. Whether or not that number is exactly right for your business, the direction is clear: capacity is going to appear.
Now what?
This is where most businesses stall. The capacity materialises, but nobody planned for it. It does not get redeployed — it evaporates. People find ways to fill their day. Meetings expand. Processes get more elaborate. The admin shrinks but somehow everyone is still just as busy.
In a franchise network, this is an even bigger missed opportunity, because the things that actually drive franchisee performance are precisely the things that get squeezed when everyone is drowning in operational tasks.
Think about what your franchisees would do with an extra day a week:
More time with customers. Not processing transactions — actually talking to people, understanding their needs, building the relationships that drive repeat business and referrals. The thing that most franchisees got into business to do, before the admin consumed them.
Better local marketing. Not the corporate campaign that head office pushes out — the local stuff. The community events, the partnerships with nearby businesses, the social media presence that actually reflects their location and their customers. The marketing that franchisees know works but never have time to do properly.
Coaching and development. The team meeting that keeps getting cancelled. The one-to-one with the new starter. The training on the new product line. The conversations that build capability and reduce staff turnover — which, in a franchise network, is one of the biggest hidden costs there is.
Compliance done properly. Not the rushed checkbox exercise at the end of the month, but genuine attention to the standards that protect the brand, the customer, and the franchisee’s own business.
The franchise networks that will get the most from agentic AI are not the ones that deploy it fastest. They are the ones that decide, in advance, where the freed capacity goes. They make it deliberate. They build it into their plans, their KPIs, and their conversations with franchisees.
Because if you automate the admin but do not redirect the time, you have not transformed anything. You have just made your existing operation slightly cheaper to run. And your competitors — the ones who redirected that time into customer experience, into coaching, into growth — will overtake you while you are still congratulating yourself on the efficiency saving.
The franchises that will win
Agentic AI is not a chatbot upgrade. It is a genuine shift in what technology can do inside a franchise operation. But the technology is the easy part.
The hard part is the thinking that needs to happen before and after you deploy it:
Before: Which processes deserve to survive? Which should be retired? Where are the real accuracy benchmarks — not the imagined ones — and what verification do you need?
After: Where does the freed capacity go? Who decides? How do you make sure it flows into the work that actually drives franchisee and customer value, rather than simply disappearing?
The businesses that answer these questions honestly will build franchise networks that are not just more efficient, but genuinely better — better for franchisees, better for customers, and better positioned to grow.
The ones that skip the questions and rush to automate will just be doing the wrong things faster. With 95 percent accuracy.
Colin Rees is the founder of Xpera, where we help franchise networks and multi-site businesses make smarter technology decisions. If you are thinking about where AI agents could make a real difference in your operation — or whether your processes are ready for them — we would welcome a conversation.