AI has been overpromising and underdelivering for years, and the reckoning is coming. PwC’s 2026 Global CEO Survey found that 56% of CEOs achieved no financial return from AI, and only 12% reported both revenue and cost benefits. In 2026, as mass layoffs hit and promised productivity gains remain elusive, businesses will realise they skipped an essential step: laying the foundations AI needs to deliver. These are the main mistakes I see service businesses make again and again, and how to avoid repeating them.
1. Automating broken operations
In 2025, too many businesses tried to automate before they fixed their operations, deploying AI onto chaotic and undocumented workflows. The assumption being that automation would “clean things up”, but AI can’t fix broken processes. Exceptions still happen, work gets held up, information stays scattered across systems, and teams manually intervene as much as ever.
It’s no wonder that most AI projects are still stuck in pilot, with McKinsey’s 2025 AI report finding that while most businesses are using AI in at least one function, two-thirds are yet to scale it across the business. If you want to see results from automation, you need to do the unglamorous work of understanding your own operations. Map your processes first. Then, you can seal the cracks, start building from steady ground, and begin to scale.
2. Making staff cuts based on unrealistic AI goals
Another common trap is cutting jobs in anticipation of productivity improvements that haven’t arrived yet. Many service businesses have banked on future efficiency gains and jumped the gun, assuming AI would absorb the extra workload quickly.
The reality is that most AI deployments deliver improvements gradually. Teams are then left understaffed and overstretched, while still handling the same volume of work. Instead, AI gains need to be proven, and staff redeployed/upskilled to other areas of the business. Measure impact first and then make resourcing decisions based on evidence, not optimism.
3. Ignoring process debt
Many service businesses have a growing problem that their balance sheet doesn’t track: process debt. That’s the gap between documented workflows and messy operational reality. On paper, work follows clean, predictable steps. In practice, it zigzags across inboxes and spreadsheets, held together by constant firefighting. Often, leaders have no idea what’s actually happening on the ground.
This is why so many automation projects are set up to fail. Agentic AI depends on structure and clarity. These systems need to understand what work exists, how it moves, and when to act. The only way to address process debt is by learning how tasks actually move, and where the problems and breakdowns occur. Document how work actually flows before introducing automation tools.
4. Believing “AI-powered” is enough
“AI-powered” used to be a selling point. In 2026, it isn’t. After too many unmet promises, service operators aren’t impressed by flashy demos or feature lists anymore. They’ve seen too many tools that looked clever but changed very little day to day.
Now they want concrete numbers. How much manual effort does this remove? Which bottlenecks does it eliminate? How much money have businesses like ours actually saved? Solutions that add AI features without solving real operational problems will struggle to get through procurement. Focus on real impact, not potential.
5. Keeping technology decisions locked in IT
Traditionally, IT teams have made decisions about every tool and system. That model no longer works. The pace of operational change is too fast, and decision-making is too far removed from the reality of day-to-day work. More technology decisions need to be made by the customer service managers, finance operations heads, and workflow owners who are closest to the problems.
They understand where the problems are and where automation would actually help. Modern platforms have made this shift possible. You no longer need extensive technical expertise to deploy effective solutions. The CIO role has to shift from gatekeeper to enabler, prioritising governance and security, while empowering teams to improve their own operations.
6. Not understanding what work actually gets done
I regularly ask leaders simple questions: Where does work pile up? Who is responsible for what? Why does this task take so long? Most can’t give clear answers. That blind spot is one of the reasons why some jobs will disappear in 2026. The roles being eliminated never created value in the first place. They existed to compensate for broken operations, like manual email routing or re-keying data between systems.
When those roles go, leaders will be forced to ask why this work existed at all. Once you build visibility into real workflows, you can understand who does what and why. Only then can you decide what should be automated, redesigned, or removed entirely.
The bottom line for 2026
AI isn’t going away, but the excuses are. Service businesses that pull ahead in 2026 will be the ones that stop chasing the shiniest tools and fix their foundations first. That means understanding real work and addressing process debt before automating deliberately. Anything else is just repeating the same mistakes, faster.
