Three times this month somebody has asked me whether CCG “does automation”. The first time I gave a full answer. The second time I gave a shorter one. By the third I had stopped answering altogether and started asking a question back, because something in the way that question is put had been bothering me for weeks and I could not name it.
I can name it now. The question is almost always about the technology. It is almost never about the work.
What the business automation industry actually is
It is worth being precise here, because “automation” has become one of those words that means whatever the person saying it needs it to mean.
The business automation industry is a real and layered thing. At its base sits robotic process automation, software that drives the same screens and keystrokes a person used to drive. Above that sits workflow and business process management, which handles the routing, the approvals and the handoffs between teams. Alongside both runs integration, the unglamorous plumbing that lets one system speak to another. Then there is document and email processing, which reads the unstructured material that used to land on somebody’s desk. And now, sitting on top of all of it, agentic AI, which does not follow a script at all. It decides.
Around that technology sits an entire economy. Vendors and licences. Implementation partners. Certifications. Centres of excellence with their own governance forums and their own quarterly reporting. A sales motion that is, in fairness, extremely good.
None of that is a criticism. The tools work. Some of them are genuinely remarkable, and a process that used to take a team four days now finishes before the coffee does. We are not sceptics at CCG. We have watched clients take real cost and real drudgery out of their businesses this way, and we have been glad to help them do it.
But here is what had been bothering me. What the industry sells is capability. What a client actually wants is a changed organisation. Those are two different purchases, and the gap between them is where the money quietly goes missing.
The industry has solved the technology. It has not solved the adoption
We are not guessing about this. The 2026 AI and Data Leadership Executive Benchmark Survey put the question to senior data and AI executives at nearly 110 Fortune 1000 companies and leading global brands. The answer was blunt: 93.2% named cultural challenges as the greatest impediment to adoption. Only 7% pointed at the technology.1 Read that the other way round for a moment. Fewer than one in ten of the people whose entire professional life is technology believe that technology is their problem.
We are not surprised. We would have been surprised by the opposite.
Automating a process does not remove the work, it moves it
This is the point that gets missed in the demonstration, and it gets missed for a simple reason. The demonstration is always of the happy path.
When you automate a process you are not deleting effort. You are relocating judgement. And judgement, unlike effort, does not sit quietly wherever you put it. Every automation programme we have worked on eventually arrives at the same four questions, usually later than it should have:
- Who owns this process now that nobody actually runs it? Ownership does not disappear simply because the running does.
- What happens to the exceptions? The awkward tenth that the machine cannot handle is now the entire job of the team that remains, and it is by some distance the hardest tenth.
- Who is accountable when the output is wrong, and how would anybody even know that it was?
- What does a supervisor now supervise, and has anyone taught them how?
Not one of those is a technical question. Every one of them is a question about how the organisation is designed and how its people work. Which is to say every one of them is change work, and none of it comes in the box.
In our experience the programmes that disappoint rarely fail at the build. The build is usually fine. They fail three months later, quietly, when the exceptions pile up, a workaround appears, and somebody in the team starts keeping the old spreadsheet “just to be safe”. Nobody reports that to a steering committee. It simply becomes the way things are done, and the benefits case dies of natural causes.
Automating a process you have not first understood is like buying a faster car to shorten a journey you should not be making.
This is no longer a pilot conversation
The same survey shows how far the technology itself has already travelled. Firms reporting AI in production at scale rose from 4.7% two years ago to 39.1% this year. Limited production rose from 24.5% to 54.5% over the same period.1 The tools are in the building. Which means the people questions are not theoretical any more either, and an organisation that has not answered them is not being prudent. It is being exposed.
Why we still call this strategy implementation
At CCG we have argued for years that “change management” is a misnomer, and automation programmes are the cleanest illustration of the point we have had in a long while. Nobody needs their feelings managed about a piece of workflow software. What they need is for a new way of operating to be designed, decided, resourced and led. That is strategy implementation, and it is a different discipline with different demands.
We are a bunch of straight-talking, no-nonsense people, so let us be plain about the sequence, because the sequence matters at least as much as the effort:
Design the operating model before you buy the tool. That is the work of CCG Consult, which turns an automation ambition into changed processes, defined roles, escalation paths and clear ownership, so that the technology has somewhere to land.
Find out where you actually stand, which is rarely where the pilot said you were. That is CCG Analytics, measuring readiness, adoption and sentiment so that decisions rest on evidence rather than on the confidence of whoever ran the demonstration.
Build the judgement that the automation now depends on. That is CCG Learn, developing the capability people need in order to supervise, question and override these systems, which is precisely the work that is left once the routine work has gone.
Settle accountability at the top. An automated process with no owner is not efficient. It is unattended. That is CCG Advisory, working with Boards and executives on the oversight and ownership that decides who answers for the output.
So here is the answer I gave the third person who asked, and it is the one I have settled on.
Yes, we work on automation programmes. No, we will not sell you the robot. There are excellent firms who will do that, and they will do it better than we would. What we will do is make sure that when the robot arrives somebody owns the process, somebody can handle the exceptions, somebody is accountable for the output, and the people who remain understand what their job has just become.
That is the half of the programme that nobody ever demonstrates. It is also, on the evidence of the very people running these programmes, the half that decides whether any of it works.
If you would like help with Doing Change Better, we would welcome the conversation. Get in touch with us.
