When a leadership team first talks to us about AI, the conversation tends to have two halves.
The first half is about capacity. Reply to enquiries faster. Take on more work without adding headcount.
The second half often arrives unprompted, and with more feeling. It is a list of what must not change: a call that a person always makes, a sign-off the founder still gives, the way the company sounds when it writes to a client. Leaders will defend one small step in their process at length.
Most AI plans treat that second list as a constraint to work around. We think it is the plan’s first page.
Most AI spending starts from the wrong end
A common sequence is tool first: buy the licences, then look for places to use them. Returns have been modest. When IBM surveyed 2,000 CEOs in 2025, they reported that only 25% of their AI initiatives had delivered the return they expected over the previous few years. In the same survey, 64% acknowledged that the risk of falling behind drives investment in some technologies.
Employees see it from the other side. In a Gallup survey of US employees published the same year, 44% said their organisation had begun integrating AI. Only 22% said it had communicated a clear plan or strategy for doing so. That leaves a lot of people holding a tool and working out for themselves what it is for.
Our read is that both findings describe the same gap. A team that doesn’t know what a change is protecting can easily assume it is protecting nothing, their jobs included. This is how a first rollout stalls. Nobody refuses. People carry on working the way they did and the new tools go unused.
The “do not touch” list shows you where trust is made
In a business that grows on trust, a handful of moments carry the relationship. The first real reply to an enquiry. The conversation where a client’s actual problem gets named. The judgement call on what is good enough to send. These are what clients are paying for, even when the invoice says something else.
People are cautious about AI. A 2025 study by the University of Melbourne and KPMG of more than 48,000 people in 47 countries found that only 46% are willing to trust AI systems. That is the general public, and your clients are part of it. When you insist that a person makes a particular call, you are pointing at where your company’s trust is made.
That is why we ask for the list early, and from more than one person. Founders name some moments. The people closest to clients name others. The list is rarely written down, which is a risk in itself: what makes the company distinctive lives in a few people’s habits. Turning it into standards the whole team can carry is the work of our mission Scale your signature.
Human work shouldn’t be the leftovers
There is a quieter mistake on the other side. A plan that starts with “what can we automate?” defines people’s work as whatever remains once the automation is finished. It was never a decision. It is what was left.
We would rather a leadership team chose it on purpose. Three questions help draw the line:
- Where does a client decide to trust us? Keep a person there, and give them more time for it.
- Which judgements would we have to explain if they went wrong? Those stay with someone who can be accountable for them.
- Where do our people learn the standard? If AI writes every first draft, juniors can lose the practice that used to build their judgement. Decide where they will get it instead.
Everything outside those lines is a candidate for AI, and there is usually a lot of it.
Then put AI to work around the people
Protecting a moment doesn’t mean leaving it alone. The reason to draw the line is to make those moments better.
Take the call a person always makes. Before it, AI can pull together everything the company already knows about the client, so the caller starts informed instead of from scratch. Afterwards it can draft the follow-up and update the record. The human part gets more of the person’s attention because the rest takes less of it.
This is what we mean by co-intelligence: AI that knows your business, your clients and your voice, working beside people whose judgement and relationships are the reason clients stay. A fast first reply is worth having. A fast first reply that sounds like everyone else’s is worth much less.
It also changes what you can say to your team. “Here is what stays yours, and here is what we are taking off your plate” is a different announcement from “we are rolling out AI”. People can work with the first one. They can also tell you where you have drawn the line wrong, and you need to hear that before the budget is set.
Try this before you plan next year
If you are setting an AI budget for 2027, do one thing first. Ask each person on your leadership team, separately, to write down the three moments in a client’s experience they would refuse to hand to a machine. Then compare the lists.
Where they agree, you have the start of your standards. Where they disagree, you have found the conversation to have before any money is spent.
Sources
- IBM Institute for Business Value, IBM Study: CEOs Double Down on AI While Navigating Enterprise Hurdles, 6 May 2025. A survey of 2,000 CEOs in 33 countries.
- Gallup, AI Use at Work Has Nearly Doubled in Two Years, 15 June 2025. A survey of employed adults in the United States.
- University of Melbourne and KPMG, Trust, Attitudes and Use of Artificial Intelligence: A Global Study 2025. More than 48,000 people in 47 countries.