The CEO has returned from an event where everyone was already using artificial intelligence. A vendor has a solution. Employees are quietly trying it for themselves. The head of IT must turn all of this into something that will still work when the presentation is over.
I recently dug out two pieces I wrote in 2013. One about the future of IT, the other about Bine, an IT manager who measured his worth mainly by how many things stopped working without him.
Rereading some of my predictions made me smile. With Bine, I laughed a little less. The character was intended as satire even then, but his way of thinking seems durable enough to survive several generations of technology.
Let us leave Bine at his whiteboard for a moment. The more interesting question is what we expect today from the person who connects technology and management.
We might call them the CIO, IT director or head of IT. In a smaller company, they may have a less impressive title and a much longer list of duties. The essence of their role becomes clear when someone brings a big technology promise into the boardroom.
Today, artificial intelligence is the easiest setting in which to imagine that conversation.
“We need to do something,” the CEO says. “Our competitors are already using it.”
The CIO nods.
“For what, specifically?”
“Well, that's for you to tell us. You're the IT person.”
And so the technology gets an owner before the business problem even gets a name.
Competitors usually deliver their best results over lunch
Technology enthusiasm spreads in an interesting way. It starts with an idea, moves on to a presentation and, by the next business gathering, appears as something every serious player is already using successfully.
Just think of the conversations about blockchain. The word could travel much further than any understanding of the problem it was supposed to solve. Enthusiasm was enough to join the discussion; the details were someone else's job.
With artificial intelligence, the temptation is even greater because the first encounter is so direct. Type a question, get a sensible answer. Request a summary, receive a summary. Something that used to take time is suddenly done before your eyes.
That is convincing. Justifiably convincing, too.
The trouble begins when we jump from that experience to the conclusion that we can sort out an entire department in the same way.
There is considerable work between “the tool produced a good answer” and “the company can rely on this process”. In a presentation, that distance often disappears between two slides.
The same applies to stories about competitors. “We've already implemented it” can mean successful everyday use. It can also mean an experiment in one department, a few purchased licences or a project whose costs nobody has properly added up.
There is no need to assume everyone is lying. It is enough for each person to tell the best part of their story.
The CIO's job is to establish which part of that story could create value in their own company.
That takes curiosity, some technical knowledge and enough composure not to compete in a contest of enthusiasm.
Between the gatekeeper and the resident evangelist
I see two extremes in responses to new technology.
On one side is the IT leader who begins by listing reasons we do not need it. Security, compatibility, maturity, staffing, other priorities. Sometimes they are right on every point. Sometimes legitimate professional concerns also provide a rather comfortable defence of their own way of working.
On the other side is the enthusiast who returns from a presentation almost acting as the vendor's salesperson. They see possibilities, connect ideas and want to start immediately. Questions about maintenance, accountability and costs seem annoyingly petty given the historic nature of the moment.
The first can close off useful paths for the company. The second can lead it down those paths without checking how to get back.
These are not necessarily two types of people. They can be two responses from the same person: cautious about unfamiliar technology and surprisingly forgiving towards the technology they have fallen in love with.
That is why I believe a CIO needs an approach to assessment that can withstand their own enthusiasm.
If they demand evidence of benefits from colleagues, they must demand it for their favourite solution too.
Start with the work we want to improve
Imagine a company that wants to use AI to prepare quotations faster.
That is already a much more useful starting point than “we're going to deploy agents”.
Now we can examine what actually happens with quotations. How long does preparation take? Where do people look for information? What do they copy manually? Who checks prices and special terms? Which exceptions cause the most trouble?
We might discover that employees spend most of their time searching documents. Perhaps they are waiting for a discount approval. Perhaps they have to enter the same information into three different applications.
Those are three different problems. It is not a given that they all need the same solution, or that they all need AI at all.
The CIO must be able to bring the discussion down to this level. They need colleagues from sales, finance and operations, because they cannot understand the actual work from an application description alone.
And someone on the business side must take responsibility for the outcome.
If sales wants faster quotations, it must help define what counts as a good quotation, how much checking is needed and which errors are unacceptable. That cannot be handed over entirely to IT, only for someone to conclude when trouble arises that the solution “doesn't understand our business”.
Understanding the business starts with the people doing the work.
The human voice is a user interface. Accountability is still ours.
A distinctive feature of conversational tools is how quickly we can start judging them as conversational partners.
The answer flows. The sentences are well formed. The tone is confident. The tool understands the question well enough to make us feel that it also understands every consequence of the answer.
Here, the CIO must pay particular attention to the difference between a pleasant user experience and a dependable business process.
For a quotation, polished wording is not enough. The price must be correct. The discount authorised. The delivery date achievable. The customer's information used appropriately.
If the system also sends the quotation, the consequences differ from merely drafting it for review. If it can change a price, confirm an order or initiate a payment, the difference is greater still.
The word “agent” does not remove those distinctions.
The scope of permissions must reflect proven reliability and the consequences of error, rather than the persuasiveness of the performance.
Telling employees which tool they may use is therefore not enough. They need to understand what it is useful for, where it needs review and when they must not hand over a decision.
Even the instruction “a human should check it” can be rather empty. Which human? With what expertise? How much time will they have? Can they access the information needed to verify the result?
If the system produces more content in a few minutes than that person can meaningfully review in a working day, we may chiefly have automated the creation of a new backlog.
When the presentation ends, the service begins
In a demonstration, we see a question and an answer. Behind them may be a whole chain of components: a model, a data connection, a search engine, an external service, access permissions and rules governing the process.
None of this is inherently bad. This is how solutions are built.
What matters is knowing who understands the whole and who will act when part of the chain changes or fails.
Who will notice that a connection still works but returns incomplete data? Who will check the effects of a new model version? Who will be able to explain why the system produced a different result from last week?
Here, the CIO must ask the vendor questions that feel slightly awkward during a successful presentation.
What will we have to maintain? What will you maintain? How will we monitor quality? What happens when an external service is unavailable? Which dependencies are we taking on?
A good vendor does not need to answer every question in the first minute. But they must distinguish between what they have verified, what they assume and what they do not yet know.
If every unknown is dismissed with “the platform will handle that”, what we have received so far is mostly reassurance.
Time saved is not yet money saved
When assessing benefits, we quickly become very precise where the arithmetic is easiest.
If a tool saves ten minutes per task, we perform the task a hundred times and multiply the result by an hourly labour cost, we get an attractive number.
A business decision needs a few more answers.
How much time will review take? How many corrections? How many exceptions? What will employees do with the time freed up? Will we handle more work, shorten response times or improve quality?
All of these can be real benefits. They do not have to be dressed up as headcount reductions to look sufficiently businesslike.
On the cost side, there is more than the subscription too. There is integration, data preparation, monitoring, user support and adapting to changes.
I would therefore expect the CIO to provide estimates for normal use, heavier demand and a less pleasant scenario. I would expect finance and management to help assess which benefits actually matter to the company.
Unknowns are acceptable. The problem begins when we assign them a value of zero in the business case.
Caution must come with a next step
After all this, it would be too easy to say that the CIO's main task is to slow down hasty decisions.
Indecision has consequences too.
If every initiative receives an excessively long list of objections, employees and departments have a good reason to find their own routes. The head of IT can retain a sense of control while actual technology use moves beyond their field of view.
That is why a warning must come with a practical next step.
For quotation preparation, that might mean a limited trial: selected types of quotations, well-organised data, a defined budget and mandatory review before sending. Before starting, we agree what to measure and what result would justify continuing.
We test the awkward cases too. A missing price list. Conflicting data. A special agreement with a customer. A request the system should not carry out.
A successful trial gives us a reason to expand. An unsuccessful one must allow us to stop or change direction.
A project that cannot be stopped because we have already boasted about it too much has lost an important part of its controls.
What should we expect from a modern CIO?
I do not expect someone who can predict every change in technology. Nor someone who personally understands every detail of every tool.
I do expect them to bring the right people together, clarify the purpose, expose dependencies and present a decision to management in a way that makes its less glamorous side clear too.
Alongside “we can do this”, they should explain what it requires, who will be accountable and how we will know whether it was worthwhile.
I also expect management to make that role possible. It is difficult to demand a strategic partner if you involve them only after the solution has been promised, the deadline set and the vendor invited to sign.
The beauty of this work is the moment when technology genuinely makes people's work easier. When sales can respond faster, a colleague stops copying the same data repeatedly and the company becomes able to do something it could not do before.
The trap is that the novelty quickly takes the credit, while responsibility for its consequences stays with the people who must live with it.
That is why I would end the opening conversation differently.
“Our competitors are already using it,” says the CEO. “When will we?”
“I suggest we start with quotations,” the CIO replies. “Sales will define what needs to improve. Together, we'll check the results, costs and errors. Then we'll decide how much work we can entrust to the system.”
The CEO looks at them.
“So you believe in AI?”
“Enough to test it. And I take my job seriously enough not to declare success in advance.”