I've sat in a fair few boardrooms over the years — banking, insurance, higher education, manufacturing — and I've noticed the same conversation happening again and again lately, almost word for word.
“What should we be doing about AI?”
And nine times out of ten, the answer starts with technology.
Which model should we use? Which platform should we buy? How do we integrate it? How do we secure it?
Before you know it, the whole thing has landed on IT's desk, and everyone's rather relieved to have handed it over.
I understand the instinct. But I think we're still putting AI in too small a box.
AI is a technology — but its implications go much further
We tend to describe AI as a technology, and that's perfectly reasonable. But AI is also a multidisciplinary capability: part science, part technology and engineering, and increasingly intertwined with business, data, people, processes and governance.
The science gives us machine learning, neural networks, language models, computer vision, speech recognition and increasingly capable systems that can reason, plan and use tools.
Technology makes those capabilities available.
Engineering makes them reliable enough to put into production.
But none of that, on its own, creates business value.
For that, we have to think about the business itself: what we do, how we do it, how decisions are made, how customers are served, how people work and where value is actually created.
And that's where things get interesting.
Not the first technology to change the nature of work
There is a tendency to describe AI as fundamentally different from everything that came before it.
I'm not convinced that's quite right.
Technology has been changing the nature of work for centuries.
The personal computer didn't merely help typists type faster. It fundamentally changed clerical work.
The spreadsheet transformed financial analysis.
The internet didn't simply accelerate communication — it changed distribution, commerce and entire industries.
Travel agents, newspaper publishers, bank branches, manufacturing workers and countless other occupations have been reshaped or displaced by successive waves of technology and automation.
So AI isn't the first technology to transform work.
What is different is the breadth of the capability it is beginning to automate.
Previous digital technologies largely automated the handling, movement and processing of information. AI increasingly operates on the information itself — interpreting language, analysing documents, generating content, writing software, identifying patterns, making recommendations and carrying out bounded tasks.
That potentially reaches much further into knowledge work.
So perhaps the better analogy isn't caffeine versus steroids.
It's this:
Digital technology has progressively automated more of the business process. AI may accelerate that progression into areas of work we previously assumed required human cognitive effort.
That's significant enough without pretending it has no precedent.
The real question isn't “How do we use AI?”
For many organisations, the first stage of AI adoption will be relatively mundane.
People will use AI to write emails, summarise meetings, analyse documents, generate presentations, research subjects and write code.
Some will call this superficial adoption.
I wouldn't.
It's actually an important part of the process.
Technology adoption has rarely happened through perfect top-down planning. People experiment. They discover things. They find shortcuts that management never imagined. They expose inefficient processes simply by trying to work around them.
A person using an AI assistant to answer customer enquiries might discover that 40% of those enquiries are essentially the same.
That may lead to a knowledge base.
The knowledge base may lead to automated responses.
Those responses may eventually become an AI customer advisor integrated with the CRM, quotation system and scheduling platform.
The original productivity experiment has now become process transformation.
Bottom-up experimentation and top-down strategy aren't alternatives. They should reinforce each other.
The mistake is not giving people AI tools.
The mistake is giving people AI tools and then assuming the job is done.
AI strategy isn't an IT strategy — but it isn't separate from IT either
This is another distinction I think we need to reconsider.
It is tempting to say:
The business decides what to do. IT decides how to do it.
That division made considerably more sense when technology was largely an internal service function.
It is becoming less useful.
With AI, the architecture can directly affect the business proposition.
Consider a customer-facing AI advisor.
Whether it can access customer records, make recommendations, generate quotations, arrange appointments, take payments or hand a case to a human isn't simply an IT decision.
Those are business decisions.
But they are also technology decisions.
The security architecture determines what the AI is allowed to see.
The data architecture determines what it can know.
The model determines what it can understand and generate.
The integration architecture determines what it can actually do.
The governance framework determines what it is permitted to do autonomously.
And the business strategy determines why any of this should exist in the first place.
So I don't think AI should be owned by either the business or IT.
AI increasingly sits at the intersection of business strategy and technology strategy.
The organisations that get this right will bring the two together rather than choosing one over the other.
Capability isn't the same as reliability
There is another trap we should avoid.
We sometimes talk about AI as though we have created a new form of autonomous intelligence that can simply be handed a business problem and left to solve it.
We're not there.
Today's generative AI systems can display remarkably sophisticated reasoning-like behaviour. They can analyse, infer, plan, code, compare alternatives, use tools and adapt their responses to context.
But capability does not automatically mean reliability.
These systems can hallucinate.
They can produce confident but incorrect answers.
They can reproduce bias.
They can fail unexpectedly on seemingly simple problems.
And their performance can vary depending on the data, context, instructions and system architecture surrounding them.
That matters enormously when AI moves from assisting people to acting on behalf of organisations.
The question therefore isn't simply:
“Can AI do this?”
It is:
“Can AI do this reliably enough, within clearly defined boundaries, with appropriate oversight and consequences?”
That's a much more useful enterprise question.
From assistance to agency
Perhaps the most interesting transition is the one from AI as an assistant to AI as an active participant in a business process.
Imagine a traditional customer journey:
Customer → employee → software → employee → manager → customer
AI can potentially change that into:
Customer → AI → business systems → AI → customer
with people becoming involved where judgement, accountability, relationships or exceptions genuinely require them.
That's where the economic implications become much more interesting.
The question stops being:
“How can AI help our employees work faster?”
and becomes:
“Which parts of our work should now be performed by people, which by machines, and which by a combination of both?”
That is a much bigger strategic question.
And with greater capability comes greater responsibility
The more deeply AI becomes embedded in a business, the more important governance becomes.
AI can amplify good processes.
But it can also amplify bad ones.
It can scale a useful decision — or scale a flawed one.
It can reduce human error — or introduce a different kind of error at extraordinary speed and scale.
So ethics, governance, security, privacy, resilience and compliance cannot be something we bolt on after implementation.
They need to be designed into the system.
But even here, I would resist reducing the discussion to:
“Are we compliant?”
Compliance establishes boundaries.
Leadership judgement determines what we should actually do within those boundaries.
Trustworthiness is therefore not merely a regulatory requirement. It is a business requirement.
So whose job is AI?
Perhaps that is the question worth asking in the boardroom.
Not:
“Which AI technology should we buy?”
And not even:
“How can we introduce AI into the business?”
But:
“What could our business look like if machine intelligence became a scalable organisational capability?”
IT has a critical role in answering how.
The business has to determine why, where and what.
And increasingly, those conversations cannot be separated.
AI is not the first technology to transform work.
It isn't magic.
It isn't a replacement for judgement.
And giving everyone a chatbot does not constitute transformation.
But neither is AI simply another piece of software to install.
It represents the latest — and potentially one of the most consequential — stages of the digital transformation of work.
The organisations that benefit most won't necessarily be the ones with the most impressive models or the largest number of AI tools.
They will be the ones that learn to combine human judgement, machine intelligence, technology, data and redesigned processes into a coherent business capability.
And, importantly, they will learn to do that responsibly.
Technology has been helping us work faster for generations.
AI may increasingly change who — or what — does the work.