AI adoption: where to start when your ideas are confused

Anyone working through their dose of conferences, webinars, presentations, mini-courses and tutorials on artificial intelligence knows the feeling: the more you listen, the less clear it is what to do.

When I meet companies, the same sentence keeps coming back: “After a feast of webinars and conferences, we have fewer ideas, and they are more confused than before.” It even has a name: AI decision paralysis, the state in which management keeps postponing the choice because there are too many options.

In this article I explore why it happens and where to start.

Taking apart what holds us back

The question that holds us back is one: “Why should we do it?” If the answer is “because it’s fashionable”, the question needs rephrasing.

Following fashion is conformism, and conformism pays little.

Behind all the hype there is a paradigm shift, a change that rewrites the rules of work.

To give this perception (fashion) meaning and direction, in my work I usually turn the statement around and uncover three needs, as deep as they are keenly felt.

“We are thinking about it, but it’s not the right time yet”

This is what people say when they postpone: “We’re not ready: our employees don’t use it.” They often believe their company is not ready, that it needs more reflection, another conference, another quote.

Without realising that employees are already using it, behind the company’s back: a phenomenon called shadow AI, employees using artificial intelligence on a do-it-yourself basis, outside any company rule. It brings serious security and compliance problems, from privacy to trade secrets, and I covered it in detail in this article on shadow AI.

Shadow AI is far more widespread than most people think. Employees turn to do-it-yourself tools to work faster or to make a tedious task lighter, even when the company has explicitly banned them. And these employees prove two things:

1. We did not listen to them. They had a need, and they solved it on their own.
2. We are already using artificial intelligence, we just don’t know it.

So the time has already come. The real choice is between governing it and putting up with it.

Why generic courses don’t meet our needs

There are the well-known multi-company courses and the webinars with twenty thousand participants. It is the classic one-to-many approach: one person speaking to many. Someone who doesn’t know us talks about the sector, and the sector is an average that rarely matches the needs of any single company.

Then there is a second limit. Our competitors are sitting in the room, or on the call. We have to be very careful about what we say, because one careless remark can leak confidential information and break compliance. So the real questions stay unasked.

The result? We are watching a documentary that doesn’t reflect our reality. A beautiful documentary about lions, when we live in Milan and there are no lions in Milan. Or a documentary about oil opportunities, when there is no oil where we are. Interesting, well filmed, and useless for deciding what to do on Monday morning.

What we need: in-company training and consulting

In-company training comes into the company and works on its processes, data and opportunities, with its own people and no competitors in the room. To show what it brings to light, I use a framework that is more than twenty years old.

On 12 February 2002 Donald Rumsfeld, then US Secretary of Defense, split knowledge into categories during a press briefing. At the time it raised a few smiles. Today it is one of the most useful tools for a company approaching artificial intelligence, because it forces you to take stock before you go shopping.

1. Known knowns. Common, generic knowledge about artificial intelligence, the market and the sector. The same knowledge everyone has, competitors included.
2. Known unknowns. What artificial intelligence can do for our internal processes, how, and with which tools: what is specific to us. We are aware of it, because we leave the course saying “we’re missing the specifics”.
3. Unknown knowns. Skills, processes and uses of AI we already have in-house. We worked convinced these resources were missing, yet they were there: the consultant brings them out during consulting and training.
4. Unknown unknowns. Opportunities and risks no one imagined. We didn’t know certain things were possible, and we didn’t know our employees, using AI on their own, were giving our secrets away.

A curious detail: Rumsfeld named only three. The third quadrant, the unknown knowns, was added later by the philosopher Slavoj Žižek. And it is precisely the one companies overlook most.

A generic course sees only the first quadrant. In-company training, backed by consulting, sees all four, because what matters emerges in exactly the quadrants the course leaves in the dark.

Not sure who should take the first step?

One knot remains: who does what. There is a poem attributed to Charles Osgood, the CBS broadcaster, about four people named Everybody, Somebody, Anybody and Nobody.

There was an important job to be done. Everybody was sure that Somebody would do it. Anybody could have done it. Nobody did it.

In a company it sounds like this: marketing waits for IT, IT waits for management, management waits for HR, HR waits for the next conference. The matrix tells you what you know; consulting decides who moves, on what and by when.

How I work

The method matters as much as the content, so here it is:

1. Mapping the quadrants with management and senior staff in marketing, sales and HR. This is where shadow AI comes to light and becomes an asset to govern.
2. Owners and scope: one person, one area, one date. That is how Nobody becomes Somebody.
3. Tailored training, cut to the company’s processes. I experiment together with its people on real cases and deliver measurable results, agreed before we start.

Which quadrant is your company starting from? It is the first thing we look at together.