Where do you start with AI in your organization?

The question “where do we start with AI?” leads to a long run-up in many organizations. First draw up a policy, choose a platform, form a working group. Months later nothing is running yet, and doubt grows about whether it will deliver anything.

It can also work the other way around, and that order works better: start with one process, book a visible win there, and use that win to choose the steps that follow. Below you will read how to pick that first process and what you do and do not need to arrange up front.

The pitfall: starting too big

Anyone who takes AI seriously wants to do it properly. That is understandable, and yet it often stalls precisely there. The reflex is to put up the whole structure first: an AI policy on paper, a chosen platform for the entire organization, a working group to drive it. All sensible on the face of it, and all work that takes months before a single employee notices anything.

In those months little is visible. Management asks what it delivers, and there is no answer yet. The energy that was there at the start fades away. AI itself is not the problem here. The problem is that you started building a foundation without first knowing what should go on top of it.

Start with one process, not a plan

The reverse approach: pick one process in your organization where AI can already make a difference today, and let it work there. No policy first and no platform for everyone, just one defined application that you notice results from within a few weeks.

Starting small is not the same as fiddling around without commitment. You deliberately pick a process that matters, you agree on what you want to achieve, and you track what it delivers. The difference with the big approach is that you learn by running something, instead of by thinking everything through in advance. What you discover in that one process about your data, your people and the tool is worth more than a policy document you wrote without that experience.

A first result that people see with their own eyes does something a plan cannot: it removes the doubt. Colleagues who see a tedious job go twice as fast will ask of their own accord whether it can work for their work too.

How do you pick that first process?

Not every process is a good first choice. The suitable candidates have a few things in common:

• It is repetitive. Work that comes back in the same form again and again delivers results faster than a one-off exception.

• It costs noticeable time now. A process that eats up hours every week makes the win visible. With a five-minute task there is little to gain.

• It runs on information you already have. Reports, emails, documents, data from your own systems. The more the input is already in place, the easier the start.

• The stakes are manageable. Start with work where a mistake is no disaster and a person still checks the result. Not with a process that goes straight to the customer or the court.

Think of drawing up quotes from a fixed format, summarizing long reports, pre-sorting customer questions, or preparing recurring reports. Weigh the candidates against each other on two things: how much time the process costs now and how feasible the first step is.

Make that first win visible

Picking a first process is half the step. The other half is showing what it delivers, because that is where you get the support for the next steps.

That starts before you change anything. Measure how long the process takes now, or how much work goes into it. Without that starting point you cannot prove afterwards what the win is, and it stays a gut feeling. A simple measurement will do: how many hours per week, how much lead time, how many steps.

Also agree on when the trial has succeeded. “Twice as fast” or “done in a day instead of a week” is a more concrete goal than “see whether it helps”. After a few weeks you know whether it works, and you have a result you can share across the organization. That result is your best argument for the expansion that follows.

From first win to a course

Once that first process is running and the win is visible, the moment comes when loose trials need a direction. Which processes you tackle next, who becomes responsible, and where do you want to be in a year. That is the point where an AI strategy becomes useful.

The order is deliberate. A strategy you draw up before the first experience is guesswork. A strategy you make after you have seen in one process how it plays out rests on real experience. You then know where AI catches on in your organization, where it chafes, and what your people think of it. That course does not have to be a hefty report; one page with choices and a goal is enough to steer by.

What you do arrange up front: data and agreements

Starting small does not mean skipping everything. One thing you arrange before you put real company data into an AI tool: the agreements around your data.

As soon as personal data or confidential information goes in, the GDPR and the agreements you have with customers apply. With free versions your input may be used to improve the model; in business subscriptions you can often turn that off. Check that per tool, and for your first trial deliberately choose work where the data is not the most sensitive. That keeps the start simple and keeps you on the right side of the rules, without having to write a whole policy for it first. That policy comes later, as part of the course.

How we can help

Do you want to get your management team aligned in a short time on where AI delivers the most and which process lends itself to a first trial? Then the workshop AI Strategy for Managers is the starting point. You go home with a shared picture and a first choice to get to work with.

Do you then want to actually set up that first process, from choice to working application? Then we help with the execution in AI in Business. Together we map where the win is, set it up within your own environment, and make sure the first win is measurable.

Schedule a no-obligation call, and we will look together at where the first win is for your organization.

Frequently asked questions

Sources

• Dutch Data Protection Authority, GDPR preconditions for generative AI (https://autoriteitpersoonsgegevens.nl/en/documents/gdpr-preconditions-for-generative-ai): what the GDPR requires when using AI with personal data.