When we started Kritmatta, we wanted to make AI useful for people who do not have a technical background. The thing that becomes obvious quickly is that the best place to begin is not the technology at all. It is the work your team already does every day.

If information is hard to find, if people are copying the same details between systems, or if actions quietly disappear after meetings, those are problems worth solving. They are also problems you can describe without using the word AI once, which is usually a sign you have found a real one.

Here are five ways to start.

1. Map the workflow before you change it

Buying an AI tool before you understand the problem leaves too much to chance. You end up with a tool looking for a job.

Ask someone to walk you through a task from beginning to end instead. Where does the information come from? Who needs to approve it? Which steps involve chasing an update, or entering the same details twice?

A simple list is enough. It does not need to be a diagram or a process document. Once you can see the process written down, you can decide which single step is worth improving, which is a much easier decision than choosing a platform.

2. Ask your team where the time goes

The people doing the work usually know exactly where the friction is. They have been working around it for months.

Ask each person a direct question: if I could take one repetitive task off your plate, what would it be? Then estimate how long that task actually takes.

The arithmetic is worth doing properly. In a team of ten, losing just one hour per person each day adds up to 50 hours a week. That is more than a full-time role's worth of time, spent on work nobody chose to do.

Thinking in hours makes the opportunity much easier to grasp than an abstract efficiency percentage. Nobody has a feel for what a 12 per cent improvement means. Everybody understands 50 hours.

3. Start with one small, measurable task

It is tempting to build an elaborate automation straight away, because the elaborate version is the one that sounds impressive.

Start with something you can compare against the current process instead. Arranging interviews. Preparing a customer quote. Something that happens often enough to measure and matters little enough to get wrong once.

Record how long it takes now. Test a new approach. Check the quality of the result after a week, then again after a month. If it does not help, adjust it or move on.

You learn faster when the first test is manageable. A small test that fails costs you a week. A large one that fails costs you the appetite to try again.

4. Turn information into actions

A meeting transcript that nobody reads has limited value. The recording was never the point.

Make sure notes become a task list, with owners and next steps attached. The same principle applies to email: a useful daily summary should separate what needs action from what is only there for reference.

When I came back from holiday to more than 400 unread emails, I did not want a summary of all 400. I asked for a digest of the priorities first. That gave me somewhere useful to begin, which is a different thing from being told what happened.

That distinction is most of the work. Information tells you what occurred. An action tells you what to do about it, and who is doing it.

5. Keep reviewing what matters

Automation needs attention after it is set up. This is the step most often skipped, because switching something on feels like finishing.

Check whether it is still saving time, still producing reliable work, and still helping customers. For customer enquiries, an initial response may be quick to automate, while a complex question still needs a person to answer it properly. Knowing which is which is a judgement call, and it changes as the business changes.

Underneath all of it, I use a simple daily check-in. In the morning: what is the most important thing I can move forward today? At the end of the day: what did I actually move?

That habit does more for workflow efficiency than most software does, because it keeps the whole thing pointed at results rather than at activity.

This post is adapted from an article by Kritmatta co-founder Kalpesh Baxi, published by Business Advice in September 2026.

Simple AI. Real results.