Let me save you some time. Should you use AI? One hundred percent, yes. Everyone should, for their own productivity at a minimum. If you write emails, summarise documents, plan anything, or make decisions with incomplete information, AI will make you faster today, and it costs less than a coffee habit.
The interesting questions are the next two: where should you use it, and how should you use it well? That is where most people and most businesses get it wrong, and it is what this article is really about.
The mistake almost everyone makes
Most businesses start with the tool. Someone sees a demo, gets excited, and asks "where can we use this chatbot?" That is backwards. You end up with technology looking for a problem, a pilot that impresses nobody, and a team that concludes AI is overhyped.
I start from the opposite end, and it comes straight from my engineering background. Before we talk about AI at all, we look at the business as a system.
The process I use with every client
Step one: look at the whole business. Not the org chart, the actual flow of work. What comes in the front door, what goes out the back, and everything that happens in between. Where does work queue up? Where do things get dropped? What do people complain about on a Friday afternoon?
Step two: map the processes. We write down, step by step, how the important work actually gets done. Not how the procedures manual says it gets done, how it really happens. Who touches it, what software they use, how long each step takes, how often it happens. For some clients we go all the way to a full value stream map, a lean manufacturing technique that traces every step from customer request to delivered value and puts numbers on the time and waste at each stage. It sounds heavy. It is also where the money hides.
Step three: identify the highest value targets. Once the processes are on paper, the opportunities are usually obvious. We score them on a few simple factors: how often the task happens, how long it takes, what it costs when it goes wrong, and whether it sits on a bottleneck that slows everything else down. Frequent, time hungry, error prone, and bottlenecked is the jackpot combination.
Step four: deploy at the top of the list first. Not the easiest target, not the flashiest one, the highest value one. Early wins need to be worth talking about, because they fund the appetite for everything that follows.
Step five: make the experience stupidly simple. This is the step people skip, and it is the one that decides whether the whole thing survives. Once a solution works technically, we ask: how do we make using this so simple that it genuinely makes someone's life ten times easier? If the person doing the job has to fight the tool, they will go back to the old way within a fortnight, and they will be right to. Adoption is a design problem, not a training problem.
Try it yourself this week
You don't need a consultant to run a lightweight version of this. Here is the exercise I would set you:
- List the ten most repetitive tasks in your week or your business
- For each one, note roughly how often it happens and how long it takes
- Note what it costs when it is done late or done wrong
- Circle the two or three with the biggest numbers
- Take just one of them, and in the coming articles I'll show you the tools to attack it
A simple process map with the bottleneck highlighted
One more thing before you race off and automate your top target. There is an order of operations to this, and getting it wrong wastes money. That is the subject of the next article, and it involves a funnel.