I watch businesses do the same thing every quarter. They buy a clever new tool, use it twice, and let the subscription rot. Then they decide AI does not work for them.
The tool was rarely the problem. The workflow around it was.
A tool is a hammer. A workflow is the house you are trying to build. Buying more hammers does not get the house built, and most teams own a drawer full of hammers they never picked up again.
Start from the friction, not the feature
Most AI projects begin with the wrong question: “what can this tool do?” The better question is “where does our week get stuck?”
Map the work as it happens. Watch where things wait, where someone re-keys the same data, where a job sits in an inbox until a person finds time. Find the step everyone on the team dreads. That dread is the signal. It marks the place where a small change pays back fast.
Build the smallest thing that helps
You do not need a transformation programme to start. You need one painful step to hurt less.
Pick a single workflow. Strip it back to what it is meant to produce. Then put automation around the boring middle: the copying, the formatting, the chasing, the first draft. Ship that one improvement, measure the time it gives back, and use the result to fund the next one.
A worked example. A property firm I worked with spent four hours a week turning enquiry emails into a spreadsheet by hand. We did not buy an “AI platform.” We connected the inbox they already had to the sheet they already used, and let a model pull out the fields. Four hours became ten minutes. They got their afternoon back.
Why this beats buying
Tools change every month. Workflows change far less often, because they map to how your business earns money. When you fix the workflow first, the tool becomes a detail you can swap later. When you buy the tool first, you bend your business around someone else’s product and hope it fits.
The businesses pulling ahead found the three or four jobs that ate their week and made those jobs smaller.
Where to begin this week
Open your calendar and your team’s. Look for the recurring blocks that produce admin rather than value. Write down the top three. For each one, ask what a good version of that work would look like if the dull part disappeared.
You now have a backlog worth more than any tool demo. Start at the top, build the smallest fix that helps, and keep going.