AI demos are built to be seductive. A neat screen. A clean example. A confident person clicking through a result that appears right on time.
Then your team buys the tool and discovers the demo was not your Tuesday. Your work has odd source material, local language, half-finished notes, board preferences, public records, and three people who all mean something different when they say "make it concise."
This is why I want teams to do one boring thing before they buy. Not write down the work. Talk through the work.
The short version
Before buying AI software, pick one small workflow and record yourself explaining the steps out loud. If the recording is longer than three minutes, the workflow is probably too big. Use the transcript as raw material, then ask your preferred LLM to turn it into a documented SOP for your organization.
Name the task before the tool
Pick one job you want AI to help with. Not "communications." Not "research." A real task. Drafting a business visit summary. Preparing a company brief. Turning meeting notes into board-ready next steps. Comparing grant requirements against a project idea.
Then open the voice recording app on your phone. Ideally, use one with transcription. Talk through the process as if you were explaining it to a new person on your team. Where does the information come from? Who touches it? What gets copied? What gets checked? Where does the work slow down? What makes the final version acceptable?
If your recording is longer than three minutes, you probably picked a process that is too big. Good. That is useful information. Break it down into a smaller chunk and try again.
The tool should pass the workflow test
Once the transcript exists, give it to your preferred LLM and ask it to turn the rough explanation into a documented SOP for your organization. Not a fancy binder. Not a six-week process-mapping project. A first draft your team can correct.
Now the buying conversation changes. You are no longer asking, "Does this have AI?" You are asking, "Can this help with step three without breaking step five?"
That is a much better question.
Maybe the tool is worth buying. Maybe the cheaper answer is training your team to build the workflow with tools they already have. Maybe the real problem is not software at all, but a process nobody has cleaned up in five years because everyone was too busy surviving it.
Buy last, not first
I am not anti-tool. I am anti-confusion. Buying first can feel decisive, but it often delays the harder work of naming what your team actually needs.
Record the workflow. Keep it under three minutes. Turn the transcript into an SOP. Find the human judgment points. Then decide what software, training, or outside help belongs in the picture.
You may still buy the tool. You will just buy it with your eyes open.