Map the work before you buy the tool

The most common way organisations waste money on AI is by going straight to buying. A vendor shows a polished demo, everyone gets excited, a contract is signed, and six months later the tool sits unused because it never fit how the work actually happens. The fix is simple and unexciting: map the work first, then choose a tool, if you still need one.

Pick one piece of work that happens often. A customer sends an email with a problem and your team replies. Write down every step: the email arrives in a shared inbox, someone passes it on, that person reads it and looks up the customer, drafts a reply, asks a manager if it is tricky, the manager edits and sends, someone records what happened. Seven steps. For each one ask how often it happens, how long it takes, and what makes it slow or wrong.

Ask the people who do the steps, not only the people who manage them. You will find the bottleneck, and it is rarely where anyone expected. If the slow part is finding context, you do not need a chatbot, you need a better way to search what you already have. If it is drafting, you need help writing replies that bring in the right context. If it is deciding what counts as tricky, you need sorting, not writing. Many support chatbots deal with the first two steps, which were never the problem. That is why they disappoint.

Vendors sell what they have. A chatbot company will offer a chatbot whether or not your bottleneck is somewhere else. That is not dishonesty, it is simply what they make. The map protects you. When you know the bottleneck, you can sit through any demo and say: show me how you help with step three.

A team of fifteen answering customer questions for a distributor was close to signing for a full platform, chatbot and all. An afternoon spent mapping a week of emails showed the chatbot would answer questions that were never slow. The real drag was that people looked for answers in three places, and every tricky email waited for one long-serving colleague. What that team needed first was one place to search and a way to write down what she knew. Smaller, cheaper, and tried within a month.

Buy the tool when you know what it replaces. Do not buy it to find out.