Applied AI
Your first AI project should solve a boring problem
17 October 2026 · FinAnalytic Editorial · 2 min read
A proposal arrives with missing information. Someone searches old emails, opens three spreadsheets and asks a colleague for the latest document. That is a useful starting point for an AI project.
The current signal
An August 2026 research paper on organisational ChatGPT use found activity across writing, technical work, communication and information synthesis. It describes adoption patterns; it does not establish that every deployment improves productivity.
Our view: the first question should be which piece of work is worth improving.
Pick one task you can finish
Consider a supplier-document checker. Its job is to identify missing fields, point to the relevant source and prepare a review list. Payment approval stays with a person.
That scope gives a team something concrete to evaluate. An assistant expected to handle every procurement question is much harder to judge.
Count the checking time
Compare the current process with the pilot using representative examples, including incomplete documents and awkward exceptions. Measure total time to an accepted result, including corrections and human review.
If the draft takes seconds but checking it takes longer than doing the task manually, the workflow still needs work.
Make the next decision small
Agree what success looks like before the pilot: acceptable accuracy, permitted data access, review responsibility and operating cost. Expand only after the team can explain what improved.
A useful first project gives you a better process and evidence for the next investment.