Most businesses we speak to have tried AI the same way: a chat window, a few impressive answers, and then nothing changed. The concern isn't the technology. It's that AI sitting beside the work rarely changes the work.
The AI that pays off is quieter. It's connected to your own records, does one specific job well, and lets a person check what matters.
Where AI genuinely earns its keep
Paperwork that reads itself
Invoices, purchase orders, forms, receipts: if your team types information from documents into a system, AI can read the document and fill in the fields, with a person checking only the exceptions.
Answers from your own records
Staff spend hours asking each other "where's that policy?" or "what did we quote this client?" An assistant that searches your documents, not the open internet, and shows where each answer came from, saves that time without guessing.
Follow-ups that don't slip
Drafting the reminder, summarising the call, preparing the next quotation. AI is good at the first draft, so people spend their time on the judgement.
Sorting and routing
Which enquiry is urgent? Which department should handle this request? Classifying incoming work is a strong, low-risk first project.
Where not to start
- Anything where a wrong answer is costly and nobody checks it. Keep a person in the loop for money, health and legal decisions.
- A public chatbot as your first project. It's the most visible, the hardest to get right, and the easiest to embarrass you.
- Concerns you can't measure. If you can't say how long the task takes today, you won't know if AI helped.
Three rules that keep it safe
- Permissions first. AI should see only what the person using it is allowed to see, and act only where you've allowed it.
- Show the source. Every answer should point to the record it came from, so people can check it.
- Humans approve the important steps. AI drafts; people decide.
A 30-day plan for your first AI project
Week 1 — Pick one task. Choose something repetitive, measurable and low-risk. Write down how long it takes today and how often it happens.
Week 2 — Connect the data. Give the AI access to only the records that task needs, with clear permissions.
Week 3 — Run it alongside people. The AI does the work; your team checks every result. Note what it gets right and wrong.
Week 4 — Decide with numbers. Compare time saved and error rates. Keep it, adjust it, or stop. Then choose the next task.
The takeaway
The question isn't "how do we use AI?" It's "which hour of work should never be done by hand again?" Start there, connect AI to your own records, keep people in charge of the decisions, and let the results earn the next project.
Want to find your best first AI project? Book a strategy call and we'll look at your work with you, and tell you honestly where AI will and won't help.
- Start with one repetitive, measurable, low-risk task, not a public chatbot.
- Connect AI to your own records, with permissions and sources shown.
- AI drafts, people decide: keep a human on every important step.