Not every repeated task needs AI.
You want less rework, not another tool to look after. Start by separating what always follows the same rule from what needs understanding, drafting or judgement.
Discuss my taskThe simple distinction
A template preserves a useful structure. Automation follows agreed rules. AI can interpret varied information or prepare a draft, but can make mistakes. An integration connects systems. These are different contributions, not a ranking from old to modern.
One enquiry. Four possible contributions.
A customer asks for a proposal. A template prompts the necessary information. A fixed rule routes the enquiry to the right person. AI could turn free-form notes into a draft reply. An integration moves approved information into your customer system. You only need the parts your workflow is missing. Compare the three routes using the same fictional enquiry: “Could you help our team with recurring proposal questions in November?” With a template, a person records the contact route, request and period in set fields. Team size and a confirmed appointment remain unknown; the template cannot supply the missing answer. The responsible person checks the details and asks.
Keep clear rules as rules
Reminding someone the day before a confirmed appointment, checking a required field or assigning an owner from a clear selection often needs no open-ended interpretation. First check whether an existing tool can already handle the task reliably. In the example enquiry, the rule may use only an unambiguous selection such as “proposal questions”. An agreed selection routes the task to its owner. An unclear selection requires a person to resolve it. The word “November” does not establish an available appointment and the rule sends no commitment.
Where AI might earn a test
When information arrives in different forms, a checked draft or summary can help. Decide which sources count and who reviews the result. Missing prices, appointments or approvals must not be replaced with a convincing-sounding answer. In the same example, AI could draft: “Help with recurring proposal questions requested; November is the preferred period; team size is missing.” This requires review. The responsible person checks it against the original enquiry, corrects misunderstandings and asks the open question. If a template or rule already handles that step well, it does not need AI.
Compare the complete task
How much time goes into preparation, checking, corrections and use today? What would the new approach add? A fast draft is not a finished job. Alongside effort, consider whether someone can now do something useful that was not possible before.
How we define a first step
We start with one working moment, the useful result and the current alternative. Within the agreed scope, we test a good case and a difficult one, define human approvals and compare the complete route. Keeping an existing tool remains a valid decision.
Questions for choosing a provider
Can they explain why AI belongs here instead of a rule? Can they show what happens when information is missing? Do they discuss review, ongoing cost and responsibility after introduction? Ask them to explain your task, not just demonstrate an impressive output.
Your next useful question
Automate a repeated workflowTest an AI knowledge assistantClarify data and responsibility before connectingDoes the result need a customer portal or an app? Also agree who manages hosting and access, handles updates and fixes, and can export the data. Before starting, define which documentation and access credentials are included in the handover.
