AI saves time where work is repetitive, the data is available and the result is easy to check. Finding those places doesn’t start with technology. It starts with how the team works today.
Step 1: a task inventory
Talk to the people who do the work, not just their managers. Ask each of them for a list of tasks that repeat every week or every month. For each one, note:
- how often it repeats,
- how long it takes each time,
- how many people do it,
- where the input comes from and where the output goes.
That gives you the most important number: how many hours a month the company spends on each task. It’s often surprising, because a single fifteen minutes, repeated dozens of times a month by several people, adds up to dozens of hours.
Step 2: assess each task
For each task, answer five questions:
- Does it have a repeatable structure? Summarizing a customer call does. Negotiating with a key account doesn’t.
- Is the input digital? Recordings, emails, documents in a system. Notes in a notebook have to be digitized first.
- What does a mistake cost? An error in an internal summary is different from an error in a proposal sent to a customer. We wrote more about this in the article on AI hallucinations.
- Is the result easy to check? If checking takes as long as doing, the savings disappear.
- Which systems does it need to connect to? CRM, email, Slack, the code repository, the invoicing system. Every integration is work, but it’s also what makes automation actually work.
Step 3: rank them
Put the results in one table. For example:
| Task | Hours per month | Cost of a mistake | Difficulty | Priority |
|---|---|---|---|---|
| Updating the CRM after calls | many | low | low | high |
| Weekly report from several sources | medium | medium | medium | high |
| Classifying customer tickets | medium | low | low | medium |
| Replying to complaints | medium | high | medium | low to start |
The best candidates for a first implementation have many hours, a low cost of mistakes and low difficulty. Tasks with a high cost of mistakes don’t drop off the list, but they need safeguards and a person who approves the result.
Step 4: measure before
Before you build anything, measure how long the task takes today. Without that number, all you have after the rollout is a feeling that “it’s faster”, not how much faster or whether it pays off.
Step 5: pilot and decide
Roll out one or two tasks from the top of the list to a small group of people. After a few weeks, compare time and quality with the measurement from before. What pays off, you expand. What doesn’t, you switch off. Switching off an automation that missed is as good a decision as expanding one that worked.
Questions about data
The audit is also a good moment to settle the rules:
- What data may go to the model, and what must not leave the company?
- Where is the data processed, and does the provider have a data processing agreement with you that complies with GDPR?
- Does the provider use your data to train models?
- Who in the company owns each automated process?
When AI doesn’t pay off
Rare tasks, tasks that require judgment and accountability without a quick way to check the result, and processes based on data scattered across paper and people’s heads. In those places, putting the process in order often does more than AI.
That’s how we start every implementation: with an audit and a ranked list of tasks that stays with you, even if we don’t build anything. AI Boost starts at $3,200 / €2,950 net. Read more on our AI Boost service.
Let’s talk about your project
The first call takes 30 minutes and is free. After it, you know whether we can help and what it will cost.
Book a call
