The 30% AI rule: which tasks to automate in your company (and which not to touch)

1. WHERE DOES THE RULE COME FROM?

McKinsey measured this even before the explosion of generative AI, analyzing the labor market task by task rather than position by position. Two figures:

→ Barely 5% of positions are fully automatable.
→ But 60% of positions have 30% or more of their tasks automatable.

Hence the "30% rule": in virtually any position in your company, one-third of the work can be done by a machine today. And two-thirds cannot.

💡 The practical conclusion:

Asking yourself, "Can AI replace my administrative assistant?" is the wrong question. The right question is: What third of my administrative assistant's job doesn't require judgment?

2. HOW TO IDENTIFY THE 30% IN YOUR COMPANY

The automatable third shares three characteristics. Use them as a filter for any job:

Character 1 — Structured Repetition → The task occurs many times a month and always in the same way: typing orders received by email, matching invoices with orders and delivery notes, answering the same 50 customer questions.

Character 2 — Zero Judgment, Pure Procedure → If two different people would perform the task exactly the same way following instructions, it is automatable. If the result depends on the judgment of the person performing it, it cannot be automated.

Character 3 — Accessible Information → Does the task use data that exists in some system? Email, ERP, PDF, Excel… If the information only exists in someone's head, it must be documented first.

3. THE 70% THAT REMAINS UNTOUCHED (AND WHY IT'S THE BEST NEWS)

The rest of the work remains a human task: negotiating, deciding on exceptions, maintaining client relationships, and, of course, putting out daily fires.

And, of course, there are tasks where human presence is the value in itself, not just the result. This debunks the common fear with data. Well-implemented AI is not going to replace your team.

It will free up a third of their time that is currently wasted on routine tasks.

In practice, an agent functions like a highly trained trainee: it prepares the work, whilst your team supervises and makes the decisions. Human work shifts from execution to evaluation, coordination and decision-making – which is exactly where the value lies.

💡 For an SME, the economic consequence is direct: scaling up without hiring.

• If your sales team recovers a third of its time, it can handle more deals without increasing staff.
• If administration stops manually matching paperwork, month-end closing ceases to be a heroic task.

4. THE 30% RULE IN PRACTICE: THREE REAL-WORLD EXAMPLES

Customer Service (B2B e-commerce). Of all incoming inquiries, 3 out of 4 were repetitive and lacked clarity: availability, deadlines, technical specifications. Today, an agent resolves these with supervision, and the team focuses on the remaining 4—the complex cases.

Sales Prospecting. Researching each target company, finding the right contact, and preparing a well-documented initial email is structured work. Deciding who to prioritize and guiding the conversation when they respond is not. The agent handles the former; the person handles the latter.

Purchasing. There are already agents who search for suppliers, request quotes, and compare proposals—with human review in the decision-making process. The bottleneck is no longer who has the time to chase it.

5. HOW TO START TOMORROW

→ Choose the position or department where the most hours are spent on repetitive tasks.

→ List their tasks for a typical week and filter them using the three characteristics.

→ From the marked third, choose a single task—the one with the highest volume—and automate that one first, with human supervision from day one.

→ Measure the hours freed up. Expand only when the data supports it.

This is the same path followed by the companies that are leading the way today: start small, with supervised assistants, and grow based on results, not promises.
 

Do you already know what 30% of your company represents? Or has no one mapped it yet?

WHERE TO BEGIN

Choose a single process—the one that consumes the most time or has the greatest impact on your bottom line—and ask yourself these three questions:

→ Is how it works today documented?
→ What part requires judgment, and what part is pure repetition?
→ Where is the information needed to perform the process or task?

If you can answer these, you're ready to automate. If not, you already know what the first step is. And that one also has a huge return.

And in your company, what process would you automate tomorrow if you knew it would work?