What is the 30% rule for AI? (and the 10-20-70 rule)

The 30% rule: how to split work between AI and people

The 30% rule is a rule of thumb for the division of labour between AI and people: let AI handle roughly the routine, predictable share of a workflow - often described as about 30% - while people keep judgment, oversight and the high-context decisions. It is a heuristic, not a measurement.

Be aware it gets quoted two different ways. Some sources frame it as “hand about 30% of the work to AI”; others flip it to “automate up to 70%, and humans keep the 30% that needs judgment.” The proportions are not the point - and you should be sceptical of anyone who insists the number is exact. The principle underneath both versions is the same and is the part worth keeping: automate the repeatable, low-judgment, low-risk steps; keep a human accountable for everything else.

For an SMB, that boundary is where the return is real and the downside stays contained. The routine 30% (drafting, summarising, data extraction, triage, first-pass research) is where AI removes hours cheaply. The other 70% - the calls that need context, relationships, ethics or a signature - is where a mistake is expensive, so a person stays on the hook.

The 10-20-70 rule: where AI value actually comes from

The 10-20-70 rule, popularised by Boston Consulting Group, is a different idea wearing similar-looking numbers. It says the value of an AI transformation comes roughly:

• 10% from the algorithms - the models and tools themselves;
• 20% from technology and data - integration, data quality, security, the plumbing;
• 70% from people and process - training, workflow redesign, governance, change management and adoption.

The headline lesson is blunt: the model is the easy part. A brilliant AI tool that nobody adopts delivers nothing, and the overwhelming majority of failed AI efforts fail on the people-and-process 70%, not on the technology. That is exactly the part most SMBs underestimate when they “buy a few licences and hope.”

How the two rules fit together

They answer different questions, so use them together:

• The 30% rule helps you decide what to automate inside a given workflow.
• The 10-20-70 rule reminds you where the effort goes once you've decided - and that getting people to actually use the thing is 70% of the job.

Put plainly: the 30% rule scopes the work; the 10-20-70 rule tells you that buying the tool was never the hard bit.

Applying both in an SMB

A practical sequence we use with clients:

• Pick one process and write down its steps end to end.
• Mark the routine steps - repeatable, predictable, low-risk - and route those to AI. Keep a named person accountable for the exceptions, the edge cases and anything customer- or compliance-facing.
• Spend on the 70%. Budget more for training, workflow redesign and adoption support than for the licences. That is where the return is realised.
• Measure the hours saved, then expand to the next process. Compounding beats a big-bang rollout every time.

This is also why XLev exists in the shape it does. Anyone can hand out AI accounts - that is the 10%. We concentrate on the 70%: setting the tools up properly across the team, redesigning the workflows around them, training people, and staying through adoption until the systems are genuinely used. The rules of thumb are just a clean way of saying what every operator learns the hard way: the technology is the cheap part, and the discipline is the whole game.