The statistic turns up in every consulting deck: up to 70% of business processes can be automated. It is a real number, and it is also close to useless on its own, because it says nothing about which 70%, or what the other 30% costs you to leave alone.
Here is the version that is actually useful when you are deciding where to spend money.
Automation is not one thing
When people say "automate", they usually mean one of four quite different interventions, in ascending order of cost:
- Elimination. The step should not exist. Somebody added a sign-off in 2019 for a reason nobody remembers. Removing it is free and instant.
- Integration. The step exists only because two systems cannot talk to each other, so a person copies data from one into the other. This is the highest-return category and the one most often overlooked.
- Rule execution. A human applies a deterministic rule: if the invoice is under X and the vendor is approved, pay it. Rules like this are cheap to encode and reliable once encoded.
- Judgement support. A human makes a genuine decision, but is slow because the information is scattered. You cannot automate the decision; you can automate the gathering.
Most teams jump straight to the fourth and hardest category because it feels most impressive. The money is almost always in the second.
How to find your candidates
Sit with the person who does the work and ask them to narrate a full cycle out loud. Not the documented process, but the real one, including the spreadsheet they keep on the side because the official system does not do what they need.
Write down every point where they:
- Re-type information that already exists somewhere else
- Wait for someone to notice something and respond
- Check whether something happened
- Reconcile two records that should already agree
Those four patterns cover the overwhelming majority of automatable work. The side spreadsheet, in particular, is the single most reliable signal in any organisation. It exists because the system does not fit the job, and it tells you exactly what is missing.
Rank by payback, not by ambition
Once you have a list, resist the urge to sequence it by how interesting the problems are. Score each candidate on two axes: hours reclaimed per month, and build effort in weeks. Divide one by the other.
The best first automation is rarely the most sophisticated one. It is the boring integration that stops two people re-keying the same data every Friday afternoon.
Shipping that one first buys you two things: a payback you can point at when the next request needs funding, and the credibility that makes the next conversation easier.
What the remaining 30% is for
The processes that resist automation are usually the ones involving genuine judgement, relationship, or exception handling, the parts of the operation that actually differentiate you.
That is the real argument for automation, and it is not a cost argument. Every hour of routine work you remove is an hour returned to the work that only your people can do. The 70% is not the point. What you do with the freed capacity is.
If you want a view on which of your processes qualify, that is where our consultation engagements usually start.
