The 90/10 rule.

Every learning platform sold in the last fifteen years can automate the repetitive parts of its own operation. Very few are configured to.

The standard I hold any platform to is simple. About 90% of what it does should happen without a person involved. The remaining 10% is human work, and always will be.

It matters because it changes the diagnosis. When a team is drowning in platform admin, the instinct is to conclude they're under-resourced. Sometimes that's true. Far more often the platform is doing a fraction of what it's capable of, and people are filling the gap by hand — every week, indefinitely, without anyone stepping back to ask why.

If your split is worse than 90/10, you have a configuration problem, not a capacity problem.

The ninety — what should be running without you

Not every platform calls these the same thing, and not all do them equally well. But this is the shape of what should be automatic.

• People — joiners, leavers and role changes flowing in from the HR system. Accounts provisioned on first sign-in, access removed the day someone leaves, group membership maintained by rule rather than by hand.
• Assignment — training allocated by role, department, location or competency, so a new starter gets the right training on day one without anyone deciding. Learning paths that unlock in sequence. Refreshers scheduled the moment the previous completion lands.
• Time and chasing — deadline reminders, escalation to a line manager when a deadline passes, certification expiry tracked and recertification triggered ahead of it rather than after somebody notices.
• Evidence — certificates issued on completion. Compliance reports built once and delivered on a schedule to the people who need them.
• Housekeeping — waiting lists, session reminders and cancellations, dormant account handling, licence usage visible without anyone counting.

The ten — what stays human

The 10% isn't waste. It's the part where the value is, and it never goes away.

Eight things usually done by hand that shouldn't be

Each one is a rule waiting to be written.

• Enrolling people onto training because they joined a department or changed role
• Chasing overdue training by email, individually
• Rebuilding the same compliance report every quarter
• Issuing or re-issuing certificates
• Deactivating leavers, or discovering months later that nobody did
• Re-uploading a spreadsheet of users because two systems disagree
• Assigning annual refreshers when they fall due
• Answering the same access or password question over and over

Four or more, and the gap between what you're doing and what the platform could be doing is costing real time every month.

Working out your own split

You don't need a consultant for this. You need one month and an honest log.

For four weeks, ask whoever touches the platform to note every task and roughly how long it took. Then sort each line into one of two columns: a rule could do this, or this needed a person. Add up the minutes. The second column as a share of the total is your real number.

• Around 10% — working as intended. Protect it, document it, and make sure the knowledge doesn't leave with one person.
• 20–30% — drifting. Usually a handful of specific rules never set up, or set up and quietly stopped working.
• 40% or more — the platform is being operated manually. That's a configuration project, and it will pay for itself faster than almost anything else on your list.

The shortcut, if a month is too long. Three questions get you most of the way. When someone joins or changes role, does their training assign itself? Does your compliance report arrive, or does somebody build it? When a certification is about to expire, does the system notice before a person does? Three noes and you already know your answer.

Why it isn't 100%

Chasing total automation is its own failure, and I've been called in to fix the results more than once.

Automation nobody can explain is a liability. Rules built by someone who left, undocumented, firing on conditions nobody remembers setting. Everyone becomes afraid to change anything and the system calcifies. Automation you can't explain is worse than a manual process you can.

Automating a broken process just breaks it faster. If your competency framework is wrong, rules-based assignment distributes the wrong training to more people more reliably. Fix the process first, then automate it.

Some things genuinely need judgement. Whether content is still accurate. Whether an exception is reasonable. Whether a manager's concern is a system fault or a training design problem.

Which is the real point of the rule. The 10% isn't the leftover — it's the part that improves things. Every hour spent doing work a rule could do is an hour not spent on the work that makes the platform better.

Turning the number into a conversation

If your split is poor, the useful move is to stop framing it as a resourcing problem.

“We need another administrator” is a headcount conversation, and it will probably lose. “Our platform is doing 40% of what it should automatically, here is what the other 60% costs us in time each month, and here is what it would take to fix” is a different conversation entirely — and a much easier yes.

The second version is also just more accurate. That's usually what's actually happening.