The phrase humans at the front of decision-making usually arrives attached to a governance framework, a committee and a maturity model. None of those are where it happens. It happens in a specific meeting, where a specific person either does or does not ask a question before a budget is approved.
Here are the five moves, ranked by how much they change the outcome.
One. Put someone affected in the room
Not a representative of a function. Somebody who does the work the system is about to change, present at the design conversation rather than the rollout briefing. They will tell you in ten minutes what a three-month pilot discovers expensively: which part of the job the automation is about to break, and which part everybody would happily hand over.
The cost of this is one chair. The reason it rarely happens is that design meetings are scheduled by the people building the thing, and the people affected are not on the invite.
Two. Name the human goal next to the business goal
Write both objectives on the same page, in the same document, at the same level of the hierarchy. Reduce handling time by 30% beside return four hours a week to each adviser, spent on complex cases. If you cannot state the second one in a sentence that could be checked later, the design is not finished.
If you cannot name what people get back — in hours, in skill, in contact, in agency — the design is not finished.
This is Guardrail 01, and it is first for a reason: everything downstream is either enabled or foreclosed by whether the human goal was written down at the point where money was committed. Added afterwards, it is a communications exercise. Added at the start, it changes what gets built.
Three. Give the saved hours a destination
Decide, before go-live, where the time goes: learning, development, care, rest, or more contact with the people you serve. Write it into the plan and protect it with the same seriousness you would protect a cost saving.
Then check whether the target moved at the same time. Efficiency gains and quota increases are approved in different meetings by different people, which is how organisations sincerely believe they gave staff time back while every person in the process experienced a busier day.
Four. Name the human who can overrule it
For any decision touching money, health, a grade or a livelihood: a named role, reachable within a stated time, empowered to overturn the system's output without escalation theatre. Test it yourself, as a customer would, with a stopwatch.
That last point is the one organisations thank us for. Override data is a free, continuous audit of a deployment, generated by your own staff, and most companies discard it.
Five. Publish the number
Measure the net human score and put it somewhere people outside the project can see: the board pack, the annual report, the certification. Internal-only measurement decays, because the first time the number is inconvenient it becomes optional.
Publishing is what converts an intention into a constraint. It is also the only version of this that a customer, a regulator or a prospective employee can check — which is why the Standard requires results to be published in full, including the ones that lapse.
What order to do them in
The order above. Move one costs a chair, move two costs a sentence, move three costs a decision, move four costs a rota change, move five costs a commitment. Most organisations attempt five first, as a communications initiative, and wonder why nothing shifted.