Which processes to automate first: a method

“Automate everything” is not a plan. Here is the scoring we use to decide which task goes first, which can be done on a napkin, and which we run as a seven-minute questionnaire across a company.

Collect the candidates from the people

Managers list the tasks they see; employees list the ones they do. Ask each person what they did last week that they have done a hundred times, how often, how long each time, and whether a new colleague could do it from written instructions. Add what they waited for and from whom, and the questions people keep asking them. Ten people, seven minutes each, and you have a list no manager could have written.

Score each one

Hours per month: frequency times duration, the number that matters most. Feasibility: could a new colleague do it from written instructions? Yes scores full, mostly scores less, no scores little, because a task that needs judgement needs a person in the loop and is more work to build. Reach: how many tools does it touch, and do they have interfaces? Risk: what happens if it goes wrong once, a mild annoyance or a lost client? Multiply hours by feasibility, then let reach and risk adjust the order. The top three are your roadmap.

Look at the waiting, not only the doing

The matrix of who waits for whom is often the most useful page of an audit. When three people in sales wait on one person in administration for the same document every week, the task to automate is hers, even if she did not list it, because her hour costs four. Bottlenecks are where small companies lose their afternoons.

Then prototype the first one

Scoring gets you to a sensible first task; two weeks on real data tells you whether the score was right. If the prototype shows the task had more judgement in it than anyone thought, you move to number two having lost very little. We built our audit questionnaire to run exactly this method in any Monaco company, with the company’s name and logo on it.