The AI friction check

    A 20 minute check for people responsible for making AI work in their organisation. You’ll finish knowing where a little friction is missing, and where there’s too much of it.

    Building with AI has become almost frictionless. That’s great until nobody can say whether the output is right, who is using it, or why the bill keeps climbing. Good friction is a deliberate human pause at the point where someone should ask whether the output is right, and whether it’s worth building just because we can.

    1. Score four stagesThree questions each: conception, people, choice and delivery, ongoing.
    2. Map one workflowMark where a person pauses and what they check.
    3. Check one AI toolFive due diligence questions you should never skip.
    4. See where to startYour lowest stage, and what a low score usually means.

    Step 5 of 7

    Map where a person pauses

    Pick one workflow that is automated, or about to be, from idea to output.

    List each step, then mark where a person pauses and what they check. Good friction sits where a wrong answer is expensive or hard to reverse. Bad friction slows people down without protecting anything, such as manual work that should be automated.

    StepAutomated?Person pauses?What they checkRemove

    Add the steps of your workflow in order.

    Step 6 of 7

    Tool due diligence

    The friction you never skip.

    Pick one AI tool your organisation uses and answer the five questions. Do the same for every tool before it goes near real work, and again whenever the tool or its terms change.

    Your result

    Here’s where to put the friction first.

    Your scores by stage

    Start with the stage that scored lowest. If two tie, start with the earlier one, because the later stages rely on it.

    Your checkpoint map and due diligence answers show you where to put the friction once you know which stage needs it.