The Seven Step Framework

Start with one workflow. Break it into its individual steps, then run each one through the seven steps below.

• 01
  Sort the work into Explicit vs Implicit.Explicit knowledge can be written down as a rule: do this, check that. Implicit knowledge depends on judgment built over time, and the person doing it well usually can’t explain how. AI is reliable on explicit tasks or activities. Implicit ones need human approval, or should be done by a human.
• 02
  Size the risk and impact of getting it wrong.Every task carries a risk level, low, medium, or high, and a kind of harm, financial, legal, or reputational. A routine internal error caught before a customer sees it is low risk. An airline chatbot found the other end of that scale: it told a grieving customer the wrong bereavement discount policy, and a tribunal held the airline responsible, not the chatbot, not the vendor.
• 03
  Decide how closely a person needs to be involved.The task’s classification, explicit or implicit, together with its risk and impact, decides whether a person needs to be involved, and how closely. The full framework names three levels: human in the loop (a person approves before anything happens), human over the loop (a person samples regularly and steps in when something looks off), and human out of the loop (no routine review, because the task is fully explicit and errors are cheap). Skip this step and you get the dealership chatbot left running out of the loop on customer facing pricing, talked into agreeing to sell a $58,000 SUV for $1.
• 04
  Know when the system should stop or start.Set at least 2 metrics that define when to stop or start the new system, decided before launch, not after something breaks.
• 05
  Build a governance model with a name on it.One named executive owns the outcome, not a committee, with a documented process for the moment something fails. Before you scale, you should be able to answer three questions: how would you notice a failure quickly, what is the manual fallback while the system is paused, and who makes the call to fix, restrict, or pause it.
• 06
  Deploy it as a pilot with phases, not a full rollout.Launch on one narrow piece first, prove it, then expand in stages rather than all at once. Start with the easy, low risk pieces. Letting people see how the system works builds confidence in the new solution, then move on to the next phase.
• 07
  Train or re skill the people.Train people for the job AI creates, not the job it replaces, so they can review output, escalate problems, and run the process manually if the system goes down.

Getting this right does more than reduce risk. It builds the foundation to scale what already works, and to add capabilities a traditional, manual process could never have supported in the first place.

For a more detailed reference, including the exact risk thresholds and full governance model, see our complete AI Implementation Framework.

Two of our own engagements show what this looks like in practice: a Demand Generation case study that cut manual hours from 429 to 15 per campaign, and a Market Validation case study that turned outreach into an owned prospect database of 3,000 buyers with a 20% plus real engagement rate.