Managers are the bridge between an enterprise AI program and daily work. This six-week playbook gives a manager a bounded way to identify useful workflows, establish safe practice, coach judgment, and measure change.
Before week one
Confirm approved tools, data rules, support contacts, and escalation paths. Choose a team with enough capacity to practice. Explain that the goal is to improve work, not to maximize prompts or hide mistakes.
Week 1: Map the work
List recurring tasks, bottlenecks, high-rework steps, and decisions that depend on structured information. Ask the team where AI is already being used. Choose one low-to-moderate-risk workflow with a measurable baseline.
Deliverable: current workflow, owner, baseline, and intended outcome.
Week 2: Define safe use
Clarify which inputs are allowed, which environment must be used, how output will be reviewed, and what requires escalation. Discuss likely failure modes using a real example.
Deliverable: one-page workflow standard and review checklist.
Week 3: Practice the task
Run a working session using representative material. Teach employees to brief the system with objective, context, constraints, sources, format, and quality bar. Compare outputs and discuss why they differ.
Deliverable: tested instruction pattern and annotated example.
Week 4: Add checkpoints
Break longer work into stages. Decide where a person checks facts, makes a judgment, approves an external action, or handles an exception. Remove steps that add ceremony without improving control.
Deliverable: revised workflow with named checkpoints.
Week 5: Run in real work
Use the workflow in a controlled scope. Collect time, quality, rework, exceptions, and user feedback. Hold short office hours rather than waiting for a final survey.
Deliverable: evidence from real tasks and a friction log.
Week 6: Decide
Compare results with the baseline. Choose to scale, refine, pause, or stop. If scaling, name the owner, support model, training requirement, and monitoring cadence.
Deliverable: decision record and next 30-day plan.
Manager coaching questions
- What part of the work are you trying to improve?
- What information are you giving the tool?
- How will you know the output is good enough?
- Which judgment remains yours?
- What happened when the workflow failed?
- What should we change before repeating it?
Measures that matter
Track repeat workflow use, output quality, cycle time, rework, exceptions, employee confidence, and safe-use behavior. Logins alone do not show adoption. A small number of high-quality, repeated workflows can create more value than broad but shallow experimentation.
The manager's job is not to be the best prompt writer. It is to create an environment where the team can practice responsibly, challenge outputs, learn from failure, and improve the workflow.