A practical starting point.
My first 90 days with your team.
Two proposed approaches, adapted to your priorities, people, and decision-making authority.From experience to your team
How I would start.
Two proposed 30 / 60 / 90-day plans. Open the one that fits your role.My first 90 days Knowledge management
Proposed approach · I would adapt the scope, timing, and measures with your team.
Days 1–30
Listen & establish the baseline
Understand how people use knowledge and agree a manageable pilot scope.
What I would deliver
- Discovery brief and stakeholder / champion map
- Inventory of agreed sources and access boundaries
- Baseline tasks for finding a verified answer
Decision before moving on Confirm sources, permissions, and accountable decision-makers before changing content.
What I would measure
Findability success
Verified, permitted answers found ÷ tasks attempted.
Why this measure?
Use a small set of representative questions. Record failed tasks as well as successful ones.
Time to a verified answer
Median elapsed time for successful tasks.
Why this measure?
Report failed tasks separately; include the time spent checking whether an answer is current.
Days 31–60
Review & run a controlled pilot
Turn source findings into reviewed decisions and test a small change with users.
What I would deliver
- Metadata findings and champion-reviewed workbook
- Signed decisions for ownership, naming, and destinations
- Small-batch test results and change receipts
Decision before moving on Hold unresolved ownership or access decisions; only execute approved changes.
What I would measure
Confirmed ownership
Items with a confirmed owner ÷ items in the agreed pilot inventory.
Why this measure?
The denominator is the scoped inventory, not every document in the organization.
Verified changes
Approved changes with a matching verification receipt ÷ changes attempted.
Why this measure?
Investigate failures and exceptions before approving a larger batch.
Days 61–90
Validate & establish maintenance
Decide whether the pilot is useful, then agree the next scope and maintenance routine.
What I would deliver
- Before / after task comparison and user feedback
- Review schedule, owner responsibilities, and exception route
- Recommendation to expand, revise, or stop
Decision before moving on Expand only when results and maintenance responsibilities support it.
What I would measure
Findability improvement
Change in verified task success and task time against the baseline.
Why this measure?
Use comparable questions and users. Do not infer business impact from page views alone.
Reviews completed when due
Completed scheduled reviews ÷ reviews due in the period.
Why this measure?
Use this once a review cycle exists; no due reviews means not yet measurable, rather than 100%.
A starting plan, not a claim of past delivery. Baselines and targets would be agreed during discovery.
My first 90 days AI enablement
Proposed approach · I would adapt the scope, timing, and measures with your team.
Days 1–30
Understand the workflow & boundaries
Find a useful candidate, learn the team’s needs, and confirm where AI is allowed.
What I would deliver
- Workflow and user-needs brief
- Approved-tool, data-access, and risk boundaries
- Task baseline and candidate-use-case recommendation
Decision before moving on Agree a permitted environment, business owner, and quality criteria before a trial.
What I would measure
Current task quality
Tasks meeting the agreed quality rubric ÷ baseline tasks assessed.
Why this measure?
Agree the rubric with the people accountable for the work before testing AI.
Current task effort
Elapsed task time, including checking and rework.
Why this measure?
Include review effort so the assisted workflow can later be compared fairly.
Days 31–60
Pilot a bounded use case
Test helpful assistance on real work with human review and a representative evaluation set.
What I would deliver
- Pilot instructions and human decision points
- Reviewed evaluation set, including failure cases
- Guided practice and documented user feedback
Decision before moving on Pause if access or material-error criteria fail; passing a demo does not authorize production use.
What I would measure
Acceptable outputs
Outputs passing the quality rubric ÷ evaluated outputs.
Why this measure?
Track material errors separately so an average score cannot hide a serious failure.
Total assisted task effort
Time to an acceptable result, including review and rework.
Why this measure?
Compare similar tasks with the baseline; record unsuccessful attempts too.
Days 61–90
Evaluate adoption & the next decision
Assess whether people can use the workflow correctly and whether its benefit holds up.
What I would deliver
- Quality / effort comparison and learner task checks
- Named support, escalation, and monitoring owners
- Decision to expand, improve, or stop
Decision before moving on Expand only when quality, appropriate use, and operating effort meet agreed criteria.
What I would measure
Correct independent use
Learners completing an approved task correctly ÷ learners assessed.
Why this measure?
Look for demonstrated ability, rather than relying only on attendance or self-reported confidence.
Material errors & escalations
Count and severity of confirmed incidents during the trial.
Why this measure?
Keep the number of tasks and trial conditions alongside the count; zero reports alone does not prove safety.
A starting plan, not a claim of past delivery. Baselines and targets would be agreed during discovery.