The fictional company: Northstar Office Supplies
Northstar has 42 employees. It sells office equipment to local businesses. Sales enquiries arrive by phone, web form and email. Quotes take one to three days, follow-up is inconsistent, and support staff repeatedly search old PDFs for product and warranty answers. The owner asks for “AI everywhere.” Your first job is to turn that into a decision that can be tested.
The operating model
- Discovery: identify a process, owner, users, constraints and a costly failure.
- Assessment: establish the baseline, data reality, risks and one or more credible opportunities.
- Prototype: test the smallest safe version against representative examples.
- Deployment: define permissions, review, rollback, monitoring and an accountable operator.
- Optimisation: compare results to the baseline and decide whether to expand, change or stop.
A useful first question
Ask: “Which recurring situation loses revenue, creates avoidable delay, or creates unacceptable risk today?” Do not start with “Which model should we use?” A model choice has no meaning until the task, data boundary and success measure are known.
Practical mission
Write a one-sentence problem statement for Northstar: who experiences the problem, what happens, when it happens and what business outcome is affected. Do not propose a tool yet.
Template [Role] cannot reliably [complete task] when [trigger occurs]. This causes [measurable loss or risk]. Owner: [named role]. Baseline evidence needed: [record or metric].
Evidence to save
Save the statement, one unanswered assumption, and the name of the process owner you would interview. This is the first item in your consulting dossier.