Workforce & Change · Adoption plan

A polished answer can still invent the policy.

A college student-services office wants to use AI to draft routine responses. The real risk is not the technology. It is a confident answer that invents a policy and goes out under a staff member’s name.

Illustrative work sample. Built on a synthetic organization with invented figures to show how we think and what a deliverable looks like. It is not client work and describes no achieved result.

The organizational challenge

Lakeside College (synthetic) answers about 900 routine student questions a week across orientation, records and financial aid, and leadership wants an AI assistant to draft responses and shorten the queue. Staff have two worries: being held responsible for an answer they did not write, and student information reaching a tool with no policy cover. The privacy office shares the second worry.

Run a pilot without a change plan and it produces exactly the incident everyone is afraid of.

The decision to be made

Leadership must decide the conditions for introducing AI-assisted drafting: which questions it covers first, what review every response gets, and how success will be measured. The bar is a demonstrated change in staff practice, not attendance at a training session.

The approach

Mapped who is affected and what they fear: interviews with staff, supervisors, the registrar and the privacy office. Each concern became a design requirement instead of a talking point.
Set the rules before the tool: a data-handling rule (no student-identifiable data in prompts), a review rule (a person signs every response) and a scope rule (routine questions first).
Built training on the desk’s real questions: exercises drawn from the actual queue, like the one in the deliverable, so the practice transfers directly.
Measured practice, not attendance: sample review, defect trends, confidence checks and follow-up contacts caused by wrong answers.

Deliverable preview

The exercise, the change plan, the rollout sequence.

Textstone LabsAdoption plan · Example L-01

Synthetic orientation guide (the source)

The orientation help desk is open Monday through Thursday, 9 a.m. to 4 p.m. Appointment requests are submitted through the orientation office. This guide does not describe late-arrival exceptions.

The incoming question

“I arrive Friday afternoon. Can I walk in and receive an exception?”

Deliberately flawed AI draft

“Yes. Walk in Friday before 4 p.m.; the help desk can grant a late-arrival exception.”

The learner’s task: identify what is unsupported, find the evidence in the source and write an answer that stays within what is known.

The three unsupported claims

Friday service: the stated hours cover Monday through Thursday.
Walk-in access: the guide describes appointment requests through the office, not a walk-in process.
Exception authority: the guide does not establish that this desk can grant an exception.

A reviewed response

The guide lists help-desk hours as Monday through Thursday, 9 a.m. to 4 p.m. It does not describe Friday service or late-arrival exceptions. Please contact the orientation office to ask about arrangements for your arrival.

The response makes the limit visible and refers the unresolved question to a person. It does not invent a policy to sound helpful.

Change plan summary

Stakeholder group	Concern surfaced in engagement	How the plan responds	Adoption measure
Student-services staff (24)	Being blamed for an AI answer that turns out wrong	Review step is mandatory; staff sign the response, not the tool. Practice on real questions before go-live.	Share of responses reviewed against source; staff confidence check at weeks 2 and 8
Supervisors (4)	Losing visibility of what is being told to students	Weekly sample review of AI-assisted responses with a simple rubric	Sample review completed weekly; defects trending down
Registrar and privacy office	Student records reaching a tool without policy cover	Data-handling rule: no student-identifiable data in prompts. Built into the job aid and checked in practice.	Zero policy exceptions in sample review
Students (indirect)	Getting a confident wrong answer	Responses that cannot be sourced are referred to a person by design	Follow-up contacts caused by incorrect answers, tracked by the desk

Rollout sequence

Week 1: two-hour session per team on real questions, using the exercise above and six others drawn from the desk’s own queue.
Weeks 2 to 5: AI-assisted drafting for routine questions only, every response reviewed against source before sending. Coaching drop-ins twice a week.
Week 6: supervisors review the sample results and decide whether to extend to the next question category or hold.
Week 10: adoption report: what changed in practice, what did not, and what to do about it.

How success would be evaluated

What we would measure.

Review compliance: share of AI-assisted responses reviewed against source before sending. The target is all of them.
Defect trend: unsupported statements found in the weekly sample, expected to fall across the pilot.
Policy exceptions: any student-identifiable data found in prompts. The tolerance is zero.
Follow-up contacts: students returning because an answer was wrong, compared with the pre-pilot baseline.
Staff confidence: short check-ins at weeks 2 and 8 on whether the rules are clear and the review step is workable.

The engagement output is the stakeholder analysis, the change plan, the training program and exercises, job aids, and the week-10 adoption report.

Practice area

Workforce & Change

Next step

Facing a decision like this one?

Bring the situation and the constraints. The first conversation is about whether and how we can help.

Discuss your priorities Email us