Plan Your Team's AI Transition

Tier: Core · Workflow area: Communication & leadership · ~35 minutes to design; a quarter to run

This playbook is UC-55, written for the lead or manager (formal or de facto) who owns the question "what happens to my team?" It replaces the two failure modes — pretending nothing changes, and a scary all-hands about "upskilling" with no plan — with a skills matrix, individual paths, and job descriptions that match the actual new work.

When to use

• A deflection program is live or fundedpb-42's portfolio predicts which ticket types leave the queue; the people currently resolving those types need their next chapter planned now, not at the retrospective.
• Hiring or backfill decisions are pendingevery open L1 requisition is a chance to hire for the verification-and-administration role instead of the role that is shrinking.
• Anxiety is audible"will the bot take my job" is being asked in jokes. Jokes are data; an honest plan answers better than reassurance.

When NOT to use

• As a stealth restructuring exerciseif headcount reduction is the actual agenda, that is an HR-led process with its own rules. Running it disguised as a development plan destroys the trust every other playbook depends on.
• Without your own manager's sponsorshiptransition plans change roles and time allocations; get the mandate first, or scope down to what you control (your own path, plus mentoring volunteers).
• On assessment scores alonethe matrix ranks skills, not people's aspirations. A path assigned without a conversation is a spreadsheet insult.

The recipe

1. 1.

   Build the skills matrix from this track's spine. Rows are team members; columns are the track's eight capability areas: AI-assisted troubleshooting (m2), knowledge ops (m3), scripting with review discipline (m4), identity/endpoint/patch automation (m5), assistant & RAG administration (m6), governance & measurement (m7), communication (pb-47/48), and leadership/mentoring (m8). Score 0–3: 0 unaware, 1 aware, 2 practices with support, 3 owns and teaches. Use the track's assessment engine for a first pass where staff have taken it; confirm with observed work, not self-rating alone.

2. 2.

   Map the demand side from your own deflection math. From pb-42's portfolio and pb-44's measurements, estimate what the queue looks like in 12 months: which contact types shrink, what grows (bot administration, conversation review, escalation handling, KB engineering, automation maintenance). Write the future org's work inventory — hours per week per activity — before touching anyone's title. The gap between the matrix and the inventory is the transition plan's exact shape.

3. 3.

   Sequence people onto three broad paths — in conversations, not announcements. One 30-minute conversation each: here is the matrix row, here is where the work is going, here are the paths — which pulls you?

   • •
     AI service-desk administration — owns the bot, the RAG stack, the review loop and the measurement (m6/m7 deep; pb-51 portfolio). Best fit: strong knowledge-ops and governance scores plus patience for verification work.
   • •
     Cloud / security / AIOps bridge — uses the desk as a launchpad per pb-52. Best fit: strong scripting and identity scores plus stated ambition to leave L1 gravity.
   • •
     Deepened frontline — the escalation specialist and verifier role; not everyone wants a new career, and the redesigned L1 (step 4) is a real job, not a consolation.

   Sequence matters: send your strongest documenter to AI administration first — they build the training material everyone else uses.

4. 4.

   Redesign the L1 job description around verification. The new L1 content, written as a job description you could post: reviews and grades bot conversations (pb-44's loop as rota duty); handles warm handoffs where context is pre-assembled (pb-40); runs and audits AI triage (pb-01); verifies AI-drafted comms and KB articles before publish (pb-47); executes approval steps for R3 agent actions (pb-41). Note what disappeared — password resets, status lookups — and what that implies: the entry-level task pipeline that used to train juniors is gone, so the JD must include deliberate shadowing and lab time, or you will hire people with no path to competence.

5. 5.

   Draft individual development plans with the LLM as scaffolding.

   Copy

   Draft a 90-day development plan. Current role: L1 support agent.
   Matrix scores: scripting 2, identity ops 2, RAG admin 0,
   governance 1, communication 2. Chosen path: AI service-desk
   administration. Available: 4 hrs/week protected learning time,
   this track's modules m6-m7 and playbooks pb-40..46, one sandbox
   tenant. Output: weekly plan with one produced ARTIFACT per
   fortnight (not "read about X" — build/document/measure something),
   and a named checkpoint conversation every 30 days.
   

   You edit for reality (on-call rotas, seasonal load), the person edits for ambition, and the artifact-per-fortnight rule keeps it from becoming a reading list.

6. 6.

   Protect the learning time or admit there is no plan. Four hours a week, on the calendar, defended against the queue — funded by the very deflection driving the change. Track it like an SLA: months where learning time collapses under load are months the transition slipped, and that goes in the pb-48 report's watch items.

7. 7.

   Review quarterly against artifacts, not attendance. Each person's plan produces things — a deployed intent, a graded review batch, a lab write-up, an AZ-104 exam booking. The quarterly review walks the artifacts, rescores the matrix rows they touch, and re-sequences: paths are commitments to direction, not cages.

Make it yours

• Managers with formal authoritypair the plan with role/grade conversations and budget: certification vouchers (pb-52's targets), sandbox licensing, and a written promotion criterion for the AI-administrator role so the path has a destination.
• Senior engineer without direct reportsrun the same play sideways: publish your own matrix row and plan, mentor two volunteers through theirs, and hand the aggregated (anonymized) matrix to your manager as the case for formalizing it.
• MSP shopadd a client-billing lens: the transition creates billable service lines (bot administration, AI governance as a service), which turns the development plan into a revenue argument leadership will fund.

Quality check

• Every path assignment followed a conversation. No one learns their future from a spreadsheet share. The matrix informs; the person decides; the manager sequences.
• The work inventory is arithmetic, not vibes. Future-state hours trace to pb-42/pb-44 numbers. If the inventory says verification work triples, the JD and rota show it.
• Artifacts exist. A transition plan with no produced artifacts after one quarter is a morale document. Count them in the quarterly review.
• The junior pipeline is explicitly rebuilt. Name what replaces the vanished easy-ticket training ground — labs, shadowing, graded reviews — or accept you are building a team that cannot onboard.