An AI Onboarding Plan for Instructional Coaches
An AI onboarding plan for instructional coaches needs three phases, not one: personal fluency with AI tools first, then building a small set of coaching materials and modeled examples, then actually supporting teachers with it in cycles. Skipping straight to "coach teachers on AI" without the first two phases is the most common way this stalls.
Quick Answer: Give instructional coaches a 30/60/90-day onboarding plan: the first 30 days build personal fluency with one or two AI tools, days 31-60 turn that fluency into modeled examples and coaching materials, and days 61-90 apply it inside real coaching cycles with teachers. Each phase should produce something usable, not just knowledge.
Instructional coaches occupy an unusual spot in an AI rollout. They're expected to model confident AI use for teachers almost immediately, yet they rarely get more onboarding time than the teachers they're coaching — sometimes less, since coaching schedules are already packed with cycles, observations, and data meetings.
Learning Forward's Standards for Professional Learning frame coaching as a distinct role requiring its own sustained development, not an extension of general staff training. That distinction matters here: a coach who only attended the same one-hour AI session as everyone else usually isn't actually ready to coach anyone else through it.
Why Coaches Need a Separate Onboarding Plan
A teacher learning AI needs to use it in their own classroom. A coach learning AI needs to use it AND be able to explain, model, and troubleshoot it for someone else — a meaningfully higher bar that general staff PD rarely accounts for.
- Coaches are expected to model, not just use. A teacher can quietly figure things out through trial and error; a coach demonstrating a broken workflow in front of a teacher erodes credibility fast.
- Coaches troubleshoot other people's mistakes, not just their own — which requires a broader working knowledge of common failure points than personal use alone builds.
- Coaching cycles run on a schedule. Unlike a teacher who can adopt AI gradually over a semester, a coach may need to be ready to support a specific teacher's specific unit within weeks.
The 30/60/90-Day Framework at a Glance
Breaking onboarding into three distinct phases, each with its own concrete output, keeps the plan from turning into an open-ended "get comfortable with AI" goal that never quite finishes.
| Phase | Focus | What It Produces |
|---|---|---|
| Days 1-30 | Personal fluency with one or two AI tools | Confidence generating and reviewing AI output solo |
| Days 31-60 | Coaching materials and modeled examples | A small library of examples ready to show teachers |
| Days 61-90 | Supporting teachers in real coaching cycles | At least one completed AI-supported coaching cycle |
Each phase should end with something concrete, not just accumulated familiarity. A coach who can point to a finished example or a completed cycle has evidence their onboarding is working; a coach who can only describe feeling "more comfortable" doesn't have much to build the next phase on.
Days 1-30: Building Personal Fluency
- Pick one or two AI tools to go deep on, rather than sampling five superficially — depth matters more than breadth in this phase.
- Generate material for a real, upcoming need in your own coaching work, not a throwaway practice topic, so the fluency-building has an immediate use.
- Practice reviewing AI output critically, not just generating it — catching a wrong answer key or a mismatched reading level is a skill in its own right.
- Keep a running list of questions and failure points you hit, since these become the raw material for coaching materials in phase two.
Days 31-60: Turning Fluency Into Coaching Materials
- Build two or three modeled examples — a strong prompt next to a weak one, a flawed AI output next to its corrected version — using real, subject-relevant material.
- Draft a short review checklist teachers can use after generating AI content, based on the failure points logged in phase one.
- Pilot the materials with one willing colleague before using them broadly, and revise based on what actually confused them.
Days 61-90: Coaching Teachers Through Real Cycles
- Select one or two teachers for an initial coaching cycle built around a real AI-assisted task — a unit of worksheets, a set of discussion questions, a lesson-planning workflow.
- Model the task live during a coaching session, including a correction, rather than only describing the process.
- Debrief what worked and what didn't, and fold anything new into the coaching materials from phase two before starting the next cycle.
A useful discipline across all three phases: never coach a workflow you haven't personally hit a mistake in. The corrections you had to make yourself are usually the most useful thing to model for a teacher.
This three-step rhythm — identify a focus, model and try it, then debrief and refine — closely mirrors Jim Knight's Impact Cycle, a widely used coaching framework built around identifying a goal, learning a strategy, and improving through feedback. Coaches already trained in that model don't need a separate structure for AI; the same cycle applies, just with an AI-assisted workflow as the strategy being learned.
How to Know the Onboarding Is Working
Progress shows up in what a coach can produce and demonstrate, not in how many tools they've tried. A coach genuinely progressing through the three phases should be able to point to concrete evidence at each stage, not just a growing sense of familiarity.
- By day 30: the coach has generated and critically revised AI output for a real coaching need, unassisted.
- By day 60: at least two modeled examples exist, tested on a colleague, ready to use in a live coaching session.
- By day 90: at least one full coaching cycle has been completed with a real teacher, with a debrief that fed back into the materials.
- Ongoing: teachers who've been through a cycle can describe the AI-assisted workflow back in their own words, not just recall that a coach showed them something once.
If a coach reaches day 90 without a completed cycle, that's usually a signal the onboarding plan needs a lighter phase-one and phase-two load, not that the coach lacks aptitude for the work.
What This Looks Like in Practice
Say you're an instructional coach supporting a middle school's science department, and the department wants help using AI to draft lab-report feedback comments. You're only two weeks into your own onboarding when the request comes in.
- Use the request to accelerate your own phase-one practice — generate sample feedback comments for a real upcoming lab report, not a hypothetical one, so your fluency-building and the coaching need overlap.
- Build one strong-prompt example and one weak-prompt example from that same real material, ready to show the department before the cycle formally starts.
- Start the actual coaching cycle small — one teacher, one class set of lab reports — rather than rolling it out to the whole department at once.
That overlap between your own onboarding and a real request is common, and it's not a problem — as long as you're honest with the teacher about still being early in your own learning, rather than presenting untested material as a finished, expert-vetted resource.
Common Pitfalls Specific to the Coaching Role
Coaches run into a slightly different set of problems than teachers do, mostly because of the added layer of supporting someone else's use of the tool, not just their own.
| Pitfall | Why It Happens to Coaches Specifically | What Helps |
|---|---|---|
| Modeling only successes, never mistakes | Wanting to appear confident and expert | Deliberately include one correction in every modeled demo |
| Coaching a workflow never personally tested | Time pressure to "just show something" | Never skip phase one for a topic before coaching it |
| Overwhelming teachers with every AI feature at once | Coach's own enthusiasm after building fluency | Introduce one workflow per coaching cycle, not the whole toolkit |
| Treating AI coaching as separate from existing coaching cycles | Framing it as a new initiative instead of folding it in | Embed AI into an existing coaching cycle's goal, not a standalone add-on |
The overwhelm pitfall is worth flagging specifically. A coach who just spent 60 days building broad fluency naturally wants to share all of it — but a teacher in a single coaching cycle usually absorbs one new workflow well, not five.
ISTE's guidance on AI use in schools consistently recommends documenting where a tool's output failed, not just where it succeeded, as part of any AI-related professional-learning material. Applied here, that means the failure-point log from phase one should end up embedded directly in the coaching materials built during phase two, not filed away separately.
Adapting the Plan for Common Real-World Constraints
A 90-day plan assumes a coach has a fairly typical caseload and schedule — which many don't. The phases stay the same regardless of a coach's specific situation; what changes is the pace and the scope of each one.
| Constraint | How to Adapt | What Doesn't Change |
|---|---|---|
| Part-time coaching role | Stretch each phase to six weeks instead of thirty days | Still complete all three phases before broad rollout |
| Coach serves multiple buildings | Pilot phase three with one building before others | Modeled examples from phase two transfer across buildings |
| Coach also carries a full teaching load | Fold phase-one practice into existing lesson prep | Personal fluency still comes before coaching others |
| Large staff with high demand for support | Train a few teacher-leaders as informal AI peer supporters | One coach can't run every teacher's first cycle alone |
A coach juggling multiple buildings or a partial teaching load is usually better served by a longer timeline than a compressed one. Rushing phase one to hit an arbitrary 30-day mark tends to produce a coach who can demonstrate a tool but can't yet troubleshoot it — which shows up the first time a teacher's session doesn't go as modeled.
Tools for Coach Onboarding
A coach's tool needs shift across the three phases, from personal exploration to something teachers can pick up quickly during a cycle.
| Tool Type | Best Fit | Phase |
|---|---|---|
| General AI chatbot (ChatGPT, Claude, Gemini) | Personal fluency-building, flexible experimentation | Days 1-30 |
| EduGenius | Coaching demonstrations using a saved class profile teachers recognize from their own use | Days 31-90 |
| Shared document or slide deck of modeled examples | The actual coaching artifact used in cycles | Days 31-90 |
EduGenius can be useful specifically in phases two and three, since a class profile set up once carries grade level, subjects, and ability range into every generated worksheet or quiz — letting a coach demonstrate a full, differentiated workflow in a single coaching session rather than rebuilding constraints from scratch each time. New users start with 25 welcome credits, enough for a coach to build phase-two materials before deciding whether to recommend the tool further.
Pro Tips for Coach Onboarding
- Document your own mistakes as you make them, not just your successes — they're the most valuable coaching material you'll build in the first 30 days.
- Resist coaching a tool or workflow you tried once. One successful generation isn't the same as understanding where it tends to fail.
- Ask a teacher what they actually need before choosing what to model. A coach's own favorite AI use case isn't always the one a specific teacher needs first.
- Build a shared folder of modeled examples as you go, so phase three doesn't start with a blank page.
What to Avoid
- Don't skip straight to coaching teachers without building personal fluency first. A coach demonstrating an unfamiliar workflow live is the fastest way to lose a skeptical teacher's confidence.
- Don't model only clean successes. A teacher who's never seen a coach correct a mistake may trust AI output uncritically once they're on their own.
- Don't try to onboard on every AI tool at once. Depth on one or two tools in the first 30 days beats shallow familiarity with five.
- Don't treat the 90-day mark as a finish line. AI tools and best practices keep changing; onboarding should transition into ongoing practice, not stop.
Key Takeaways
- Instructional coaches need a distinct onboarding plan because they must model, troubleshoot, and explain AI use — not just use it themselves.
- A 30/60/90-day structure works well: personal fluency, then coaching materials, then real coaching cycles with teachers.
- Each phase should produce something concrete — a finished example, a checklist, a completed cycle — not just accumulated familiarity.
- Deliberately modeling a correction, not only a clean success, builds more credibility with teachers than a flawless demo.
- The most common coaching-specific pitfall is coaching a workflow the coach hasn't personally tested and made a mistake in yet.
- Introducing one workflow per coaching cycle, instead of an entire toolkit at once, keeps teachers from feeling overwhelmed.
Frequently Asked Questions
How long should AI onboarding take for an instructional coach?
Roughly 90 days structured in three phases works well for most coaches: 30 days of personal fluency-building, 30 days turning that into coaching materials, and 30 days applying it in real coaching cycles. The exact pace can flex around a coach's existing caseload.
Should a coach wait until fully confident before coaching anyone on AI?
No. Waiting for full confidence can delay support indefinitely. A coach two or three weeks into onboarding can still coach a narrow, well-tested workflow, as long as they're transparent that a specific tool or task is still new to them too.
What's different about coaching AI use versus coaching other instructional practices?
The core coaching skills — modeling, questioning, debriefing — transfer directly. What's different is the pace of change: an AI tool's interface or capabilities can shift between coaching cycles in a way a well-established instructional strategy rarely does, so materials need more frequent revisiting.
How many teachers should a coach work with during the first coaching cycles?
Starting with one or two teachers, rather than a whole department, lets a coach refine coaching materials based on real questions before scaling. A pilot cycle that surfaces confusion is far easier to fix with two teachers than with fifteen.
Does a coach need to be an expert in AI before starting phase three?
No. Phase three is designed to start once a coach has working fluency and tested materials, not deep expertise. Coaching a specific, well-practiced workflow honestly — including being upfront about its limits — matters more than broad mastery.
Related Reading
- AI Professional Development for Teachers: The 2026 Guide
- How to Train Teachers to Use AI for Designing Assessments
- How to Train Teachers to Use AI for Creating Worksheets
- Building AI Confidence for ESL Teachers
- How to Integrate AI Into the Curriculum-Mapping Workflow
- How School Leaders Can Roll Out AI District-Wide
References
- Learning Forward — Standards for Professional Learning.
- Jim Knight and the Instructional Coaching Group — coaching-cycle and Impact Cycle frameworks.
- ISTE — guidance on AI literacy and responsible AI use for educators.