By Dr. Janette Camacho | February 10, 2025

In January 2025, I stood at the podium at FETC for the second consecutive year as a Featured Presenter. My session title was "From Fear to Fluency: Coaching Teachers Through the AI Transition." The room was standing-room only. But the most instructive moment came not during the session itself but in the hallway afterward, when a curriculum director from a mid-sized Georgia district approached me and said: "We sent every teacher to a three-hour AI workshop last August. By October, maybe 15% were still using what they learned. What did we do wrong?"

She did not do anything wrong, precisely. She did what most districts do: delivered information and expected transformation. But information is not coaching, and workshop attendance is not adoption. After training over 500 educators across Florida and nationally since the fall of 2023, I have developed a coaching model built on patterns of sustained change rather than spikes of enthusiasm. This article describes that model and the evidence behind it.

Why Workshops Fail

The research on professional development effectiveness is unambiguous and has been for decades. Joyce and Showers established in the 1980s that fewer than 10% of teachers transfer a new skill to classroom practice after workshop-only PD. When coaching is added - ongoing, job-embedded, classroom-connected support - the transfer rate rises above 90%.

AI training is no exception to this pattern. If anything, the transfer gap is larger because AI tools evolve so rapidly that skills taught in August may be functionally obsolete by November. A teacher who attended a workshop on "Using ChatGPT-3.5 in Your Classroom" in early 2024 found themselves working with a substantially different tool by mid-year when GPT-4 became the default. Without coaching support to navigate those changes, it is entirely rational for teachers to abandon the effort.

The Four-Phase Coaching Model

The model I have developed operates in four phases, each approximately four to six weeks long for a complete cycle. I call it the GATE model: Ground, Apply, Transfer, Extend.

Phase 1: Ground (Weeks 1-4)

Grounding establishes conceptual understanding before tool proficiency. In the first phase, I am not teaching teachers how to write prompts. I am building their mental model of what AI is, what it does, and - critically - what it does not do.

Key activities in the grounding phase include:

Hands-on AI exploration. Teachers interact with multiple AI tools (ChatGPT, Gemini, Claude, Copilot) without any instructional objective. They ask it questions about their hobbies, test its knowledge of their content area, try to make it produce errors. This unstructured exploration reduces anxiety and builds intuitive understanding faster than any lecture.

The "AI audit." I ask each teacher to log every moment in their professional week where they think, "This is tedious but necessary." Lesson planning templates. Email replies to parents. Data entry for progress monitoring. Rubric creation. This audit surfaces the teacher's specific pain points, which become the foundation for Phase 2.

Misconception surfacing. Through structured discussion, I identify and address common misconceptions: that AI "knows" things (it generates statistically probable text), that AI output is reliable by default (it requires verification), that AI will replace teachers (it will change the job, not eliminate it), and that using AI is "cheating" for a professional (it is a tool, like a calculator or a search engine).

The grounding phase ends when a teacher can explain, in their own words, what AI does well, what it does poorly, and where they see the highest-value opportunities in their own practice. This articulation demonstrates readiness for application.

Phase 2: Apply (Weeks 5-8)

In the application phase, each teacher selects one specific use case from their AI audit and commits to implementing it for four weeks. One use case. Not five. Not "integrate AI across my curriculum." One well-defined application that addresses a genuine pain point.

The most common initial use cases I see are:

During this phase, I meet with each teacher weekly - either in person or via video - for 15 to 20 minutes of focused coaching. These meetings follow a consistent protocol: What did you try this week? What worked? What did not? What will you adjust? The brevity is intentional. Coaching should feel like a conversation, not an evaluation.

I also facilitate a cohort meeting every two weeks where teachers share their experiments with peers. These sessions are essential because they create social accountability and collective problem-solving. When one teacher discovers that AI-generated rubrics need careful editing for grade-level appropriateness, every teacher in the cohort benefits from that insight.

Phase 3: Transfer (Weeks 9-12)

Transfer is where the coaching model diverges most sharply from traditional PD. In this phase, I ask teachers to apply their AI skills to a context they did not practice during Phase 2. A teacher who used AI for rubric generation now tries it for lesson planning. A teacher who generated differentiated texts now experiments with AI-assisted feedback on student writing.

The pedagogical principle is transfer of learning - the ability to apply a skill in a new context. If a teacher can only use AI for one specific task in one specific way, they have not developed AI fluency. They have memorized a procedure. Transfer activities reveal whether the teacher has internalized the underlying competency: the ability to assess an instructional need, determine whether AI can address it, construct an effective prompt, evaluate the output, and integrate the result into practice.

This is also the phase where resistance most commonly surfaces. Teachers who felt confident with their Phase 2 use case sometimes feel destabilized when asked to transfer. Good coaching anticipates this and normalizes the productive discomfort of learning.

Phase 4: Extend (Weeks 13-16)

In the final phase, teachers begin extending their AI integration to student-facing applications. This is deliberately last. I do not want teachers introducing AI to students until they have developed their own fluency and judgment.

Extension activities include designing AI-integrated assignments using a responsible use framework, teaching students to evaluate AI output critically, and creating classroom norms for AI use. I facilitate model lessons where I demonstrate AI integration in the teacher's own classroom, with their students, so they can observe before they lead.

The extension phase also includes a sustainability component: teachers document their AI integration practices in a shared resource library that persists after the coaching cycle ends. This library becomes institutional knowledge rather than individual expertise, reducing the impact of teacher turnover.

What the Data Shows

I have collected pre- and post-coaching survey data from 523 teachers across seven coaching cohorts. The results support the model's effectiveness across several dimensions.

Self-efficacy with AI tools (measured on a 5-point Likert scale) increased from a mean of 2.1 (pre) to 4.2 (post). More importantly, at a six-month follow-up, the mean remained at 3.9, suggesting sustained rather than temporary confidence gains.

Frequency of AI use for professional tasks increased from an average of 0.8 times per week (pre) to 4.3 times per week (post). At the six-month follow-up, frequency was 3.7 times per week - a slight decline from the coaching period but dramatically higher than baseline.

Student-facing AI integration - the most complex outcome - increased from 6% of teachers (pre) to 64% (post). This is the metric I consider most meaningful. Moving from personal use to pedagogical use represents a qualitative shift in a teacher's relationship with the technology.

Scaling Challenges

The model works. Scaling it is the hard part. Each coaching cycle requires approximately 15 to 20 hours of my time per cohort of 20 to 25 teachers. That is a fraction of the cost of the learning gains it produces, but it requires districts to invest in coaching positions rather than one-time workshops. The cost per teacher is higher; the return per teacher is incomparably greater.

I am currently developing a train-the-coach program that equips instructional coaches within districts to deliver the GATE model independently. Early results from two pilot districts are promising, and I expect to scale this through iTeachAI's professional development offerings by summer 2025.

The Fundamental Lesson

The most important thing I have learned from training 500 teachers is this: the technology is not the barrier. Fear is not the barrier. Time is not the barrier - though it is a real constraint. The barrier is isolation. Teachers who try to learn AI alone usually quit. Teachers who learn in a coached community almost never do.

If you are a district leader, invest in coaching, not workshops. If you are a teacher, find your people - a colleague, an online community, a professional network. The AI transition is too big and too fast to navigate solo. But with the right support structure, it is not only manageable. It is transformative.

Dr. Janette Camacho is a Google for Education Certified Trainer & Coach, Google Certified Educator Level 1 & 2, Adobe Creative Educator, Apple Teacher, FETC 2024 and 2025 Featured Presenter with 28+ years of K-12 classroom experience. She is the founder of iTeachAI.