How I Actually Use AI in Customer Education At Zapier (with 5 Workflows to Make Your Own)
By Eric Mistry, Customer Success Manager + AI & Automation Transformation Lead at Zapier
AI in customer education is in a weird place right now. It’s both overhyped and underused.
I spend most of my time working on AI and automation at Zapier, and working directly with folks trying to bring both into their daily work. The initial gap is pretty consistent. There’s a lot of curiosity and some experimentation, but very few things have actually made it into day-to-day work.
Over the past year, I’ve been building and testing this in real environments, including my own work, bootcamps, customer programs, and internal teams. Some things worked. Some very much did not.
What follows are five workflows that have held up. These are not demos or one-off prompts. They are things you can actually use this week to ship faster, get better insight, and spend more time on the parts of the job that are actually human.
1. Course Outcome Alignment
Most courses drift over time. You start with clear learning objectives, then content gets added, updated, patched, and reused. A year later, you have something that feels right but hasn’t been checked against what it is supposed to teach.
The workflow is simple. Use AI to compare your course content directly against your intended learning outcomes. This removes a lot of expert blindness and gives you review capacity you probably do not have.
In practice, you export your course content as Markdown or HTML, provide your learning objectives, and ask AI to map content to each objective. From there, you can have it flag gaps, overemphasis, or areas that are missing entirely.
Markdown matters more than most people expect here. (Use pdf as a last resort.) Clean structure leads to much better analysis.
2. Source Pack Consolidation
Course inputs tend to live everywhere. Docs, tickets, Slack threads, recordings, SME notes, old decks, and half remembered conversations all end up in the mix. The result is that you are constantly rebuilding context every time you touch a course.
The workflow is to create a single, structured source pack that becomes the canonical input for everything. AI performs dramatically better when it has one coherent document instead of scattered inputs across tools.
To do this, gather all relevant materials for a course and combine them into a single Markdown document. Organize it into sections that reflect how someone would actually want to learn or build from it. Once that exists, you can reuse it for drafting, analysis, and iteration without starting from scratch each time.
This tends to feel like extra work until you do it once. After that, it becomes the default.
3. Connect Learning Data to Product Reality
Learning data and product data usually live in separate systems and rarely get meaningfully connected. You can see course completion rates and you can see product activation, but tying those together often turns into a project that never quite happens or needs heavy handholding from an already overtaxed data team.
The workflow here is to use AI to synthesize learning data, product usage, and support signals into a single view. This is where customer education starts to look less like a support function and more like a growth lever.
A practical way to approach this is to pull learning metrics such as completion, drop-off, and feedback, combine them with product data like activation and usage patterns, and layer in support signals such as ticket themes or recurring issues. From there, you can ask AI to identify correlations, patterns, and potential hypotheses.
Questions that tend to produce useful output include which lessons correlate with faster activation, where learners drop off and why, which segments complete and succeed, and what support issues could be prevented with better education.
You do not need perfect data to start, you just need enough to begin asking better questions.
4. The “And Then” Chain
A common pattern with AI is that it gives a decent answer and then the process stops. Most people use it for a single step such as summarizing or drafting, and never push further.
The real value shows up when you keep going.
This workflow is about chaining your thinking so that each output leads to the next question. For example, you might start by identifying your best performing lessons, then compare them to lower performing ones, ask what differs between them, extract repeatable patterns, and turn those into updated standards or templates.
One simple tactic that works well is to end your prompt with “and then what?” It nudges the interaction forward from observation into action and keeps the loop moving.
5. Prototype Learning Experiences Fast
Interactive learning often sounds like a great idea until you look at the effort required to build it. As a result, many courses default to content followed by a quiz and stop there.
The workflow here is to use AI to prototype learning experiences quickly so you can test ideas before investing heavily in them. The goal is not to produce something perfect, but to create something concrete enough to react to.
You start by defining the learning objective clearly, then describe the type of interaction you want, whether that is a scenario, a decision point, or a practice exercise. From there, you can generate a simple prototype such as a branching scenario, a lightweight interactive exercise, or a small game focused on concept reinforcement. (Protip: ask for the output to be a single html5 output!)
Prototypes tend to change stakeholder conversations. It is much easier to react to something tangible than to a description of what might be built.
Closing
You do not need to implement all of this at once. Pick one workflow, try it this week, and see where it holds up or breaks down.
Most teams do not need better prompts. They need a small set of actually useful workflows that make their work better, and they need to get into the habit of using them consistently. This is a good place to start!
Eric Mistry is Zapier’s AI and Automation Transformation Lead, focused on turning AI hype into habits that stick. He helps teams adopt AI and automation in practical, human-centered ways that save time and reduce toil. He also co-leads Zapier Academy and writes the Customer Education Bi-Weekly Newsletter. Known for translating between technical and non-technical audiences, Eric guides organizations from experimentation to adoption with repeatable workflows.
