---
title: "Customer Education for SaaS: Teach Inside the Job, Not Before It"
publishedAt: "2026-11-18"
summary: "Customer education is usually an academy run before the work. The teaching that sticks happens inside the job, when the AI does a step and shows where it lives."
author: "Ben Skoog"
authorImage: "/ben_skoog.jpeg"
---

Customer education is the work a software company does so its customers can get value from the product without help: courses, an academy, webinars, help articles, in-app tours. The guides that rank for it treat it as a program you build next to the product and run before the customer needs it. Our view is different. Customers do not want a course. They want one thing done this week. The teaching that sticks happens inside that work, at the moment a setting matters, when the AI does the step in the customer's account and then shows them where it lives.

We did not start there. When we first built an agent that could do setup work through a product's own tools, I thought the education question had gone away. If the AI does the work, why would the customer need to learn anything? Then the software companies we work with told me, one after another, that they wanted their customers to learn the product anyway. They were right, and it changed what we built.

So this post makes one argument the education guides do not: customer education should stop being a separate program run before the work, and become part of the job the AI owns for each customer. The AI does the work, shows where it lives, and leaves the customer able to make the next change on their own.

## What the customer education guides recommend

The ranking pages agree on the formats. [Gainsight's guide to customer education tools](https://www.gainsight.com/blog/customer-education-tools/) centres on a learning management system, with online programs, certifications earned through quizzes, video and text content, instructor-led sessions and a community. [Archbee's SaaS customer education guide](https://www.archbee.com/blog/saas-customer-education) defines it as a structured, ongoing practice of helping customers get value at every stage, and lists help articles, video tutorials, in-product walkthroughs and checklists, academies and certifications, webinars and community forums. [MeltingSpot's piece on the strategic role of customer education](https://meltingspot.io/en/blog/the-strategic-role-of-customer-education-in-saas-business-success) makes the same case: education is continuous, not a one-time event, delivered through in-app tutorials, webinars, guides, courses and forums.

That is a fair description of the field, and for a large, complex rollout an academy is still useful. But look at when each format teaches.

<img src="/blog/customer-education-for-saas/formats-and-timing.svg" alt="Table of seven customer education formats and whether Gainsight, Archbee and MeltingSpot recommend each: courses or an academy, certifications, webinars, video tutorials, help articles, community forums and in-app tours. Almost all are recommended by all three. Our read of when each teaches: before the task, on a schedule, or on request. None of the guides lists the AI doing the step in the customer's account and then showing where it lives." />

*Formats recommended by the three ranking guides linked above. The "when it teaches" column is our reading, not theirs.*

Almost everything on the list teaches before the task or waits for the customer to go looking. The one format that lives in the product, the tour, still runs ahead of the work.

## Why teaching before the task does not stick

Memory research has said for more than a century that learning which is not used fades steeply. A 2015 [replication of Ebbinghaus' forgetting curve in PLOS ONE](https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0120644) by Jaap Murre and Joeri Dros found results similar to Ebbinghaus' original data: the curve drops fastest in the first hours and days, and after a month almost none of the benefit of having learned the material once was left. The experiment used lists of nonsense syllables, not software settings, so treat it as a direction rather than a measurement. Still, a customer who sat through a course in week one and needs the setting in week six is close to starting again.

Product research says the same about in-app tours. Page Laubheimer at Nielsen Norman Group, in [Onboarding Tutorials vs. Contextual Help](https://www.nngroup.com/articles/onboarding-tutorials/), writes that tutorials interrupt users, do not necessarily improve task performance, and are quickly forgotten. Help works better when it is triggered by a signal that the user needs it at that moment, and when it sits beside each step instead of asking the user to memorise steps in advance.

Workplace learning reached the same conclusion years ago. Josh Bersin's [learning in the flow of work](https://joshbersin.com/2018/06/a-new-paradigm-for-corporate-training-learning-in-the-flow-of-work/) argued for building learning into the systems people already work in, partly because his research found the average employee has only about 24 minutes a week for formal learning. A small business owner setting up your product between other jobs probably has less.

None of this says education does not matter. It says the timing is wrong. The best moment to teach a setting is the moment it is being set.

## What we got wrong, and what we built instead

Here is the part that comes from our own calls.

My first instinct was that customers just want the outcome, so learning was beside the point. I said as much to an enablement leader at a large digital experience platform in August. He disagreed, politely. For anything more complex than a simple product, he said, a customer has to understand how their setup is built to own it, handle the edge cases and make their own changes. A screen with the setting on it, laid out visually, is faster to take in than a long text reply. His company was moving its simpler tiers from static walkthrough videos to a conversational agent inside the platform, and still kept people for the big deployments.

He was not alone. A hospitality software company we demoed to in October could have given an MCP server write access to set every setting for its customers. They chose not to make that the default, because a customer who never sees the settings never learns the product and comes back to the AI for every change. They wanted the AI to take the customer to the right page and point at the setting, so the customer knows where it is next time. A campground reservation software company told us the same in August: its click-through setup is deliberately teaching, because customers who have done it once need less support later. And a learning platform company we work with, a company that sells education software, had the clearest version of the problem: prospects logged into an empty sandbox, did not know where to start, and left. A blank account teaches nothing.

So we changed the design. Our agent does the work in the background through the product's own tools, signed in as the customer with their permissions. When it is done, it brings the customer to the screen where the work landed and walks them through what it did. The vendor decides, step by step, which steps the AI does silently, which it does and then shows, and which it walks the customer through on screen. We covered how to draw that line per action in [agentic UX](/blog/agentic-ux). For customer education, the point is simpler: every step the AI does for a customer is also a chance to teach one thing, at the one moment the customer cares.

I still think the customer's first wish is the outcome. What I learned is that the vendor's wish, a customer who can find their own settings, is not in conflict with it. It is what makes the outcome last.

## Customer education as part of the job

A service, the way we use the word, is a job the AI owns for each customer and keeps working, like getting a new customer live or keeping an account's setup healthy. It is not a task or a scheduled report. Teaching fits inside that job naturally:

- **Teach the step that will come back.** If a customer will need to change a setting again (a price, a schedule, a rule), the job does it with them or shows it after. If nobody will ever touch it again, the AI just does it.
- **Teach when the state slips, not on a calendar.** A job that notices a customer stuck halfway through setup can explain the one step they are missing, on the screen where they are missing it, instead of sending another webinar invite.
- **Start from work, not from a blank account.** The learning platform's empty sandbox is the opposite of education. The first thing a new customer sees should be something done in their account, with the reason it was done.
- **Keep the human path open.** The enablement leader was right that large, complex deployments still need a person. The job should hand those to someone on the vendor's team, with the context of what the customer has and has not done.

We wrote about the first of these jobs, getting each new customer live, in [SaaS implementation as a job](/blog/saas-implementation-as-a-job). Customer education is what makes that job's result hold after the AI steps back.

## What this means for a customer education team

If you run customer education today, your academy is not wasted. Courses and certification still serve the power users, partners and admins who want to go deep, and help articles are what both people and AI agents read when they need an answer.

What changes is where most of the teaching happens for the average customer. Less of the team's time goes into producing courses that are forgotten by week six. More of it goes into deciding which steps a customer should learn and which they never need to, writing the short explanations the AI gives at those moments, and reading where customers keep getting stuck so the next version of the job teaches that step better. The education team becomes the people who decide what the customer should know when the work is done.

There is one honest limit. This only works as far as the product exposes. If a step has no tool behind it, the AI can only point at the screen and explain. That is still better timed than a course, but it is not the same as doing the work with the customer.

## FAQ

### What is customer education in SaaS?

Customer education is everything a software company does to help customers get value from its product on their own: courses and academies, certifications, webinars, help articles, video tutorials, in-app tours and communities. The newer view is that much of it should happen inside the work itself, at the moment a customer needs a step, rather than in a program run ahead of time.

### Why is customer education important?

Customers who understand how their account is set up can make their own changes, need less support and are more likely to get lasting value. The question is less whether to educate and more when: teaching that happens long before the task is mostly forgotten by the time it is needed.

### Does an AI that does the work replace customer education?

No. It changes the timing. When the AI does a step for a customer and then shows where it lives, the customer learns that setting at the moment it matters. The software company decides which steps to do silently, which to show and which to walk through, based on whether the customer will need that setting again.

---

Aimdoc runs the jobs your customers need done, like getting each new customer live, and shows them where the work lives as it goes. See how [services](/services) work, or [start free](https://app.aimdoc.ai/signup).
