Lessons from Our Labs: 7 AI Moves for Customer Education Teams

Last month, my co-founder, Akash, and I had the opportunity to speak at CEdMA’s empowerED25 conference in Austin, TX.

Two recent-grads, who’d never clocked a day at another company before starting Clueso.

Presenting to an audience of 60+ customer education experts - people who know CEd like the back of their hand.

We couldn't teach them anything about instructional design or learning outcomes they didn't already know. We should have been in the audience taking notes from them!

But we could talk about one thing we know intimately: AI.

We spend our days knee-deep in multimodal AI: LLMs, voice and vision models, reasoning models, AI avatars - you name it, we’ve probably tried it. And we've spent countless hours talking with customer education leaders about their real challenges: creating and scaling consistent content, proving program ROI, and making the most of limited resources.

So we ran non-stop experiments with AI technologies, to test them for some of the most crucial CEd workflows. Our goal wasn't just to make incremental improvements - we wanted to completely reimagine how CEd teams could operate with AI.

What we shared at empowerEd - and what I’ll share here - is the distilled version of everything we’ve learned on that journey.

AI is NOT Just the Same Job, Faster

There’s a classic adoption pattern when it comes to new technologies. And AI analyst and author, Benedict Evans, sums it up best:‍

‍“First you make the tool fit the work. And over time, you change the work.”

AI makes this feel very real. We’ve got two things happening at once:

1. We can change how we do our work (quicker drafts, automated translations, data-driven insights - stuff that felt impossible a year ago).
2. We can change what work we do in the first place (new ways to teach, new ways to collaborate with our teams).

I like to think of that second point as a whole new world opening up for Customer Education.

It’s not just about plugging ChatGPT into your workflows and calling it a day. It’s about recognizing that your job itself is changing.

That’s why you should know what’s possible, without giving into the AI overwhelm.

We’ve all given ChatGPT or Claude a spin for content creation. Some CEd teams have used it to draft help articles or write training scripts. But that’s just scratching the surface. The gap between what's possible with AI and what most teams are actually implementing is enormous.

Here are 8 AI moves - tried, tweaked, and tested - that CEd teams can put to work today and level-up.

Move #1: Build Your Own AI Minion (aka a Custom GPT)

If you’re already using ChatGPT or Claude to draft help articles, video scripts, or emails, you  know how tiresome it can get to keep typing the same prompts:

“Write in a more conversational tone.”
“Use a how-to format.”
“Keep paragraphs short.”

Especially if you’re working with swathes of content.

Custom GPTs let you build your own “AI assistant” that remembers all your preferences and brand guidelines. It’s how you stop fussing over prompt details and start getting consistent first drafts right off the bat.

The process is surprisingly simple:

Step 1: Create a new GPT in ChatGPT Plus

Step 2: Upload your brand guidelines, style guide, or examples of content you love

Step 3: Define specific instructions (tone, format, do's and don'ts)

Step 4: Test it with a few prompts and refine as needed

Step 5: Share it with your team

Here’s a quick walkthrough of how you can create, test, and refine your customGPT:

CEd teams can create Custom GPTs for a bunch of specialized tasks. Some interesting workflows I’ve come across include:

• Course outline generators that turn learning objectives into full course structures
• Assessment creators that generate different question types for knowledge checks
• Support ticket categorizers that help prioritize what content to create next

‍Pro tip: Two ways you can get your CustomGPTs to become even more powerful:
‍
1. Ask your ChatGPT to help you make your prompts for your custom GPT.
2. Everytime your customGPT gives you a good output, let it know that it’s done a good job by giving it to the custom GPT as an example.

‍

Move #2: Configure Custom GPTs as Instructional Design Co-pilots

Once you're comfortable with basic Custom GPTs, you can explore a more sophisticated use case. Think about all those hours you spend creating course outlines and mapping out training objectives. Custom GPTs can shortcut that.

Instead of starting from scratch every time, you can tell your co-pilot what your learners need and what you want them to learn, and let it propose a cohesive outline or learning path.

• Get specific: Provide clear instructions for extracting the input as well as the output. This provides your co-pilot with context for both what the course would be about - topic, learning objectives etc - as well as what the final course output should look like. Include examples of the style or level of detail you want. See these prompt examples for training instructions:

{{2-image-layout}}

• Iterate fast: Ask the model to propose multiple course designs and see which resonates.
• Refine: Inject your own examples, scenarios, or brand stories.

Here’s a video of me creating an instructional design co-pilot using the prompts above:

The end result? You’re not just saving time - you’re exploring more course ideas than you ever could manually.

‍

Move #3: Let AI Agents Do Your “Routine Recon”

Imagine what you can do by letting AI agents test your CEd site at scale, before you bring in human testers. It’s a quick way to spot the big navigation snafus - can users actually find what they're looking for?

This is where AI agents like OpenAI's Operator come in. These tools can simulate real users navigating your academy, helping you spot UX issues without organizing focus groups. Here’s what an example workflow for this might look like; this was created for Salesforce’s Trailhead portal:

1. We gave Operator a task to go to the Trailhead portal and find courses on developing Sales Management skills:

2. It spins up a virtual browser and attempts to complete the task.

3. We watched where it struggled and identified confusing navigation paths by looking at the list of all the steps it’s performed.

We ran the same test multiple times to identify consistent pain points

How we used this at Clueso

We used this approach with Operator on our own platform to see if our users could easily find our translation feature.
[Spoiler: they couldn’t!].

We followed the same steps as we did for the Salesforce example above to better understand where the friction points in the user journey were.

This is what the Clueso UI looked like earlier:

Now, the UI looks like this:

As you can see, we’ve put a clear “Translate” text on the button here to make it more obvious, and successful completion rates have seen an upward trend since.

And this is not just useful for product managers. For CEd teams, this can help you find the most confusing steps of a given workflow, and which steps to focus on in detail while creating e-learning content.

While this doesn't replace actual user testing, it gives you quick insights on which parts of your academy might confuse real users.

‍

Move #4: Turn Raw Data into Insights Without a Data Analyst

If you’ve ever tried to prove the ROI of your training program to the C-suite, you know the pain. And others in the CEd industry agree:

You’ve got course completion stats in your LMS, renewal info in your CRM, NPS data somewhere else. And pulling it all together can be an epic headache.

But advanced AI models with solid “reasoning” can parse your CSV files, join up the data, and reveal correlations. You can:

1. Download raw data from your LMS (completions, registrations)
2. Export customer data from your CRM (renewals, churn dates, NPS scores)
3. Feed this unprocessed data to an AI reasoning model.

Consider this example. We uploaded two CSV files with dummy data. One with sample customer data with revenue numbers, renewal dates, churn dates, NPS scores, etc that you'd typically get from your CRM. And another with sample course completion rates that you'd usually extract from your LMS. And we asked ChatGPT to analyse a few parameters basis the training completion rates for these customers:

Here’s a video of ChatGPT's initial analysis as well as the data charts it generated:

If you need to make specific correlations for business impact, you could also ask: "Is there a correlation between customers who complete our onboarding course and their renewal rates?"

The AI can run the analysis and provide insights that would typically require a data analyst.

‍

Move #5: Break One Long Video Into Dozens of Bite-Sized Videos

Subject Matter Experts (SMEs) are precious. You’d hate to burn their time on a million tiny recordings. Instead, have them record one big chunk, maybe a 30-minute deep dive on your product or a key training concept. Then use AI to:

• Auto-split the content into shorter segments.
• Generate polished scripts or bullet points for each segment.
• Add professional voiceovers if you want.
• Produce micro-videos or help articles.

Psst! We’re building a feature called “Clueso Cuts” to make this process even smoother. The big idea: get multiple pieces of training material from one SME session or a webinar, that can be easily digested and converted into courses and articles for your help center. Here's a sneak peek.

‍

Move #6: Localize Everything, Faster

If you have a global audience, you know translating training videos and documents is no joke. And translation agencies are pricey and slow.

AI translation has improved dramatically in the past year, making it possible to:

• Translate video scripts while maintaining context and technical accuracy
• Generate natural-sounding voiceovers in different languages and accents
• Create localized help articles that preserve formatting and screenshots

We regularly see teams auto-translate entire libraries into 30+ languages faster than you can say “Duolingo!”. You might still want a human to do a final pass for cultural nuances, but 90% of the heavy lifting can be done by AI.

And a shameless Clueso plug would be obligatory here. Look at Clueso translates content into 37+ languages:

‍

Move #7: Use AI for Creative Content

Sometimes your text-heavy training materials need visuals - diagrams, charts, infographics - to illustrate complex learning objectives or map out an entire learning path. That’s where tools like Napkin.ai come in. You paste in a block of text and the AI converts it into a spiffy, shareable graphic.

Look at this example workflow where I used Napkin to create an infographic for a learning path:

Here are some other ways CEd teams can use Napkin:

• Visualizing learning objectives: Drop in your bullet points and watch Napkin generate a clean, digestible flowchart or infographic.
• Building learning path graphics: Summarize your key modules or lessons, and it'll turn that text into a visual roadmap.
• Iterating on designs: Tweak the text, add or remove key details, and Napkin will refresh the graphic on the spot.

Where Is This All Heading?

Bill Gates called GPT “the most important advance in technology since the graphical user interface (GUIs)”.

Not PCs. Not the Internet. Not mobile phones.

Yes, AI can churn out help articles faster. Yes, it can do simple code-like tasks. But the real transformation is more profound. AI changes how you build your entire Customer Education strategy - understand what content users need, test user paths, and analyse data.

Pretty soon, we’ll see “agentic workflows” become the norm - AI that doesn’t just help with tasks, but orchestrates entire processes, like:

• Noticing a new product feature in your UI
• Updating the relevant help articles automatically
• Taking new screenshots
• Then giving you a friendly nudge: “Hey, I updated everything - do you approve?”

This is closer than you might think.

“Will AI Replace Me?”

I won’t sugarcoat it: AI is changing what we do and how we do it.

But just like the shift from typewriters to PCs, it’s not a Wellsian-world scenario of erasing jobs. Jobs are evolving. New roles are popping up. New skills will matter more.

My two cents: AI is a multiplier. Don’t just use it to do your job the old way. Lean into the parts of your work that AI can’t replicate - your real-world empathy, your deep knowledge of your customers, your ability to shape stories that resonate. AI can crank out draft content, but it can’t replace your intuition or expertise.

Getting Started (Without Getting Overwhelmed)

1. Pick One Problem to Solve. Maybe you want faster turnaround on help docs, or you need to measure course impact. Focus on that first, so you don’t drown in random experiments.
2. Train a Custom GPT. Feed it your style guide, brand voice, and best examples. It’ll save you time and keep content consistent.
3. Experiment, Evaluate, Iterate. See what works. Ditch what doesn’t.
4. Keep the Human Touch. AI can draft, analyze, and even navigate your platform. You’re still the one deciding if it’s good enough.

Above all, don’t let the hype (or the fear) paralyze you. That's what these moves I shared above are for. Start small, learn from each experiment, and double down on what actually helps your customers learn.

At the end of the day, that’s what Customer Education is about: helping people succeed with your product. AI is just another tool in our kit - a powerful, blink-and-you-miss-the-update-evolving one, yes, but still just a tool.

And we should all be using it to do our best work.