AI will write a course on any topic you name, working only from what the model already knows. Chris Badgett, founder and CEO of LifterLMS, uses AI for online courses the other way around: his own knowledge goes in before AI builds anything, and a person reviews the result before it ships.
When you use AI for the same task again and again, two things help each run get closer to what you want: update its instructions with the corrections you made, and give it a memory that carries over between sessions.
Start with your own knowledge
Putting your knowledge in first is what Chris calls the human knowledge dump, the most important lesson in his YouTube video on creating an online course with ChatGPT.
He recommends opening with this prompt: “I want you to interview me until you're 95% certain you have my unique context, points of view, resources, influences that have inspired me…” In the video's example, his conversation with AI and the pillar sources he pasted in became roughly a 60-page Google document, and only then did he make the course. Most of his AI work looks like this: he probably spends far longer talking to it and dropping in files than it spends processing, and since he types slowly, he dictates through Wispr Flow. The model fills in the gaps from there.
LifterLMS's Name Your Price add-on shows the same habit in product work. It lets people buying courses and memberships pay flexible prices, for support-the-creator pricing, regional pricing or price discovery. Instead of pitching the idea in a meeting, Chris, a non-developer, showed up with a working plugin he had built with AI. That work was mostly conversation about why the add-on was needed, what it should do and which patterns to borrow from LifterLMS's group memberships add-on.
Review the last 10%
Without enough context, AI fills the gaps with its own assumptions, and the result can still sound plausible. To someone who isn't a copywriter, AI sales copy can look like it will sell. A copywriter with ten years in the job and multi-million dollar sales pages behind them sees copy that won't.
Chris uses the 10-80-10 principle for that gap. A human puts in the first 10%, AI does the heavy lifting in the middle, and a human reviews the last 10% to make sure it is solid. Skip the front 10 and the output goes generic. Skip the end and you are crossing your fingers. For Name Your Price, Chris wanted his engineering team, working with their own AI, to do that last 10%: review the code and make it secure before it shipped, or throw his version away.
Give AI context, then a memory
Maurice is the AI agent LifterLMS built to hold the company's context. It has every line of LifterLMS code, every YouTube video and podcast episode the company has published, twelve years of support conversations in Help Scout, Google Analytics and Stripe metrics, and twelve years of ActiveCampaign CRM history. Context is the name of the game for Chris: ChatGPT or Claude out of the box are useful, and an agent with that much context can do far more. Chris likes that Maurice reads the code itself: blog posts and documentation can fall behind the software, and in his view the code is always the single source of truth.
The challenge with a second brain like this is getting what you know into it. Most people's knowledge is spread across tools and other people, and a team multiplies that: people who talk to customers, people who write code, each with their own sources. Connecting those tools to the agent is how LifterLMS does it.
Maurice wasn't customer-facing at the time of the conversation, but LifterLMS has opened customers' own sites to AI. Connect a site running LifterLMS 10.2 or later to Claude or ChatGPT through an MCP connector (LifterLMS documents the setup) and you can work on it by talking to it instead of clicking through every setting. What the AI can do depends on what the site exposes: it can grade assignments if you run the Assignments add-on, for example, or read order data and suggest a marketing strategy. For Vic Dorfman, who hosts MemberFix Radio, the appeal is handing off jobs that used to eat up time, like pulling reports and finding the insights in them: set the goal, and AI works on it overnight.
The AI's instructions also have to improve over time, as Vic saw when he started using AI to edit his videos. Before the first edit, he had it research how AI should edit and download skills to build its own set of rules from. That first edit took about seven hours to run, much of it spent on the research and downloads. After a feedback loop with his producer and editor, an edit took an hour and a half and the result was substantially better. For Chris, that feedback loop is the most important part. In his version, you do the thing, do it again, show the AI the version that was actually published and have it update its skill, so future runs come out closer.
A memory system carries that iteration from one session to the next. LifterLMS uses AutoMem, built by Jack Arturo of WP Fusion (Jack has been on MemberFix Radio to talk about AI agents for membership sites). It gives AI a persistent memory across sessions, storing what matters and linking related ideas, so the agent knows when it has seen a situation before. Chris calls it context on steroids.
What a course adds when learners can ask AI
For Chris, one of the biggest challenges in e-learning is that “a lot of people will go to AI to learn, maybe instead of a course or a membership.” Selling pure information, like a Udemy course, is what he expects to get really challenging going forward, since AI already has the information that's publicly available in a lot of courses like it.
What he would add on top of the information is curation for a specific customer avatar or ideal customer profile, and support mechanisms such as personalized support, coaching and community, though he doesn't think every course needs all of them. Cheap, decent-quality output will have a huge market, in his view, and human-made craft will be a niche that still has demand. He thinks membership site creators should ask how they are “indisputable sages or craftsmen of their craft. Without technology.” He put it in climbing terms: beginners show up with all the gear and bring less as they get better.
What this means for a membership business
Keep a human at both ends of the work. Chris applies 10-80-10 to copy and code alike: a human puts in the context, AI does the heavy lifting, and someone who knows the craft reviews the result. In a membership business that covers lessons, sales pages and member emails. Before AI drafts, have it interview you, or attach the material your business already holds, such as your course library, past emails or a folder of PDFs, so it works from your sources instead of filling the gaps with its own. Before members see the result, have the person who owns that work check it.
Build a feedback loop into repeated AI work. Member emails, lesson outlines and reports all come around again. After each one goes out, give the AI the final version and have it update its instructions, and add a memory system such as AutoMem if you iterate on the same work across sessions.
Point AI at your weak spots. For owners who don't want to become AI power users, Chris's advice is to be honest about where they're weak and ask AI there, whether that's getting a stalled project over the line or setting up a hosting option or an email automation. LifterLMS's support queue was down about 75% when Chris and Vic spoke, with no drop in usage, and Chris thinks the reason is that customers ask AI before they open a ticket. Without a technical background, you can't always judge whether the answer is right, though Vic thinks the right context is making that less of a barrier.
Know what you add on top of the information. List what members pay you for and mark the parts a general model could already give them. Curation for your ideal customer profile and human support, such as coaching or a community, are the layers Chris would add, and not every course needs all of them.
For more on where AI fits in a membership business, see AI for Membership Sites: What to Add, What to Leave Alone, and for other course platforms on WordPress, the best online course platforms for your membership site.
Chris Badgett is the founder and CEO of LifterLMS and hosts the LMScast podcast. Find him at lifterlms.com.
Frequently asked questions
Does Chris Badgett use Obsidian for his second brain?
Not at the time of the conversation. The LifterLMS team used its AI agent, Maurice, to hold the context, and for Chris a second brain meant skills libraries, project folders, automated tasks that repeat and, above all, integrations that pull the company's tools into one place. He had been looking at Obsidian and thought something like that would be very cool.
Can LifterLMS customers or agencies use Maurice?
Not at the time of recording. LifterLMS hadn't made Maurice customer-facing yet, though Chris wanted to. For an agency building a LifterLMS site, Maurice would be straightforward about what is possible out of the box. When a client is happy with nearly everything and wants one more thing that doesn't exist, it could coach the agency through the custom plugin, tweaks or hosting configuration to close the gap.
Do you need to be a developer to install an AI memory system like AutoMem?
Not necessarily. Chris didn't know how to install a memory system either, and his advice is to ask the AI itself: tell it you want to use AutoMem, ask it to talk to you like a 12-year-old, and ask again whenever a step doesn't make sense. Vic set it up the same way with no Docker experience, working through it in Terminal for about an hour.
What is LifterLMS working on next?
Continuing on the same path, faster. Chris says the team is moving ten times faster than before, mostly because of AI and, of course, a talented team. In tech companies, he sees the bottleneck moving: engineering used to be it, and marketing, sales and lead generation are the next frontier for using AI. His own focus is helping more education entrepreneurs do what they want to do and discover LifterLMS as an option.
Running a Membership or eLearning Platform on WordPress?
MemberFix provides managed WordPress operations for membership, eLearning, and community platforms — so your team never has to worry about the tech breaking during a launch or losing members to site issues. We handle development, support, updates, and monitoring with SLA-backed guarantees. Trusted by industry leaders like Echelon Front, TechServe Alliance, and hundreds of others since 2014.
About MemberFix Radio:
We interview experts and share insights for membership site owners and subscription business leaders building sustainable, profitable recurring revenue models.
