Fast, cheap, good. Most of us will remember the mantra we were taught that you can have two out of three of these:
Want something cheap and fast? It won’t be good. Want something fast and good? It won’t be cheap. You can never have all three.
Harken back to the early days of the AI boom, though, and it felt like we had finally unlocked the holy trinity of the fast-easy-good triangle. Going from ChatGPT 3 to ChatGPT 4 was a huge leap in what we perceived AI could accomplish. The price? $20 a month. Want to roll it out to your entire organization with no limits? $40 per person a month. An insanely low price given what you could do with it.
Fast and good, but not cheap?
Obviously, once we had something that was both insanely powerful and insanely cheap, there was a rush to put it in everything. The eLearning industry was not immune to this either. Walking the expo at DevLearn in those early AI years, every authoring tool became an AI-powered authoring tool. Every LMS has an AI recommendation or skills engine to deliver just the right training at just the right time.
I don’t say this in any kind of mocking tone. The cheapness and ubiquity of AI has ushered in a refreshing surge of rapid innovation in the eLearning industry, where before, you could not fault anyone for describing it as stagnant. But it’s increasingly looking like the cheap part of the fast-easy-cheap unlocking was a lie.
In their efforts to drive adoption and growth (and fueled by mountains of venture capital cash), the AI companies massively underpriced the tech. That can only last for so long before the pressure to switch from growth to profit becomes too great. Coding tools like Cursor and GitHub Copilot are ditching their all-you-can-eat plans and moving to usage-based consumption. And for the holdouts who haven’t scrapped the flat price model, usage limits and throttling are heavily increasing. At some point, the bill comes due. And that time is now.
You can only pick two…
So what do we in the eLearning space do with this cost reckoning? All of us, whether we’re on the product provider side (as Rustici Software is) or the instructional designer side or anywhere in between, are going to have to start making hard choices about which part of the fast-cheap-good triangle we want our products and learning to be in.
In some cases, this is readily apparent. If you’re an instructional designer creating quiz questions for your training, good and cheap is very achievable within the time it takes to create and roll out new training. But if you are building an AI tutor or embedding quiz question generation on the fly inside your content, it’s tougher. You absolutely need it to be fast to be just -in-time. Do you want it cheap or good? Is there even an option for it not to be good?
While there may not be a silver bullet for deciding which route you take, we recommend starting the conversation now with your company and teams. It might start with incorporating more vetting into the procurement process or by establishing clear long-term goals for how you want AI integrated into your learning or platform. Regardless, you don’t want to be stuck with no options when the next AI bill comes in 5x as large as last month’s. Consider what works best for your team and learners, and you’ll be on the right track.
Our approach to fit into the triangle
For our products, we are very deliberate about using AI only when we feel it fills a gap and adds value. Take Rustici Generator, for example: rather than giving an AI the full course materials to parse, we build deterministic parsers that can detect and return content without involving AI at all. And when we do use AI, whether that’s generating metadata on skills and topics or creating quiz questions, we constantly test across different AI models to ensure we’re striking the right balance between cost-effectiveness and quality.
And as we look to roll out new capabilities that leverage AI, like MCP, we recognize that how we implement things like MCP servers that your own agents tie into directly affects your costs, too. A ‘no wasted expensive tokens’ approach will be a big product feature, and we want to be at the forefront of it.
Finding your balance
Just as the introduction of AI caused some amount of chaos and rethinking in the industry, so too will come the re-valuing of AI. Expertise still matters, but increasingly the expertise around ‘when to use AI and why’ is just as important, if not more so. Unless you have an unlimited budget, “AI all the things” becomes cost-prohibitive and impractical. The most successful ones will be the people and companies who can best balance the scale between cost and value.
If you have thoughts or questions about Rustici Generator or AI in eLearning, you can always feel free to reach out and ask us anything!
