There’s a version of the artificial intelligence (AI) conversation in learning and development (L&D) that sounds like a threat: Automate everything, shrink the team, ship faster. That’s not the conversation I’m interested in having. The one I want to have is messier, more honest, and ultimately more exciting—because it’s about what happens when you stop asking AI to replace human expertise and start asking it to work alongside it.
My team has spent the last year building a lean, AI-assisted training design workflow, and what we’ve landed on is this: The right tools get us about 80 to 85 percent of the way there. The last 15 to 20 percent? That still belongs to us.
Start With a Source of Truth
One of the first things we established was a clear answer to this question: Where does authoritative content live? For us, that’s our knowledge hubs in Confluence and Notion. Before any AI tool touches a training initiative, we’re pulling from those single sources. Rather than manually digging through pages of content, we use AI to pull, synthesize, and fact-check information directly from these sources. This ensures that what we build is not only faster to produce but also grounded in the most current and accurate information available.
This sounds simple, but it matters enormously. AI is only as trustworthy as what you feed it. When we ground our process in curated, human-maintained knowledge bases—by anchoring it to trusted internal systems—we reduce the risk of hallucinations and increase confidence in the output.
The discipline of maintaining those hubs is entirely human work. AI accelerates the output stage; your subject matter experts and instructional designers own the foundation.
A Small Stack that Does a Lot
We’ve resisted the temptation to adopt every shiny new tool that comes along. Instead, we’ve built around a small handful of platforms that have become genuinely integral to how we work.
For initial design, we lean on Tangelo.ai’s AI content curator—but not to hand off thinking. Depending on the project, we’ll either do all of our training needs analysis (TNA) and initial design work ourselves, or we’ll use Tangelo to help design the initial flow and kick-start the structure. When the outline is where we want it, we move into execution.
That execution phase is where Claude has quickly become our favorite partner. It’s exceptional at taking a well-formed outline and helping us bring ideas to life—sharpening language, generating content variations, and thinking through learner scenarios. The key phrase there is “well-formed outline.” Claude responds to the quality of the thinking we bring to it. Garbage in, garbage out still applies.
For video, we use Synthesia; the incorporation of avatars has been a genuinely fun creative element. It gives our content a polished, modern feel without the overhead of on-camera production. It’s not for every project, but when it fits, learners notice.
Tangelo also operates as our learning management system (LMS), where we build out pathways, AI role-plays, knowledge checks, and pushes almost all of our content. We’re intentional about this: The era of the sprawling, over-engineered LMS isn’t right for where we are. We need solutions that meet our people where they are—intuitive, accessible, fast to update. For immediate field-level support, we use Spekit, an in-app experience that puts learning directly inside Salesforce, exactly when reps need it. No navigation, no search, no friction.
The 80 Percent Rule—And Why the Last 20 Percent Is Everything
Here’s the honest truth about working with AI in training design: It will get you most of the way there remarkably fast. Structure, language, variation, ideation—all of it is accelerated. But the outcomes your stakeholders care about don’t live in the 80 percent. They live in the judgment calls, the contextual nuances, the moments when an instructional designer looks at a module and says, “This technically answers the question, but it won’t change behavior.”
That’s still us. Human expertise isn’t being replaced—it’s being freed up to work at a higher level. When AI handles the scaffolding, your designers spend more time on strategy, on learner empathy, on the hard problem of whether training is the right intervention at all.
The framing we’ve settled on internally is this: We’re enabling AI to work with us, not for us. That distinction might sound like semantics, but it changes everything about how you adopt these tools. Working for you implies delegation and reduced oversight. Working with you implies collaboration, iteration, and a human in the loop at every meaningful decision point.
What AI Still Can’t Do
I’ll leave you with a running joke on my team: I still haven’t found the AI tool that can build a deck better than a true designer. And I say that having genuinely tried. There’s something about visual storytelling—the pacing of a slide, the instinct for what to cut, the understanding of how an audience will feel moving through a presentation—that remains stubbornly, wonderfully human.
That’s not a complaint. It’s a reminder of why the conversation about AI in L&D should never be about whether we’re needed. It should be about how we show up differently—more strategically, more creatively, more focused on the outcomes that only humans can truly drive.
The Future of Training Design
As AI continues to evolve, its role in training design will only grow. But growth doesn’t mean replacement. If anything, it increases the need for skilled practitioners who can guide, interpret, and elevate what AI produces.
The future of learning and development belongs to teams who can balance speed with intention, automation with expertise, and innovation with practicality. AI can help us move faster than ever before. But it’s human insight that ensures we’re moving in the right direction. And that balance is where the real impact happens.
The tools are ready. The question is whether your team is positioned to use them with intention.
