Co-design beats off-the-shelf

Most Heads of L&D have run this experiment without meaning to. You licence a polished AI-training library, roll it out, and the dashboard looks healthy — enrolments up, completions climbing, satisfaction green. Then you walk the floor a quarter later and almost nothing has changed in how the work actually gets done. The course was finished. The capability was never built.

It usually isn't a content-quality problem. The off-the-shelf modules are well-produced. The problem is that they were made for everyone — which means they were made for no one in particular, and "no one in particular" is exactly who doesn't show up to do your work on Monday morning.

Generic knowledge doesn't survive contact with real work

There's a reason the polished library underdelivers, and it's older than AI. Hermann Ebbinghaus mapped the forgetting curve in the 1880s: knowledge acquired in a single sitting and never applied decays sharply within days. More than a century of workplace-learning research points the same way — the Center for Creative Leadership's widely-used 70-20-10 model holds that most durable capability is built through challenging on-the-job experience and learning from others, with formal courses accounting for only a slice.

Off-the-shelf AI training is almost pure "formal course." It teaches about prompting, about data, about the tools — in the abstract, with someone else's examples. The learner nods along, completes the module, and then turns to their own messy workflow with no bridge from the tidy example to the specific task in front of them. LinkedIn's 2025 Workplace Learning Report keeps surfacing the same finding: the learning people genuinely engage with is the learning tied to their role and their goals. Generic content, by definition, isn't.

So the catalogue produces a number that looks like progress — completion — and withholds the one that matters: whether the work changed.

Co-design supplies what the catalogue can't

Co-design flips the optimisation. Instead of starting from a supplier's catalogue and asking "which modules shall we assign?", it starts from your business and asks "what should people be able to do differently — and on which real work?"

In practice that's a working method, not a procurement event:

• Discover the work. Map the real workflows and live use cases where AI would move the needle — claims, service, finance, operations — before a single module is chosen.
• Co-design the waves. ASTRA and your people design content against those workflows. Your examples, your data shapes, your guardrails — not a generic sandbox.
• Deliver on real tasks. People apply techniques to the work they brought with them, in clinics and workshops, so the learning has somewhere to land the same week.
• Let evidence reshape the next wave. What landed, what didn't, which blockers recurred — the next wave is scoped from the evidence the last one produced.

This is why co-design isn't a nicety; it's the mechanism that closes the transfer gap. The three things the forgetting curve and the 70-20-10 model say capability needs — relevance, application, and reinforcement — are precisely the three a catalogue strips out, and precisely the three co-design puts back.

It's also where a genuine partner earns their place over a content vendor. A vendor ships you the modules and measures whether you opened them. A partner sits in the room, learns your workflows, and designs against them — then changes the design when the evidence says to. One optimises for their catalogue; the other optimises for your business.

What a Head of L&D can do this quarter

• Audit for relevance, not coverage. A library with 400 AI courses isn't an asset if none of them use your data, your tools, or your scenarios. Ask what fraction maps to real workflows.
• Start from workflows, not modules. Pick two or three functions where AI clearly helps, and design backwards from the task to the training — not forwards from the catalogue to the calendar.
• Co-design with a sponsor, not just a vendor. Put a business owner in the design room. Their workflows, their priorities, and their authority to make the learning land are the difference between a course and a capability.
• Build in applied practice. Reserve time for people to do real work with the technique while support is in the room. Application is where the forgetting curve gets beaten.
• Measure adoption, not completion. Track whether the work changed — usage, applied evidence, recurring blockers — not how many certificates printed.

Off-the-shelf optimises for the supplier's catalogue; co-design optimises for your business. Both can fill a calendar. Only one reliably changes the work — because only one is built from the work in the first place.

That's how ASTRA Academy runs every engagement: co-designed with your people, against your real work, and reshaped wave by wave as the evidence comes in.