The absence of “desirable difficulty” or friction in AI interactions is among the most serious concerns for learning and instructional design experts today. This concern is not due to tool speed, but rather because the lack of friction bypasses or eliminates essential stages of the learning process that strengthen human cognitive and neurological foundations. When machines respond effortlessly, the natural learning pathway described in Vygotsky’s ZPD and scaffolding is disrupted. While this may seem “efficient” in the short term, over time it can undermine higher-order cognitive skills such as problem-solving, critical thinking, and creativity.

There’s no tipping point coming where things flip and the jobs are gone. The new reality is the opposite—the more we automate, the more expert human work there is to do. Here’s why: AI commoditizes the residue of human expertise—whatever can be made explicit enough to train on. That collapses the value of default model output and creates demand for what’s different. Demand for what’s different is demand for human experts, even as we approach artificial general intelligence (AGI).

Across both forms—coworker and embedded—the pattern is the same. Employee agents take over more of the stable, repeatable, well-framed layer of work. But there is a lot of work that still requires a human being in the loop. We’ve found over and over that for any kind of complex task, the best way to get great work is to have an AI and a human going back and forth in the same workspace.

In every example, the agent needs a human in order for the work to, well, work. Someone has to point it at the right thing, decide whether the output is good, catch the places where it is wrong, and turn the result into a real-life decision or process. The further away an agent gets from a human who is in charge of making sure it works well, the less well it works.

When work is abundant and looks alike everywhere, the work that doesn’t fit the pattern becomes the rare, valuable, and high-status thing(5).

This is why, in practice, AI does not eliminate expert human knowledge work. It dramatically increases the volume of work being done, and none of that work is differentiated or valuable unless a human being is involved.

What remains after automation is not a simplified role but an arbitrary residue of the most demanding, most ambiguous, and least supported work in the entire system. The human is not replaced. In other words, the human is paradoxically left with the hardest parts, and given almost no preparation for them.

Surveying nearly fifteen hundred full-time workers across industries, roles, and seniority levels, the researchers found that intensive oversight of AI tools was the single most mentally taxing form of engagement their participants described. Workers required to monitor AI agents closely reported fourteen percent more mental effort, twelve percent more mental fatigue, and nineteen percent greater information overload than those whose AI engagement was less demanding.

The final irony of automation, she wrote, is that the most successful automated systems, those with the rarest need for human intervention, are precisely the systems that require the greatest investment in human skill.

For 11 weeks, I tracked all of my AI use. One hundred sessions. I counted the tokens processed and applied publicly available numbers on per-token energy and water intensity from Epoch AI and operator-reported data from Microsoft and Google. Anyone can run this math. In those 11 weeks, I built an iOS app from scratch and wrote policy briefs on extreme heat for nonprofits I work with. I produced documentary pitch decks and drafted a 15,000-word climate fiction piece about the Colorado River collapse. I used AI every single day, often for hours at a time. Total lifecycle water footprint of all that work: about five gallons. That accounts for everything: the water used to cool the data centers, the water consumed at power plants to generate the electricity, and the water embedded in manufacturing the hardware. When an Outside editor reached out to ask me to write this story, I was on a trip to Marble Canyon, Arizona, to train raft guide companies on what is happening with the river. I drove my diesel Sprinter van from Tucson to the site, which tallied 383 miles at 20 miles per gallon of gasoline. When I ran the numbers later, the lifecycle water footprint of my fuel was around 110 gallons. One drive to the work I do on the Colorado River used more than 20 times the water of everything I did with AI in 11 weeks. That comparison stopped me cold—and I study this for a living.

Imagine recruiting managers who have never worked at the front lines, never handled customer complaints, never written up notes from consequential meetings, never grappled with the minutiae of operational work. Leadership would become abstract, detached, and dangerously naive.

Junior roles must no longer be defined by the repetitive, automatable tasks that AI can do better and faster. Instead, they should be designed to expose people to the why behind the work.

AI is only useful when paired with critical thinking. Productivity gains are meaningless if they come at the expense of professional judgment.

The default use of AI is substitution: Let the machine do the work and cut headcount. A smarter approach is to redesign workflows, so AI handles rote execution while humans focus on framing the problems, asking better questions, and building relationships.

Consider the analogy of education: If a student outsources every essay to generative AI, they bypass the intellectual struggle that produces deep learning. It is like microwaving ideas: fast, convenient, and unsatisfying. The effort, even the pain, of thinking for yourself is what builds a student’s capacity.

And then there was Claude’s desire to help. It was relentless. I pushed back constantly on anything that broke the “fourth wall,” anything that handed the learner an insight before they’d reached for it themselves, and anything that sounded like a training exercise instead of a real situation. Productive struggle is where learning happens.

AI is a design collaborator, not a design replacement. Every meaningful improvement in the rebuilt project came from my instructional design judgment, not from Claude’s defaults. The dialogue, the model interaction structure—me insisting that learners identify what worked before being told—and the debrief—me pushing back on Claude’s first version until it asked more than it told—that was all me. Claude made those things possible faster, but without the ID judgment driving the prompts, the end result would have been slicker but shallower.

The fix is an 3-step process which gives you superpowers in AI image generation: Write a visual brief — answer six questions that close the creative and pedagogical gaps before you generate a single image. Build a mood board — gather images that capture the lighting, energy, and environment of your learner’s world. Select the 3 that look like they were shot by the same photographer on the same day and upload them individually as style references. Create character anchors — your style references fix the visual world; your character references fix the people inside it. For each named character, generate a head-and-shoulders image on a neutral background, facing forward. This is your master reference. Attach it alongside your style references every time you generate a scene featuring that character — and the tool stops making casting decisions on your behalf.

A mastery model demonstrates perfect performance from the start. They know exactly what to do, execute flawlessly, and model ideal behavior. This is what most learning designers and SMEs default to: show the “right” way immediately.   A coping model struggles, verbalizes their thinking process, makes mistakes, and demonstrates the process of overcoming challenges. They eventually succeed, but only after working through realistic difficulties.

Characters in learning design don’t need movie-like backstories. They need a recognizable problem intrinsically connected to what people need to learn. The swap question is the real test: could this character or story be replaced without changing what’s learned? If yes, it’s probably decoration.

I need to use my expertise to craft a prompt that will get the most accurate result, while still recognizing the parts that need revision.