Your team already cracked AI. You're the thing in the way.

Two things are happening at once, and they pull in opposite directions. Regular people are folding AI into their days in ways your roadmap never planned for. Meanwhile the people you hired to build with it are either burning out or waiting on permission that never comes. Let me catch you up.

The chatbot is doing jobs you didn't ask it to do

People are not using chatbots the way product decks imagine. A YouGov survey commissioned by Mashable found Gen Z leads all groups in daily use, with 37 percent reaching for a chatbot once or more a day. They also assign it roles you would not put in a spec: 20 percent call it a friend, 23 percent use it as a tutor, and a smaller share treat it as therapist. Nearly half say they would be moderately to extremely affected if it disappeared.

The use is messier than any category. Marisol Jimenez, a 23-year-old grad student, uses Claude as tutor, sounding board, and therapist, and admits she once went in "circles around the situation" during a "dark time" leaning on a chatbot for emotional processing. Larz May, a digital wellness advocate, puts the emotional pull plainly: "Social media hacked our dopamine. But AI, in many ways, is hacking our oxytocin."

If you build products, this matters. People are bringing their whole lives to these tools, not just tasks. Design for that reality, not the tidy assistant use case.

The mess is the opportunity, not the bug

Work already has structure: calendars, trackers, inboxes. Personal life does not. That gap is where Sida Li placed her bet. She built Cue, a personal AI that lives inside iMessage, aimed at the long tail of remembering birthdays and settling group dinners. Her rule: "If it's not solving a fundamentally human need, it won't be an enduring product."

Two of her calls are worth stealing. She dropped the word "agent" because IDEO's Tim Brown told her "agents have their own agenda," and she wanted something that acts in the user's interest. And she made every memory visible, editable, and deletable, to earn trust before adding features. When she tested an enterprise version, everyone liked it but no one had a budget line for it, so she killed it. That is the desirability, viability, feasibility test doing real work.

Your best people are already good, and hiding it

Here is the number to sit with: MIT found 95 percent of corporate AI projects go nowhere, and the cause was the companies, not the tools. Dan Maccarone tells of a head of UX at a hundred-year-old manufacturer who cracked how to build with AI fast, then got stuck waiting on IT to approve the same tools her own PMs were already using without permission.

Ethan Mollick calls those PMs "secret cyborgs." Microsoft and LinkedIn's Work Trend Index found roughly three in four knowledge workers already use AI at work, most smuggling in their own tools, and more than half won't admit using it on anything that matters. McKinsey caught leaders lowballing their own teams' AI use by a factor of three. The bottleneck is not your workforce. It is the approval machinery above them.

More power, fewer excuses

Andy Budd frames the shift for design without the hype. The real change is not that designers can make more screens. It is that they need less permission. A good designer can now go from "we should fix this" to "I fixed it and shipped it." That is the bull case.

The bear case has teeth. Autonomy removes the cover. "I had a better idea but never got the engineering time" stops working when you can prototype the better idea yourself. Budd's warning for leaders: many companies cannot tell great design from plausible design. Plausible design looks coherent in a review, uses the right components, and ships because nobody objects. AI produces a lot of it. If your org already thinks design is mostly screens and polish, AI gives product and engineering a cheaper way to bypass design entirely.

When the tool starts running your team

Speed has a cost your top performers feel first. A designer who vibe-codes all day describes AI fatigue setting in when the workflow ate his free time too. His fix is worth passing along: if you already run 5x to 10x faster with AI, chasing 100x leads to burnout, not output. Invest in work quality, not quantity.

His other point cuts against the panic you hear in every all-hands. You are not falling behind by skipping the newest model. He waited out Sketch, jumped straight to Figma when it won, and arrived less tired than peers who chased every tool. Picking the winner once it is clear beats learning everything out of fear.

The deep cut

Adoption does not spread through a company-wide email or blanket Claude access. Nancy Baym's research at Microsoft found the reason rollouts stall is that the learning stays invisible. Your secret cyborgs cracked it in the background, so nobody else sees what good looks like. The fix is small and specific: take one person people already trust, give them real cover to work out loud, and let the rest watch. That is one desk at a time, not one memo.

So on Monday, do two things. Find the person on your team already building with AI and clear their approval path this week, not next quarter. And name the difference out loud between design that is coherent in a review and design that survived contact with real users, so plausible mediocrity does not ship while you are looking the other way.

Three questions for your team

1. Who is our secret cyborg, and what exactly is blocking them from working in the open? If the answer is an IT approval, why is it still sitting there?
2. When we shipped something recently, was it good design or just plausible design, and could anyone in the review actually tell the difference?
3. Where are our users bringing emotional or messy-life needs to our product that we designed only for clean tasks, and are we serving that or ignoring it?