Recorded October 20, 2026. Hosted by Joe Moore and Katrena Drake.
Here is the tension I cannot shake: AI can help us move faster, but speed is not the same thing as understanding. It is not the same thing as trust, either.
That thread ran through this week’s AI Frontier Collaborative Mastermind and followed us straight into the AfterShow with Laura Szumylo The room started with a question about students learning to think with AI instead of letting AI think for them. Laura then carried the same problem into product design: builders can generate a polished interface in minutes, but if they skip the human work underneath it, users feel the gap immediately.
The central question is simple: How do we use AI to expand human judgment without quietly outsourcing it?
What You'll Discover
Why critical thinking matters more when answers become cheap.
How educators might evaluate the process behind AI-assisted work, not only the finished assignment.
Why a fast prototype can still be a slow path to a useful product.
How observation, mid-fidelity testing and deliberate error paths expose problems before launch.
Why privacy choices and interface language are part of trust, not legal cleanup.
A practical human-centered AI checklist for consultants, builders and business owners.
Mastermind: Teach the process, not just the answer
1. Critical thinking becomes the main event
Katrena opened the October 20 Mastermind with a question submitted by Diana: if students use AI throughout their education, what should schools teach so young people learn to think with it instead of letting it do the thinking for them?
The first answer was not “better prompts.” Tyler Sewelll did argue for creative prompting as a way to push an AI toward a useful result. Chad Keller added editing and the ability to bring your own voice to a machine-generated framework. Laura emphasized research and validation. De Boone put critical thinking at the center.
Those are related skills, but they are not identical. Prompting gets an output. Editing improves it. Research tests it. Critical thinking asks whether the output should be trusted in the first place.
“What skills should schools focus on teaching so young people learn to think with AI rather than letting AI do the thinking for them?”
Katrena Drake, reading Diana’s question in the Mastermind at 00:16:15.
That question belongs in a school, but it also belongs in every business buying an AI tool. If your team cannot explain why an answer is sound, you do not have intelligence. You have convenient text.
2. The foundation and the tool are different things
Becky Brooks argued that education remains the foundation and AI should be treated as an additional skill. David Arago pushed on timing: students may need to learn research, math and productive frustration before reaching for the machine. His calculator comparison was useful because the point was not nostalgia. The point was fluency before automation.
Jim Hale brought the other side of the coin. His grandsons attend a school that forbids AI. He understands the protection, but he also worries they could enter college and work with no practical AI experience while their peers have years of it.
That is the unresolved tension. A total ban protects the learning process by removing the tool. Unrestricted access risks replacing the learning process with the tool. The interesting work is in the middle: age-appropriate access, visible process and adults who know enough to coach the use.
3. Show your work now includes the AI conversation
Andy Roustan described using AI as a tutor beside his daughter. The machine explains how a problem becomes an answer, then she does the next problems herself. That is very different from asking for the answer and pasting it in.
Chuck Boyce offered a practical extension: submit the AI conversation with the assignment. A teacher can then see the student’s questions, judgment and revisions instead of grading a polished artifact with no evidence of the thinking behind it.
For consultants and business owners, the translation is immediate. Keep the trail. Ask teams to show the prompts, source checks, edits and decisions behind AI-assisted work. The final document is less useful than the path that produced it when you are trying to assess judgment.
4. Guardrails are not the same as control
The room agreed that young people need guardrails, but it did not agree on who should build them or what version of “safe” they would encode. Jeff Valin pointed toward school-controlled AI environments that could monitor conversations and constrain outputs. That may protect students. It may also concentrate power over what they are allowed to explore.
No clean answer emerged, and that matters. Guardrails solve some risks while creating governance questions of their own. Who sets them? Can students and parents see the rules? Can teachers challenge them? What happens when safety turns into a single approved version of the truth?
A responsible system should make its boundaries legible. Hidden control is still control, even when the intention is protection.
AfterShow: Laura Szumylo on human-centered AI design
The AfterShow moved from classrooms to products, but the underlying problem stayed put. Laura Szumylo joined Joe, Katrena and Jeff to talk about the gap between software that looks finished and software that actually works for a human being.
Laura works at the intersection of user experience, business needs and technical teams. She sees the same failure repeatedly: a builder starts with features, navigation and visual polish before asking who the product is for and what problem the person is trying to solve.
5. Vibe coding does not excuse strategy
AI makes it easy to produce screens quickly. That is wonderful right up until speed creates confidence the product has not earned.
“Before you even prompt, who is this being built for? What problem is it solving?”
Laura Szumylo, AfterShow at 00:06:18–00:07:08.
Laura’s warning was sharper than “do more research.” One confusing screen can make a user doubt the whole product. People do not separate your clever architecture from the moment they cannot find the next button. They experience one system, and trust can collapse in a few seconds.
Her rule of thumb is worth keeping: structure is expensive; polish is cheap. Fixing hierarchy, flows and assumptions after development is painful. Changing color, type and surface styling is comparatively easy. A glossy interface can hide a structural mistake, but it cannot repair one.
6. Watch what people do
Interviews matter, but people are unreliable narrators of their own behavior. They describe the path they think they take. Then they click somewhere else.
Laura recommends testing before the product feels finished. Use a mid-fidelity, mostly grayscale version. Give three to five people the same task. Ask them to think aloud. Do not guide them. Watch where they pause, what they misread, and which route they invent when the “obvious” route is not obvious to them.
Use different testers at later stages so familiarity with the prototype does not mask a problem.
Then test failure on purpose. Enter the wrong information. Take the alternate path. Shrink the screen. Try to proceed without granting an optional permission. The happy path is where demos live. The wrong path is where customers decide whether your product respects them.
7. Privacy is an experience
Laura asked the cleanest ethical question of the afternoon:
“Are we asking for this information because we need it, or are we asking it because AI made it easier to do so?”
Laura Szumylo, AfterShow at 00:12:48.
That question should sit beside every form field. If you need the information, explain why. If a permission is optional, let the user decline without breaking the rest of the product. Do not bury the reason in policy language and call it transparency.
Laura shared an example of a personal-training form that asked for far more information than the service needed because marketing wanted the data. That is not a technical failure. It is a trust decision disguised as onboarding.
The same applies to button language and accessibility. “Submit” tells the user what the software will do. “Send my request” tells the person what will happen next. Two extra words can remove doubt. Clear hierarchy, readable type, sufficient contrast and cues beyond color are not finishing touches; they are evidence that somebody considered the person on the other side of the screen.
Put human-centered AI to work
These are editorial exercises drawn from the discussion, not assignments issued to every participant.
Preserve the reasoning trail. For one AI-assisted deliverable this week, save the prompts, source checks, edits and final decision. Review the path, not just the artifact.
Test the structure in gray. Remove branding from one important workflow. If people cannot understand the next step without the polish, the structure is not ready.
Run the wrong-path test. Enter bad data, refuse an optional permission and take an alternate route. Write down every moment that becomes confusing or punitive.
Cut one unnecessary question. Review a form and ask whether each field serves the user or merely feeds a future marketing idea.
Use three fresh observers. Give them the same task and stay quiet. Repeated behavior is stronger evidence than a confident opinion.
People in the conversations
Hosts: Joe Moore and Katrena Drake.
Mastermind attendees: Laura; Richard Augenti; Chad Keller; Greg; Chuck Boyce; Tyler Sewelll; David Arago; Jeff Valin; Diana | Lemon Lift VA; Christopher Lane; Greg Long; Ben; Becky Serenity; Michael Saunders; Luke A MacLeod; Dan Hansvick; Gerrett Archambault; Tarun B; Vladimir Gutenmakher; De Boone; Andy Roustan; Justin Sahota (Steward OS by Dynasty Systems Inc.); J Hale; Praveen Mayakar; Jason Almand; blakeArt®; WayneWallace.info – GoRizeIQ.com; Tylan Miller.
AfterShow participants: Laura Szumylo, Joe Moore, Katrena Drake and Jeff Valin.
Attendance does not imply that every person spoke or endorsed every view. A Zoom notetaker account and duplicate recording tracks were excluded from the human roster; private contact details were removed.
Explore the people and ideas
Meet Laura Szumylo of UX Gaps · Explore more AfterShow conversations · Browse all articles
Carry this into the next conversation
We still need to answer who sets the guardrails, how visible those rules should be, and how schools can introduce AI without erasing the productive struggle that builds competence.
Builders have a parallel set of questions. Which user behavior would disprove your current design? What data are you collecting because you truly need it? Where has speed convinced you a product is ready before a human has tried to use it?
Continue the conversation inside the AI Frontier Collaborative community. Bring one real workflow, one uncomfortable observation and one decision you are willing to change. That is where human-centered AI stops being a slogan and becomes a practice.
Register for the AI Mastermind → Register once to join future Mastermind sessions. Watch the Aftershow free on YouTube—no signup needed ↗
