In Development from Capably Human

Load-Bearing

Designing AI Products That Keep the Thinking Where It Belongs

Michael D. Lapico

The second book in the series: a practical guide for product teams building AI features that keep the thinking — and the capability it builds — with the people who use them.

  • AmazonComing soon

    Hardcover · Paperback · Kindle

  • Apple BooksComing soon

    Ebook

In Development in hardcover, paperback, and ebook.

About the Book

Designing AI Products That Keep the Thinking Where It Belongs

Written for

  • Product and design leaders
  • UX practitioners
  • AI product teams

Every feature teaches the people who use it a pattern of practice, and practice is how capability is built or lost.

Load-Bearing, the second book in the Capably Human series, is written for the people who decide what that practice is: product and design leaders and the teams building AI products.

It follows Heavy Lifting, which argues that the mind now needs its own fitness revolution and introduces the rules of engagement for working with AI: what to delegate and what to keep, when to trust and when to challenge, when to let it run and when to step in. Load-Bearing is the designer’s version of those rules.

Product Design Is Practice Design

You are always designing practice. The only question is whether you are doing it intentionally.

Interface choices, defaults, prompts, constraints, feedback, and timing all decide what the user does again and again. A tool that produces finished answers trains a different pattern of use than one that asks questions, offers staged hints, or supports revision. Neither pattern is visible in a usability test. Both compound over months.

What design decides

  • Interface
  • Defaults
  • Prompts
  • Constraints
  • Feedback
  • Timing
shapesbecomesbecomescarriesinto the next taskDesignUsePracticeCapability
The practice loop: design shapes use, use becomes practice, practice becomes capability — and that capability (or its absence) is what the user carries into the next task, inside your product and beyond it.

The first decision: support or substitution

The same feature can support a capability or substitute for it. Which one it does depends on the task, the user, the stakes, and whether independent capability matters in that context.

  1. 1

    Does independent capability matter here?

    Some work exists only to be finished — automate it. Some work is how the user’s judgment, understanding, or skill is built and maintained — that effort is load-bearing.

  2. 2

    Who is this user, on this task?

    A novice and an expert need different things from the same feature. Assistance that appropriately supports an expert’s oversight can strip a novice of the practice they haven’t had yet.

  3. 3

    What is the user left able to do?

    After a hundred uses of this feature, is the user more capable of approaching the task without it, equally capable, or less?

Rule 1

Decide support versus substitution deliberately, feature by feature. Both are legitimate; only one develops capability; the worst designs are substitution that believes it is support.

From Designing for Capability (CH-G-001), Capably Human

Read the guide

Human-Strengthening AI

Eight principles, as design rules

The objective is never to maximize cognitive effort — it is to keep the meaningful work with the person while removing the rest.

  1. 1

    Attempt Before Assistance

    People are encouraged to think, respond, or attempt before receiving substantial support.

    Try this

    Default to a blank input before the generate action: ask for the user’s rough take, framing, or guess first, even one sentence of it.

  2. 2

    Questions Before Answers

    The system asks when questioning develops more understanding than an immediate answer would.

    Try this

    Respond to vague requests with one sharp clarifying question instead of a confident guess.

  3. 3

    Adaptive Friction

    The effort needed for attention, learning, and meaningful participation is preserved; unnecessary difficulty is removed.

    Try this

    Audit each feature: list what effort it removes, then classify each item as incidental or load-bearing. Automate the first list; protect the second.

  4. 4

    Progressive Support

    Assistance is calibrated to the person and reduced as capability develops.

    Try this

    Decay the scaffolding: as a user succeeds without hints, stop offering them proactively; keep them reachable.

  5. 5

    Visible Uncertainty

    Established information is distinguished from interpretation, assumption, and speculation.

    Try this

    Visually distinguish sourced fact, inference, and speculation in generated output — three registers, not one.

  6. 6

    Reflection After Action

    People are prompted to identify what they learned and what they would do differently.

    Try this

    Add a lightweight reflection beat at natural completion points: one question, skippable, never a form.

  7. 7

    Appropriate Delegation

    Systems help people decide what to delegate to AI and what to keep as human responsibility.

    Try this

    Make delegation a visible, revocable choice rather than a silent default — show what the system did on the user’s behalf.

  8. 8

    Capability Over Dependency

    The system creates value without requiring permanent dependence on it; graduation is a success state.

    Try this

    Define a graduation state per feature: what would this user doing it unassisted look like? Instrument it and celebrate it in the UI.

From Designing for Capability (CH-G-001), Capably Human

Every technique, in the free guide

The Domain Lens

A design review in five passes

Walk your feature through each domain and ask what it asks the user to exercise, and what it performs for them.

What’s at risk

Framing, reasoning, evidence evaluation, judgment — the domain most directly at risk when the product supplies conclusions.

The design tell

Users paste a situation in and adopt whatever framing and answer come back.

The counter-move

Ask for the user’s framing first; show reasoning chains and assumptions; make evidence inspectable.

  • The weakest necessary sub-skill sets the ceiling

    A feature that develops five capabilities while silently eroding the one that binds them has not helped.

  • Healthy AI use is an outcome, not a feature

    It emerges from strength across the domains plus deliberate delegation judgment. You cannot bolt on an “AI literacy” tooltip and call the job done.

From Designing for Capability (CH-G-001), Capably Human

The Capability Design Review

Ten questions to run against any feature that participates in users’ thinking

They work in critique, in a PRD template, or as a pre-ship gate.

  1. What does this feature ask the user to do repeatedly — and what capability does that pattern build or erode?

  2. Does independent capability matter in this context, for this user? If not, have we simply designed the cleanest possible automation? If so, which effort is load-bearing?

  3. Does the user attempt before the system assists? What anchors the interaction — their thinking or our output?

  4. Where the goal is understanding, does the system ever ask instead of answer?

  5. Which friction have we removed, and was any of it the practice? Is there a one-keystroke exit from exactly the difficulty that matters?

  6. Does support decrease as the user’s capability grows? What does this feature look like for a user at Prompted, and at Independent?

  7. Can the user tell fact from inference from speculation in our output? Are the load-bearing assumptions visible?

  8. Is there any moment where the user reflects, summarizes, or owns the reasoning — or does output arrival end the interaction?

  9. Are delegation decisions visible, revocable, and weighted by stakes rather than convenience? What guards against oversight decay?

  10. What would graduation look like for this feature, and would we recognize it as success in our metrics?

If a feature fails several of these and independent capability matters in its context, you are shipping substitution dressed as support.

From Designing for Capability (CH-G-001), Capably Human

Early Readers

Read Load-Bearing before it’s published

Reviewers, journalists, booksellers, librarians, and educators can request an advance reading copy.

Request an advance copy
Portrait of Michael D. Lapico

About the Author

Michael D. Lapico

For more than thirty years, Michael D. Lapico built technology meant to make things effortless. As a founder, Chief Product Officer, and design executive, he brought web and mobile products to market, from consumer apps to enterprise platforms, and built the first UX organization at an established enterprise technology company. Then AI made thinking itself optional. Now he is turning human-centered design on a harder question: how to build technology that makes people more capable, not less. He is the Founder and CEO of Capably Human, which he founded to pursue that question. He lives in Windsor, Ontario.

The Series

The Capably Human series

Load-Bearing is the second book in the series. It follows Heavy Lifting and carries that book’s rules of engagement for working with AI to the teams who design AI products, as three questions to design every assisted moment around: who does this part of the work, how is the output checked, and who is in control right now.

Press Kit

For press, booksellers, and event hosts

Cover art, the author’s photo and bios, and the book’s details, free to use in coverage of the book.

Downloads

Author bios

  • Short bio (about 45 words)

    Michael D. Lapico is the Founder and CEO of Capably Human. A product design leader since 1996, he is the author of The Capability Fitness Framework and Heavy Lifting: Capability Fitness in the Age of Thinking Machines.

  • Author bio (about 95 words)

    For more than thirty years, Michael D. Lapico built technology meant to make things effortless. As a founder, Chief Product Officer, and design executive, he brought web and mobile products to market, from consumer apps to enterprise platforms, and built the first UX organization at an established enterprise technology company. Then AI made thinking itself optional. Now he is turning human-centered design on a harder question: how to build technology that makes people more capable, not less. He is the Founder and CEO of Capably Human, which he founded to pursue that question. He lives in Windsor, Ontario.

Contact

Talks, workshops, and team orders

Write to the Capably Human team about the book.

  • Talks & workshops

    Sessions with Michael D. Lapico on designing AI products that strengthen the people who use them, for product, design, and engineering teams.

  • Team & bulk orders

    Copies for product teams, design organizations, and courses.

  • Media & interviews

    Interview and podcast requests, and review copies for journalists and reviewers.

Contact Form

* indicates a required field

Launch Updates

Be the first to know when Load-Bearing is published

Leave your email and we’ll let you know when the book is available to pre-order and buy.

Launch Updates

We’ll only use your email to send news about Load-Bearing from Capably Human, and you can unsubscribe at any time. See our Privacy Policy.