Itamar Medeiros

Preventing Agent Drift: Designing AI Systems That Stay Aligned With Human Intent What would it take for AI agents to truly support human work—capturing intent, preserving context, and staying aligned with real jobs-to-be-done?

Everyone’s excited about AI agents. But a harder question remains: what would need to be true for them to actually stay aligned with the human job-to-be-done over time? Not just at the prompt, but across decisions, context, and collaboration. Curious how others think about this.

Usability in the Age of AI Traditional usability metrics fail for AI agents. Learn what changes when users delegate instead of navigate—and what stays the same.

Time-on-task doesn't mean what you think it does for #AI agents. Neither does "task success." The #usability metrics we've relied on for 30 years need fundamental rethinking—but performance and satisfaction still don't always correlate. What changes, what stays, and what's missing? #AgenticAI

Designing Awareness in Agent-Mediated Collaboration How should AI agents support awareness in teamwork—without overwhelming it? A design standard for agent-mediated collaboration.

#AI enhances individual cognition like summarizing and synthesizing, but #teamwork requires coordinating interdependence and shared awareness.

We must decide how to design for when agents stay peripheral or they should intervene.

#AgenticAI #CSCW #Collaboration

Collaboration in the Age of AI: From groupware to Agent-Mediated Teamwork Collaboration in the Age of AI begins when agents move between people, reshaping teamwork from simple assistance to active mediation.

We’ve spent years optimizing individual productivity with #AI.
But what happens when agents start mediating how #teams coordinate, negotiate, and build #sharedunderstanding?

Are we designing #collaborative systems — or just automating #distributedwork?

Designing for Explainability: From Interrogability to Intent Alignment When agents act on our behalf, trust depends on more than answers. It depends on showing the work through explainability.

If #AI agents can interpret intent and execute #workflows autonomously, what makes them trustworthy? Accuracy isn’t enough. #Transparency isn’t enough. So what is?

This piece explores why #explainability must move from feature to infrastructure — and what that means for design.

Taste Is the New Bottleneck: Design, Strategy, and Judgment in the Age of Agents and Vibe-Coding Taste isn’t magic or “good eye” folklore. It’s a learnable decision-making skill for designers and strategists in the age of AI.

If #AI can already “do the doing”, what’s left for us? I argue that #taste —as a learnable mix of judgment, sensitivity, and ethics—is now a core decision skill in tech. How should #designers, #devs, and #ProducManagers train it and use it to steer agents, not just ship outputs? #aesthetics

For Inspiration: “Canção da América” by Milton Nascimento Composed in 1979 to honor a friend of Bituca, "Canção da América" remains emotional for Brazilians, recalling loved ones who have taken new paths.

In addition to 'Canção da América', Milton Nascimento and Fernando Brant are the authors of other great Brazilian songs, such as 'Travessia' and 'Maria Maria' #Music #Brazil #Inspiration

When AI Becomes a Design Teammate: Applying Human-Agent Centered Design to Our Own Work AI shouldn’t just change what designers build—but how we design. This post explores agents as collaborators inside the design process itself.

If #AI can already generate #ideas, #mockups, and #prototypes, why are so many teams still misaligned?
Maybe the real opportunity isn’t automating #design—but redesigning how we collaborate with agents. Curious where you’d draw the line.

Designing Human-Agent Interaction: Principles for Trustworthy Collaboration Agentic systems demand new interaction models: visible reasoning, preserved human control, trust via dialogue—not blind automation.

If agentic systems hide their reasoning, trust collapses.
What if interaction wasn’t about issuing commands—but about questioning, negotiating, and governing autonomy? This post argues interaction is where trust is earned. #humanagentcentereddesign #AI #UXforAI #mlUX #AIUX #AgenticAI

For Inspiration: Wagner Moura Wins Best Actor at Golden Globes Wagner Moura makes Golden Globe history as the first Brazilian Best Actor winner for The Secret Agent. A triumph for Brazilian cinema and a personal inspiration rooted in Pernambuco.

How does The Secret Agent's portrayal of Brazil's dictatorship era challenge us to design with memory and ethics today? Wagner Moura's historic Globe win—from public payphones to global stages—sparks debate. What's your take on art confronting power?

Arts in Brazil: “The Secret Agent” Sparks Oscar Buzz with Nominations Momentum "The Secret Agent" surges with Cannes wins and Globe nods, channeling Pernambuco's tense shadows into Oscar anticipation. Recife's thriller grips globally.

Discover Recife and Pernambuco's soul in The Secret Agent's "orelhão" calls and dictatorship intrigue—how does this local pulse fuel global cinema? Dive into the culture behind Wagner Moura's breakout buzz. What's Recife's untold story for you? #IMDB #Brazil #Cinema

R.I.P. Orelhão Brazil is finally saying goodbye to the “orelhão” — the big-ear payphone that once made our cities talk long before mobiles existed.

Before Brazil became a country of cheap data plans and group chats, conversations flowed through a fiberglass “big ear” on the corner. The end of the orelhões in 2026 is more than urban cleanup; it is a quiet farewell to a different idea of connection. #DesignHistory #Telecommunications

Managing Unintended Consequences: Designing Encapsulated Workflows to Anticipate AI Side Effects Well-intended decisions often create unintended consequences. Understand why product discovery rituals matter before AI decisions harden into systems.

Most #AI failures don’t come from bad intent—but from unexamined decisions. What if discovery rituals, not principles, are where Unintended Outcomes really surface? How do second-order effects sneak in? #unintendedconsequences #Ethics #Design #HACD #HumanAgentCenteredDesign

2025 Year in Review A reflection on 2025: from AI tools to consequences, from speed to responsibility, and why Human–Agent-Centered Design started to matter.

2025 wasn’t about better #AI tools—it was about the consequences we quietly locked into #workflows, #agents, and systems.

This post reflects on what changed, what we overlooked, and why responsibility became a design concern again.

#AgenticAI #Ethics #HumanAgentCenteredDesign

For Inspiration: “Querem Meu Sangue (They Harder They Come)” by Titãs featuring Jimmy Cliff A reflection on Jimmy Cliff’s passing, reggae’s global journey, and how its rhythms shaped my musical upbringing in Brazil.

Jimmy Cliff didn’t just influence a genre—he shaped how #reggae traveled, adapted, and spoke across cultures. How did reggae enter your musical life, and what did it change for you? #RIP #Music

Operationalizing AI Ethics in Design: From High-Level Principles to Encapsulated Workflows How can designers operationalize AI ethics? Turn ethical principles into workflow guardrails that make agentic systems safe, fair, and contestable.

Designers: If AI agents start deciding for users, who fights for the guardrails? Principles fail—workflows win.

Your duty: Advocate where autonomy meets risk. Who's with me?

Workflow Encapsulation in Human–Agent Centered Design: From Doing Work to Governing Outcomes H-ACD's "Workflow Encapsulation" explains how to design for intent, authority, and governance when flows collapse and agents execute tasks on our behalf.

Agents are collapsing our workflows into invisible steps. If flows move “inside” the system, what should designers actually be designing: screens, or human–agent contracts? How are you handling authority, guardrails, and trust in your agentic products today? #HumanAgentCenteredDesign #HACD #AI

Dual Evaluation: Measuring Both Human Experience and Agent Effectiveness H-ACD’s "Dual Evaluation" principle explains why great agentic experiences succeed only when both humans and agents perform well together.

#AgenticAI experiences only “work” only when both humans and systems succeed in tandem. Not just one. This post explores how H-ACD’s "Dual Evaluation" principle helps build trust, create value, and make better product decisions. #Usability #HumanAgentCenteredDesign #AI

For Inspiration: “She Bangs the Drums” by The Stone Roses Remembering Mani, the Stone Roses’ legendary bass player, whose groove defined an era and left an indelible mark on British music history.

Was Mani the heartbeat of Madchester or did his influence go even further? Let’s talk legacy, basslines, and why his impact still echoes today. Join the debate and share memories of the Stone Roses’ iconic bassist. #Music #Madchester #BassPlayers #MusicHistory

Flows in the Age of Agentic AI: Our Core UX Models No Longer Apply? » { design@tive } information design Design is shifting from mapping steps to understanding intentions. What happens to user flows when Agents perform the work instead of humans?

We’ve spent decades perfecting user flows. What happens when AI collapses the steps and users just express outcomes? Is “flow” still a useful concept in modern UX—or do we need a new mental model? #AI #AgenticAI #HumanAgentCenteredDesign #UX #ConversationDesign

www.designative.info/2025/11/20/f...

Rethinking Design Critiques in the Age of AI Prototyping » { design@tive } information design How design critiques must evolve to balance AI-driven prototying speed with vision, intent, and shared understanding.

As #AI makes concept design generation and #prototyping instantaneous, we risk confusing speed with insight. #DesignCritiques used to be about understanding intent — aligning teams around the “why” behind a solution. In the AI era, this is even more important. www.designative.info/2025/11/10/r...

Self-Portrait: Rubik Cube » { design@tive } information design A designer’s reflection (pun intended) among Rubik cubes in Budapest — where the world’s most famous puzzle was born.

In a recent trip to #Budapest, I stumbled upon Retro Cube — a shop celebrating #Hungary’s most ingenious invention, the Rubik’s Cube. www.designative.info/2025/11/03/s...

#puzzles #rubikcube #selfportrait #traveljournal

Reimagining Creativity When Agents Do the Doing Exploring how Human–Agent Centered Design (H-ACD) can preserve meaning, mastery, and creative flow of experiences in the age of AI.

As agents grow smarter, what remains distinctly human in the act of creation? This piece explores how to use Human-Agent Centered Design (HACD) principles to protect authorship, craft, and the flow that makes #creativity meaningful. #ArtificialIntelligence #AgenticAI #HACD

Becoming a Reflective Practitioner: A Field Manual for Designers and Strategists How becoming a Reflective Practitioner helps designers and strategists develop self-awareness, emotional intelligence, and foster team learning cultures.

Designers are trained to deliver, but how often do we pause to reflect? Reflection isn’t a luxury—it’s how we grow. How do you make space for reflection in your practice? #ReflectivePractice #Mindfulness #Retrospective #ReflectionInAction #ReflectionOnAction

Jobs-to-Be-Done and Intention Mapping: Translating Human Needs into Agent Actions Jobs-to-Be-Done uncovers what people need, but in an agent-mediated world, we must translate those needs into actionable intentions.

In an age of autonomous AI agents, are traditional human-centered methods of understanding user needs sufficient? In this article, I make an argument that Jobs-to-be-Done (JTBD) can help create intention-based designs deliver Human-Agent Centered Experiences.

Human–Agent Centered Design: Why Our Profession Must Evolve with AI Artificial intelligence transforms design, and Human–Agent Centered Design (H-ACD) keeps humans central while involving agents as trusted mediators.

AI is everywhere—in our feeds, tools, and workflows. But does design need a new paradigm to keep up? From User-Centered Design to Human–Agent Centered Design, let’s debate how agents change the way we design for (and with) humans. #HACD #AI #ArtificialIntelligence #AgenticAI #UX

Using Jobs To Be Done to Elevate your Design and Research Practice Elevate your design and research practice: use JTBD to uncover user outcomes, win buy-in, and guide product strategy with clarity.

Jobs-to-be-Done isn’t just theory—it’s a toolkit for influence. Elevate your Design and Research practice by uncovering user outcomes, creating strategic clarity, and win buy-in so your research shapes real product decisions.

“Using Jobs to be Done to Elevate your Design and Research Practice” workshop at Hatch 2025 Reflections on Hatch Conference 2025: from JTBD insights to debates on process and the surprising sense of community in Berlin.

At Hatch Conference 2025, I led a workshop on #JobsToBeDone, joined debates on #designprocesses, listened to the chatter on #AI and found a rare sense of community in #Berlin. Here are my reflections.

What I’ve learned living 20 years overseas: Healthcare Lessons from 20 years overseas on how Brazil, China, and Germany shape public health and access to care.

From universal coverage in Brazil to efficiency in Germany and rapid expansion in China—what 20 years abroad taught me about #healthcare access and public health priorities. #AIDS #Cancer #Obesity #Hospitals #Geriatrics

Book Review: “Made to Stick” by Chip Heath & Dan Heath "Made to Stick" is a playbook for turning complex ideas into clear, compelling narratives. A must-read for design strategists and product leaders.

Great strategy isn’t just built—it’s told. Made to Stick teaches you how to make your ideas clear, memorable, and persuasive. #DesignStrategy #Storytelling #DesignManagement