The agentic organization

TL:DR: The conversation about AI and work keeps asking the same question: which jobs will disappear? It’s a reasonable thing to wonder, but it’s probably the wrong frame. What’s happening right now (quietly, inside organizations that have started deploying AI agents at scale) is way more interesting than job replacement. AI agents are dissolving the organizational logic that created those roles in the first place

What an agent actually does

An AI agent, in the current technical sense, is a system that doesn’t just answer questions but it takes action. It executes tasks, makes decisions within defined boundaries, calls other systems, loops back on its own outputs. It can run autonomously for minutes or hours before a human checks in.

What makes this different from automation as we’ve known it is the scope. Traditional automation replaced specific, repetitive steps in a defined workflow. Agents can replace the coordination layer (the emails, the handoffs, the “let me loop in the team,” the waiting etc).

NN/g’s article (thanks Pablo Fernandez Vallejo) of how service design will evolve with AI agents captures the shift precisely: AI will no longer just be a tool within a service. It will be an actor in it, mediating interactions between users and service providers, making decisions that previously required human judgment, and in some cases interacting not with people at all, but with other AI systems. So we are now designing services where the primary “user” of certain touchpoints is another AI (not a surprise in 2026).

The org chart problem

The most useful thing I’ve read on this recently is HBR’s piece on redesigning marketing organizations for the agentic age. The argument is sharp: most organizations are struggling not because they lack AI tools, but because their operating model (sequential, siloed, coordination-heavy) was designed for a world where humans needed to pass work to each other. Agents don’t need that scaffolding. So the scaffolding becomes friction.

The proposed solution is a “brand code”: a machine-readable knowledge base that encodes brand strategy, customer insights, and business rules, which both people and AI agents can act on. Specialized agents then handle content, experimentation, distribution, and reporting at scale, while humans shift toward strategic direction and judgment.

It sounds elegant. And structurally, it’s a fascinating service design challenge: you’re essentially writing the rules that govern autonomous behavior at scale, before you’ve seen all the situations those rules will encounter.

That’s not an IT problem. That’s a design problem.

The different angle HBR published after two months

In March 2026, HBR published a piece arguing that to scale AI agents successfully, you should think of them like team members: give them defined roles, clear authorities, onboarding plans, and accountability structures.

In May 2026 a different HBR piece arrived with the headline: Research: Why You Shouldn’t Treat AI Agents Like Employees. The finding: anthropomorphizing AI agents reduced individual accountability, increased unnecessary escalation, lowered review quality, and left employees more uncertain about their own roles.

Both perspectives are credible. Both are probably right in different contexts. And together, they illustrate that the field is working through a genuine conceptual problem in real time. The tension is real: you need to give agents enough structure to be useful, but if people start relating to them the way they relate to colleagues, you lose the critical distance that good oversight requires. The agent becomes a black box wearing a name badge.

The convergence trap

There’s another risk that doesn’t get enough attention: when companies deploy similar AI systems trained on similar data with similar objectives, those systems tend to arrive at identical decisions. Pricing, content, product recommendations, hiring criteria: the outputs converge.

For any single organization, each decision might look locally optimal. Across an industry, you get homogenization and occasionally, regulatory scrutiny for behavior that looks a lot like coordination, even if no one coordinated.

This is, again, a design problem. If competitive advantage is the goal (so your brand, your service, your experience is supposed to feel distinct) then deploying the same foundation models with the same configuration as your competitors is a strategy for sameness. The agents aren’t to blame. The design intent was never specified clearly enough. As AI agents increasingly mediate how users evaluate and choose services, AI-to-AI compatibility will become a new dimension of differentiation. It won’t just matter how your service feels to a human. It will matter how legible your service is to someone else’s AI making decisions on their behalf.

The question nobody is asking loudly enough

When an AI agent makes a bad call (e.g. gives wrong advice, takes a damaging action, fails in a way that affects a customer) the question of accountability surfaces fast. Who is responsible?

A paper from April 2026 on designing accountable agents argues that accountability can’t be bolted on after deployment. It has to be embedded in how the agent is designed: what it can and cannot do, how it communicates uncertainty, how it escalates, how its decisions can be audited. This is territory that service designers have been working in for decades, just with human actors. The service blueprint exists precisely to make visible the backstage decisions that affect frontline experiences. Now the backstage is populated with autonomous systems, and the blueprints need to account for them.

The opportunity here isn’t small. Service design has always been the discipline that maps the invisible: the processes, the policies, the moments of truth that users never see but always feel. In the agentic organization, that invisible layer just became orders of magnitude more complex. And the tools to think about it could be more relevant than they’ve been in years.

What this actually requires

The organizations getting this right share a few traits, from what I can tell.

They are mapping their workflows before deploying agents, not after, understanding where human judgment is genuinely irreplaceable versus where it was simply filling gaps because automation didn’t exist yet.
They are designing the decision boundaries of agents as deliberately as they design the user experience.
They are maintaining human oversight not as a checkbox, but as an active practice. People who understand what the agents are doing and why, and can intervene when the logic breaks.
And they are resisting the temptation to treat organizational redesign as a technology project.

The question isn’t whether to build agents, but whether the problem you’re solving actually warrants agentic complexity. Most don’t.

That judgment (knowing when not to automate) is itself a form of design.

QUOTE OF THE MONTH

“An agentic AI doesn’t just change what organizations can do. It changes what organizations are.”

— Andrew McAfee & Erik Brynjolfsson (The Second Machine Age)

BEST OPPORTUNITY OF THE MONTH

Global Digital CX Head at Roche

As the Digital CX Head, you are the strategic owner of defining and driving the Digital CX vision, E2E non-product specific DigCX strategy and DigCX roadmap across both B2C (patients) and B2B (clinicians, lab professionals) portfolios. You will oversee the creation, development, and validation of customer and user-centered, impactful digital customer touchpoints to deliver business value and you will be accountable to define and track quantitative behavioral analytics for our digital customer experience.

Strategic Vision & Roadmap (B2C & B2B)

Define and track implementation and impact of end-to-end Digital Customer Experience (Digital CX) roadmap changes across B2C and B2B portfolios
Create strategies that align digital experience design (including CX/UX/UI principles) with the business’s overall goals and deliver consistent brand differentiation
Define short- mid- to long-term plans to enhance the digital experience and drive meaningful impact for customers and the organization

Governance & Cross-Functional Collaboration

Ensure consistency in digital customer experiences across the end-to-end customer journey and intuitive and seamless flows with digital touchpoints and also product teams while adapting to specific local needs and requirements as necessary
Establish and chair the Digital CX committee, engage with functional teams across the division, regional and affiliate stakeholders, and Roche Data & Technology (RDT) to ensure alignment and governance
Facilitate collaboration across customer areas, networks, and functional partners to translate organizational strategies into actionable customer-focused projects

People Leadership & Team Development

Lead, mentor, and inspire the Diagnostics Digital CX team, fostering a positive and productive working environment
Coach team members to unlock their full potential, supporting their career development.
Manage resource planning and agile assignments to effectively deliver on cross-portfolio digital CX goals

Let’s build a healthier future, together.

BEST UPCOMING EVENTS

Designing with AI 2026 Rosenfeld Media’s focused online event for designers figuring out AI together — elevating effectiveness, avoiding pitfalls and keeping human-centred thinking at the core.

When: June 9–10, 2026
Where: Online

UX on the Beach 2026 An intimate international UX conference held steps from the Venetian lagoon, mixing serious design thinking with one of the most beautiful settings on the conference circuit.

When: June 9–10, 2026
Where: In-person at Venice, IT

uxday 2026 Italy’s dedicated one-day UX event in Bologna — compact, practitioner-focused and a reliable meeting point for the Italian design and research community.

When: June 11, 2026
Where: In-person at Bologna, IT

PDC 2026 — Participatory Design Conference Hosted by Politecnico di Milano under the theme “Peace, Dialogue, Coexistence”, this biennial academic conference explores how participatory and co-design can foster dialogue, give voice to marginalized communities and tackle the complexity of living together.

When: June 15–19, 2026
Where: In-person at Campus Bovisa Durando, Politecnico di Milano, IT

UX360 EU A Berlin-based conference that takes a full-circle view of UX — from research and strategy to craft and systems — across two intensive days in the German capital.

When: June 23–24, 2026
Where: In-person at Berlin, DE

UX Australia 2026 Now in its 17th edition, Australia’s leading UX conference brings together practitioners from across the Asia-Pacific region for two days of research, design and strategy.

When: August 27–28, 2026
Where: In-person at Sydney, AU

SmashingConf Freiburg 2026 One of Europe’s most beloved web and design conferences, known for its no-slides-without-demos format and unusually high signal-to-noise ratio.

When: September 7–10, 2026
Where: In-person at Freiburg, DE + Online
Tickets: from ~€200 (online), ~€495 (in-person)

Hatch Conference 2026 A Berlin conference at the intersection of design, technology and leadership — aimed at senior practitioners ready to think beyond the screen.

When: September 16–18, 2026
Where: In-person at Berlin, DE + Online

uxcon vienna 2026 Vienna’s flagship UX event, now a fixture on the European design calendar, balancing hands-on workshops with sharp conference programming across three days.

When: September 16–17, 2026
Where: In-person at Vienna, AT

Design + AI Summit 2026 CreativePro’s online summit cutting through the AI noise for creative professionals — practical sessions on Firefly, Photoshop, Illustrator and the tools actually reshaping production workflows.

When: September 17–18, 2026
Where: Online

Future Product Days 2026 One of Europe’s largest gatherings of digital product professionals, three days in Copenhagen blending thought-provoking keynotes with deep-dive workshops on the future of product.

When: September 22–24, 2026
Where: In-person at Copenhagen, DK

AI Summit Barcelona 2026 Billed as Europe’s biggest AI event with 10.000+ attendees, the Barcelona summit is where enterprise strategy meets cutting-edge AI research on the shores of the Mediterranean.

When: September 22–23, 2026
Where: In-person at WTC Barcelona, ES

SmashingConf Antwerp 2026 The autumn edition of SmashingConf moves to Belgium for another round of rigorously curated talks on design, performance and the evolving craft of building for the web.

When: October 12–15, 2026
Where: In-person at Antwerp, BE + Online
Tickets: from ~€200 (online), ~€495 (in-person)

PUSH UX 2026 Munich’s dedicated UX conference, two sharp days focused on pushing the boundaries of user experience design with a strong mix of international speakers and local community.

When: October 22–23, 2026
Where: In-person at Munich, DE

Y Oslo 2026 A Scandinavian conference exploring the future of design and technology, set in Oslo and known for its thoughtful curation and strong emphasis on design leadership.

When: October 26–27, 2026
Where: In-person at Oslo, NO

SDGC26 — Service Design Global Conference The Service Design Network’s flagship annual gathering, this year in Wiesbaden in partnership with World Design Capital 2026, under the theme “Beyond Boundaries: Enabling Solutions That Shape Tomorrow” — workshops on day one, keynotes and panels across days two and three.

When: October 28–30, 2026
Where: In-person at RheinMain CongressCenter, Wiesbaden, DE + Online
Tickets: Early bird available until May 15

Leading Design London 2026 The go-to event for design leaders — not about tools or trends but about the harder questions of influence, culture, strategy and what it means to lead design organisations today.

When: November 11–12, 2026
Where: In-person at London, UK