This Week’s Updates: Agent-ready products, career traps, ideas worth testing, Fin’s pricing, human oversight, design craft and measuring AI’s impact on delivery.
2026 Is Becoming The Year Of The Agents
Agents have been around in coding for much of this year, but we’re now seeing the same idea move into the rest of our digital lives. OpenAI launched Dots this week to take on Meta’s Muse, while Grok Bot and Instinct are also competing in the emerging personal and work-agent space. Instinct’s valuation reportedly jumped to $10bn this week, showing just how much money is chasing the category.
So what does it mean for product teams? Well, for one, the per-seat SaaS model is in trouble because one agent can effectively replace multiple human seats. And how do you optimise for an agent that cares about price, reliability and outcomes rather than your carefully designed onboarding flow?
Amazon and Shopify illustrate the strategic choice. Amazon is blocking agents like Muse, partly because it wants control over the customer relationship, data and shopping experience. Shopify is doing the opposite, opening its infrastructure to agents because it can still earn from the transaction even when the human never visits the store.
That leaves product teams with a real strategic trade-off: do you make your product a destination that humans need to visit, or infrastructure that agents can use on their behalf? And if agents become a meaningful distribution channel, being easy for an agent to discover, understand, trust and transact with could become as important as having a great human UX.
Last week’s poll: 36.5% saw the most value in routing work and flagging when people should step in, almost matched by 34.5% choosing classification and organisation of incoming data. Another 25% had not identified a clear use case yet, while 4% chose adapting experiences for different users or agents.
This Week’s Updates
Enabling the Team
If we were wrong, somebody would tell me by Martin Eriksson
Leaders need deliberate channels for dissent, named assumption owners and pre-mortems so bad news reaches decisions before a plan fails.
10 Career Traps (Thoughtful People Fall Into) by John Cutler
Thoughtful people can become trapped by the problems they take on; recognise early warning signs and clarify ownership before helpfulness backfires.
Product
How we develop pricing and packaging at Fin by Sophie Woods
Fin combines customer research, willingness to pay and commercial modelling to choose pricing units, package value and monitor the results.
Why Good Product Ideas Don’t Survive by Itamar Gilad
Good ideas survive when teams test assumptions early and revisit evidence, instead of letting politics and delivery pressure decide their fate.
Research
AI has made my job as a researcher more social, not less by Jen Murphy
AI helps researchers recover existing knowledge and create more room for workshops, shared interpretation and direct contact with users.
Diffusion of Responsibility on Human-AI Teams: Assuming the AI Checked by Victor Yocco
Research must examine whether people actually verify AI output, rather than assuming a human in the workflow provides meaningful oversight.
Design
On making room for craft by Karl Koch
Make craft part of the plan by estimating complete interactions, assigning finishing ownership and cutting scope before cutting essential details.
Delegative UX: designing the agent experience by Marie-Claire Dean
Delegating to agents requires shared understanding of intent, boundaries and consequences, including effects on people who never used the agent.
Dev
The Missing Architecture for Agentic Software Development by Purna Doddapaneni, Chris McLaughlin, Stanley Liu and Bill Radzevych
Engineering leaders need explicit context, automated quality checks and ownership of the delivery system as AI moves bottlenecks beyond coding.
Measuring AI’s Impact on Delivery Speed by James Shore
Measure AI against team delivery over time rather than counting code or pull requests, while accounting for changing work and measurement limits.
Seen an interesting article online? Share it with us, and we might feature it in our next issue! Click here to share an article
The Full UXDX San Francisco Agenda Is Live
Which AI Conversation Does Your Team Need Most?
AI is showing up in every part of product development, but not every team is facing the same problem. Some are trying to work out how roles should change. Others are focused on trust, research quality, agent safety, design systems, measurement or turning experiments into business impact.
The UXDX San Francisco agenda is now completely live, so you can choose the sessions most relevant to where your team is right now. Explore the talks, panels and workshops for 5 and 6 November, and use NEWSLETTERSF26 for 10% off tickets.
Free Community Events
Global Product Week is happening this week. Join local product, UX, design and engineering communities to compare how work is changing and meet people facing similar challenges. Find your local agenda and reserve a place.
UPCOMING IN-PERSON EVENTS
6 Oct: Amsterdam
7 Oct: Austin
7 Oct: Oxford
8 Oct: Boise
8 Oct: Dublin
8 Oct: Lisbon
8 Oct: San Francisco
17 Oct: Washington, DC
27 Oct: Munich
Want a UXDX Community event in your city? Apply to become an ambassador
Video Of The Week
Solving the Design System Problem When Products Live for Decades
A shared design language is difficult enough when teams ship every week. At Siemens, David Sward has to make it work across products that can stay in use for decades.
David explains how Siemens brought specialised design systems together, supported different technology stacks and planned for acquisitions and long-lived products.
Watch the full session for a practical look at building consistency without forcing every team into the same implementation.
Solving the Design System Problem When Products Live for Decades
Want to go into how careers and leadership are shifting as AI compresses the ladder? My ebook Managing Your Career In The Age Of AI explores how to build judgement, relationships and influence in a world that keeps trying to automate the surface of the work.