The Technical Debt Machine & Managing AI Slop

AI will build whatever you ask for, quickly and confidently. That is exactly why it is dangerous. It will not challenge the brief, spot the long-term trade-offs, or warn you when a “reasonable” choice breaks your architecture, your design system, or your product strategy.

Before AI, teams accumulated debt slowly enough that reviews and critiques could catch most of it. Now you can generate a year’s worth of technical and product debt in a single afternoon, because every session starts with missing context and the model fills the gaps with plausible guesses.

That is what creates “AI slop”: work that looks finished, but quietly violates your patterns. The fix is not better prompts. It is explicit principles that travel with the work.

Principles have always mattered, but teams got away with keeping them in people’s heads. AI cannot absorb that osmosis. If constraints are not written down, they do not exist. The teams that do well here keep a short set of guardrails for architecture, product, and design, then link out to deeper docs when needed.

The unexpected upside is that this rigour pays back immediately for humans too: faster onboarding, cleaner reviews, fewer opinion battles, and a more coherent product.

This Week’s Updates

Enabling the Team

The Technical Debt Machine: How To Manage The Risk Of AI Slop by Rory Madden
AI is the perfect yes-man. And yes-men are how organizations accumulate debt. This gap produces what we call "slop": output that looks finished and sounds plausible but contains debt that will slow down future work. 

The 5 AI Tensions Leaders Need To Navigate by Rebecca Hinds and Robert I. Sutton
Introducing AI into work exposes five core tensions – experts vs novices, centralised vs decentralised, steep vs flat hierarchies, fast vs slow, and top down vs bottom up change – so effective leaders treat these as design constraints to balance rather than problems to eliminate, using them to shape org structure, incentives, and ways of working around AI.

Product Direction

Governance By Principle, Not By Template by John Cutler
Treating governance as principle-driven risk management instead of project templates means shifting from faux precise codes and business cases to durable units like products, platforms, and capabilities, so investment oversight actually matches how value is created. 

The AI Bubble Isn’t Bursting — It’s Diffusing by Ian Batterbee
Treating AI not as a bursting bubble but as a diffusing technology reframes the moment as a shakeout, where hype gives way to slow, uneven integration into workflows, infrastructure and roles.

Continuous Research

Mastering The Art Of Research Communication | Great Question by Jesse Livingston (Sponsored Content)
Jesse on why most research fails to land, the four principles of effective research communication, and why plain language is powerful language.

The Business Is The Only Stakeholder That Matters by Josh LaMar
Treating UX and market research as rival camps misses the point; research becomes indispensable only when it clearly makes the business money, so aligning questions, methods, and storytelling to business outcomes matters more than defending disciplines or methods.

Continuous Design

Stop Chasing AI Design Demos. Start Designing Better Decisions by Allan Cardozo
Start from outcomes and kill criteria, build interaction systems that govern AI behaviour, and align resources and KPIs around value streams so strategy actually shapes what ships instead of decorating AI hype.

Silicon Clay: How AI Is Reshaping UX Design by Andrew Tipp
Five years of academic studies show UX teams already use AI most in discovery and testing, gaining speed and lower costs. However, risking generic, biased work and skill atrophy unless human judgment, critical thinking, and ownership stay at the centre.

Continuous Development

What Happens When The Coding Becomes The Least Interesting Part Of The Work by Obie Fernandez
Treating coding agents as partners, not threats, lets senior engineers offload mechanical typing and stay focused on judgment, trade-offs, and intent, turning “writing code” into the least interesting part of the job.

Why Does Development Slow? by Kent Beck
Every new feature quietly shrinks your future options, so keeping development fast means alternating between adding capability and restoring optionality through refactoring, tests, and simplifying decisions.

Seen an interesting article online? Share it with us, and we might feature it in our next issue!
Click here to share an article

EMEA VOTE NOW - For Your Favourite Talk!

Which speaker should be on stage at UXDX EMEA 2026?

You get to choose who takes the final speaking slot on the UXDX EMEA 2026 agenda in Berlin. Four amazing sessions are up for the vote:

• Angela Pesta (GitLab) on designing for “unknown unknowns” in the age of AI, and why human creativity still matters when the problem isn’t even clear yet.

• Caroline Arvidsson (Danmarks Fængsler) with a grounded reflection on the analogue, human side of design in a tech-heavy world.

• Rahul B (Amazon) on moving from opinions to outcomes, using clarity and shared language to turn debate into decisions.

• Tim Makin (CDL Software) on the AI delivery lifecycle, with real examples of where AI speeds collaboration, where it adds friction, and what teams learn the hard way.

Voting is live on LinkedIn. Click on your favourite and help decide the final slot here: https://www.linkedin.com/feed/update/urn:li:activity:7431627647711059968

FREE COMMUNITY EVENTS 

UXDX 2026 Workshop Announcements

Two new hands-on workshops just landed for UXDX 2026, built for product, design, and engineering leaders who want real progress on AI implementation (without breaking quality, trust, or delivery).

In Berlin, Rina Volovich and Haim Repael Azoulay (accessiBe) lead “AI & Accessibility in Practice: Closing the Design to Code Gap”. If accessibility keeps getting lost somewhere between Figma, handoff, and production, this workshop is built for you. You will work through an end-to-end flow using AI-powered tools (Figma plugins, code assistants, automated tests) and leave with a practical mental model for an accessibility aware pipeline across design, engineering, and deployment.

Jim Morris (Product Discovery Group) is bringing to New York “AI Prototyping for Non-Engineers: Demo + Hackathon”. This is about moving beyond basic prompting into higher leverage work: tooling, agents, prototyping, and shipping learning fast. You will get an intro to vibe coding, watch a live demo, then build your own interactive prototype in a hands-on hackathon, and share what you learned with the group.

Missed the announcements of other speakers? You can find the highlights of the speakers announced in January here.

Already have a ticket but need a place to stay? Find exclusive UXDX Hotel deals here:

Video Of The Week

From Cost Centre to Strategic Asset: Earning a Real Seat at the Table

Paul Strike (Principal: Design and Research Strategy at Reforme Design) and Falguni Desai (Managing Director, Strategy, Banking & Capital Markets at Microsoft) get specific about why design keeps being treated as execution in complex organisations, especially in banking and capital markets, and what actually shifts it. Not a rebrand. A practice: literacy, framing, and partnerships that change decisions early, before momentum makes change expensive.

They cover how to translate design into the language leaders trust (risk, cost to serve, time to serve), why “UX vs design” confusion persists when leaders only see the visual layer, and how to build allies through product and strategy roles so influence becomes durable rather than dependent on one sponsor. Watch the full talk now:

The Results of Last Week’s Poll

The question: How does your organization typically respond when product features fail to deliver expected results?

Last week’s poll asked how organisations typically respond when product features miss their expected results, and the answers show a familiar instinct: add structure rather than improve learning. A combined 51% either add more approval layers (29%) or hire more coordinators and project managers (22%). Another 26% double down and push harder on the original plan, which usually signals sunk-cost pressure and a fear of admitting the bet was wrong.

Only 23% said their organisations empower teams to iterate and learn faster. That gap matters because feature failure is rarely solved by more control. More gates and more coordination can reduce visible risk, but they also slow down feedback, stretch decision cycles, and make it harder to uncover what actually caused the miss.

The healthiest pattern is the one that treats “failed expectations” as a learning moment, not a governance problem. Empowered teams with clear outcomes, decision boundaries, and tight feedback loops can run smaller experiments, correct course earlier, and build real confidence through evidence. Without that, organisations often end up with the worst of both worlds: slower delivery and the same uncertainty, just wrapped in more process.

If you want to go into how careers are shifting as AI compresses the ladder, my ebook Managing Your Career In The Age Of AI digs into the levels of thinking and how to keep building judgment in a world that keeps trying to automate it.