# 3 risks the AI design process poses

> Reflections from Mostly Working

Published: 2026-10-23
Canonical: https://iamtomwhite.substack.com/p/3-risks-the-ai-design-process-poses
On this site: https://www.iamtomwhite.co.uk/notes/3-risks-the-ai-design-process-poses

Yesterday was [Mostly Working’s first Summit](https://mostlyworking.co/), a fascinating morning of talks and discussion, where 100 senior designers and design leaders gathered in a room to compare notes on the state of design with AI over the last year.

![](https://substackcdn.com/image/fetch/$s_!5UAt!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fe3476611-b570-4d5d-a3e8-aa8b253f44af_3311x2207.heic)

After listening to two super interesting panels with industry experts, we broke out into small groups, each digging into questions about our general experience using AI over the past 12-18 months, the challenges we see and the potential we’re excited about. It was great to feel the shared energy and excitement in the room around the technology advancements we’re living through.

Much of the day was about AI tools helping us with craft (execution), which makes sense because they do it incredibly well, and the thing that stuck with me through the panels, our discussions and the lunchtime debrief with my good friend [Adam](https://www.linkedin.com/in/adamrmorris/?isSelfProfile=false) that followed, wasn’t doubt about the tools. It was three themes about what can go wrong when they work this well. I wrote these up straight after the event, so they’re still a bit raw.

## 1\. High fidelity too early

It’s tempting to jump to high fidelity because you can. A year ago, getting to something polished took days. Now it takes minutes or hours, so why wouldn’t you? And you can get lost in that refinement and polish.

The trouble is that polish masks whether you’re solving the right problem. It looks great and it works, so it feels correct. Critique has the same problem. If the only thing you have to anchor to is taste, a polished thing wins every time. You need something more objective to push against. As Joel from Granola put it, you need to be returning to what the jobs to be done (JTBD) are for this?, return to design principles and, I’d argue, your research.

## 2\. Attachment to the solution

You can go very deep, very fast, and spend a lot of time perfecting a prototype. The time you put in can build your attachment to your solution, and with it a defensiveness. It feels rock solid, so you stop asking whether it’s the right thing.

> Easy to get lost in the detail because it’s so gratifying. – Joel, Granola

This came up in the panels too: senior people critiquing work, and juniors challenging it, sometimes confrontationally, when someone is convinced their solution is the best one.

## 3\. Personal silos

People at the round table described designing with these tools as isolated, even lonely. The real risk is that you now have enough intelligence on tap (even with Claude) to go deeper on your own without talking to anyone, so you skip the cross-functional feedback.

What you lose is the hive mind, early critique and course correction, and finding out about constraints and limitations before you’ve worked it through to the end and fallen for it. Which leaves me with a question I haven’t answered: are the tools themselves encouraging us to silo our work and our thinking?

## What this means for juniors

Senior designers and leaders tend to know you can’t get attached, and that constraints and collaboration make the work better. I’m not sure those entering the industry now are learning that. They may be picking up the opposite habits: fidelity first, work alone, defend the output.

None of this makes me want to use the tools less. If anything it makes me want to get better at using them, and at teaching the habits that keep the work good. I don’t have a neat answer for how, yet.