- UX for AI
- Posts
- UX Abides: How to Create a RAG-Centered Process that Really Pulls Your AI Product Together
UX Abides: How to Create a RAG-Centered Process that Really Pulls Your AI Product Together
“That RAG really tied the room together.” — The Big Lebowski (with apologies to The Dude) This is how UX can drag that RAG into the middle of the floor, invites everyone over, and says, “Okay, folks, let’s fix this.”
In The Big Lebowski, it’s a rug.
In AI product teams, it’s RAG — Retrieval‑Augmented Generation — that holds the room (and the AI-driven product) together.
And this is how UX can drag that RAG into the middle of the floor, invites everyone over, and says, “Okay, folks, let’s fix this.”
The Scene: The Big L Shop’s Wobbly AI Search Assistant
The Big L Shop, our fictional e‑commerce player, launched an AI search assistant powered by RAG. On launch day, it wowed execs. Two weeks later:
Customers: “Your bot returns $300 kicks when I search for ‘affordable red running shoes’ – WTH?!”
Support: drowning in tickets.
Sales: losing high‑intent buyers.
Engineering: swamped with backlog and no clear repro steps.
Everyone has data, nobody has the full story. Classic silo vibe.
The Cast of Characters
UX Lead: Voice of the customer, keeps everyone laser‑focused on user outcomes.
AI / Engineering: Can tweak prompts, retrieval rules, or add docs in real time.
Customer Support Rep: Fresh pain points + phrasing users actually type.
Sales / Product: Revenue lens: “This fix saves X carts per week.”
QA (optional): Can convert every live fix into an automated test.
(Important: limit the meeting to five/six people max. If you’re not needed to fix today’s bug, catch the recording.)
Enter UX: The RAG‑Wrangler
UX speaks all disciplines – we are the ultimate collaborators and generalists with the training to ask really good questions, so it makes sense for a UX lead to pull everyone together to solve customer problems collaboratively and in near real time:
Suggest a Collaborative RAG Session: “Folks, let’s drop the jousting emails and Slack messages. Give me 30 minutes live. We’ll edit the RAG prompt and grounding docs together and test fixes on the spot.”
Schedule It: Weekly, 30 minutes, same Zoom link. Agenda is a set of problematic prompts and current output, that lives in Fig Jam so anyone can peek in.
Run It (Like a Mini User Test Meets Engineering Team Stand‑up) Starting with the representative real failed customer query from the week, we reproduce it, and then we fix it—live, and retest as a team. No Jira tickets unless we hit real code.
Interlude: The Last Week to Pre-Order Our New UX for AI Book!

Our new book, UX for AI: A Framework for Designing AI-Driven Products (Wiley) dives deep into these UX‑led AI workflows— and it’s Amazon’s #1 New Release in Data Modeling & Design.
It’s the last week to pre-order the book on Amazon: https://amzn.to/4l2ShyL
(Now back to your scheduled programming…)
Collaborative RAG Session — Your Next Blockbuster
AI-driven products are trained, not programmed. A certain amount of ambiguity and hallucination is often acceptable and actually good, for certain creative responses (how much “creativity” is ok? You can calculate using Value Matrix).
This unique property of AI-driven applications gives us the freedom to fix things faster. Much, much faster – often in minutes instead of months:
RAG Fixes ≠ Full SDLC (Software Development Life Cycle): A prompt nudge or new grounding doc can go live in minutes—no need to wait weeks for the next release train.
Everyone Sees the Same Screen: Silos crumble when Support watches Engineering update the retriever and see instant improvements.
Loop Closes in One Call: Users complain → Support shares → UX frames → Engineering fixes → QA tests → Sales cheers. Done!
For the nitty‑gritty workflow, lean on the AI‑inclusive process outlined here: The New AI‑Inclusive UX Process. It’s the perfect blueprint for rapid, cross‑functional AI work.
The Script: A Typical 30‑Minute RAG Session at The Big L Shop
00:00 — UX re-runs a failed customer query: “Search ‘affordable red running shoes’ returns $300 kicks—WTH?”
05:00 — After a quick team brainstorm, the AI Engineer adds a prompt tweak to filter the price.
12:00 — We test live: assistant now shows $60–$90 sneakers.
18:00 — Support logs the fix; QA writes an automated test.
25:00 — Sales confirms this solves their biggest churn pattern.
30:00 — End call. One high‑impact problem crushed. Everybody’s smiling.
Release notes go out the same afternoon. Conversion upticks by Friday. That’s a UX‑led win—because the RAG session tied the whole room together.
The Big Finale: UX Helps Weave the RAG That Brings the Entire AI Room Together
UX is so much more than the pretty screens! UX professionals are uniquely positioned to unite customer voice, business needs, and AI technology. When we lead the RAG session, we:
Slash fix time from weeks to hours.
Turn user complaints into teachable moments for the model.
Make cross‑functional collaboration the default, not the exception.
So grab your RAG, Dude/Dudess, and tie that AI room together!
Want the Full Playbook?

Our new book, UX for AI: A Framework for Designing AI-Driven Products (Wiley) dives deep into these UX‑led AI workflows— and it’s Amazon’s #1 New Release in Data Modeling & Design.
This is the last week to pre-order the paperback version — or get your Kindle version now: https://amzn.to/4l2ShyL
Peace,
Greg, (Who Occasionally Abides)
Reply