Research answers questions. Designing answers hundreds of them: which word goes on the error, what’s on by default, whether to confirm or allow undo. None of that was ever a study, or the study never reaches you as a designer. AI is now doing a lot of the generating, and that judgement doesn’t transfer to a machine on its own.

Context design is how you hand it over — turning what you know about your users into constraints an agent reads at the moment it decides.

This is a working session, not a conference talk. We’ll cover what belongs in that context and what doesn’t, how MCP lets an agent read it live instead of from a stale copy, and how to tell whether it worked. Drawn from how we build at Looppanel, and from what we see customers like Swiggy and Freshworks doing with theirs — including the parts that didn’t work.

You’ll leave with something you can try this afternoon.

What you’ll walk away with:

🔍 A clear split between research that informs you and context an agent can actually use

🧩 What to put in that context — and what to leave out

✍️ How MCP lets an agent read it live, instead of from a copy that went stale last week

🚧 An honest picture of what broke when teams tried this in production

🤝 Time after the session to ask about your own product, not just the examples on stage

Who this is for: Product designers, UX designers, design engineers, researchers, and anyone who’s watched an AI tool generate a perfectly average screen and thought “that’s not how our users work.” Every experience level works here — if you already use AI in the workflow, you’ll have sharper questions; if you’re just starting, you’ll see where judgement still has to live.

Event details

📅 Saturday, 26 September 2026

🕒 9:30 AM – 12:30 PM IST

📍 Yuniq, TICEL Bio Park, Tharamani, Chennai

💸 Free to attend

Run of show

9:30 Welcome, intro to Friends of Figma, how the morning works

9:45 Context design — what belongs, what doesn’t, and how an agent reads it

11:00 Break

11:15 Questions, including the bits that didn’t work

11:45 Try-this: a constraint you can take back to your own file

12:10 Connect, snacks

12:30 Close