HK co-create workshop · plan for review

20 colleagues.
One day.
ai.com shaped by the people who'll use it.

Mon 5 or Tue 6 OctHK office
For Matthew Chan (CTO) and Aaron Kurzak (CPO) · Jeet, with SunnyDraft · 28 Sep 2026

Why

Before we commit the roadmap, we want to see ai.com through users' eyes

What we have today
  • Our own view of what users need
  • Feedback from alpha users who got past setup
  • Few chances to see new users meet it for the first time
What the workshop gives us
  • We watch where new users get stuck
  • We test it on their real work, with their own data
  • They shape what we build next, and what matters most

The goal isn't to validate our plan. It's to find out what we've got wrong, while changing course is still cheap.

HK workshop · plan

What we need from you

Four things, about 35 minutes of your time

Sign off

the plan in this deck

Back 20 ICs

released for one day, through HR and their managers

Open the day

5 minutes, then leave, so people speak freely

30 min on Day +2

to hear what we learned

HK workshop · plan

Who's in the room

Three kinds of user, so we can compare, not just collect opinions

5 · New

Never used ai.com. They first see it on the day.

→ Where does first use break?

5 · Set up early

Never used it. They connect their work tools this week.

→ Does their own data change first impressions?

10 · Alpha users

From the alpha user list, including some who stopped.

→ What keeps people using it, and what made them stop?

ICs only, no managersHalf from outside engineeringSmall teams who share work
HK workshop · plan

The day · 09:30–16:00

Morning: they test it. Afternoon: they reshape it.

Test · morningCreate · afternoon
1

First time

New users try it cold while we watch

2

Ask about work

Real questions, every answer rated

3

Their ideal ai.com

Sketch the screen they'd open every morning

optional: react to early concepts
4

Reimagine work

Their ideas, from their own pain points

5

Vote

A fixed budget of tokens forces real priorities

least influenced by usmost exposed to our ideas

Each session runs before anything that could bias it. People try the product, and have their own ideas, before seeing anything of ours.

HK workshop · plan · full run of show in the appendix

How we keep it honest

Evidence we can trust, not what we hope to hear

Watch, don't just ask

We count what people do: where they stall, what they come back to. What they say is secondary.

Their data, not a demo

They ask about their own work, so we learn whether answers are right, not just whether they look good.

Their ideas first

No product pitch. Every round starts with silent writing, so quiet voices count, and our ideas only appear at the end.

HK workshop · plan

What you get · Day +2

Six answers, backed by evidence, in 48 hours

01Needs

What slows ICs down at work

02First run

Where new users get stuck, and what delights them

03Daily home

What they'd want to open every morning

04Futures

Which futures they'd want tomorrow

05New ideas

What we haven't thought of

06Priorities

What wins when ideas compete for resources

Also: bugs in the tracker with owners, the same day. The readout reports what we heard, counted by group. It doesn't argue for a direction.

HK workshop · plan

Timeline

Two weeks from sign-off to evidence

Mon 28 SepSign-off

Groups assigned, invites out

today
Tue 29 – Wed 30Early setup

5 people connect their work tools

Fri 2 OctRehearsal

Accounts, data and fairness checked

Mon 5 or Tue 6Workshop

20 ICs, one day

Day +2Readout

30 min with you both

+2 weeksShare-back

Participants hear what changed

Thu 1 Oct is a public holiday, which is why sign-off today matters.

HK workshop · plan

Risks

Three risks we're managing

Tight timeline

Sign-off today. The Fri 2 rehearsal is our go / no-go.

Accounts show no data

AI Eng runs and checks all 15 accounts by Fri 2. Anyone left without data pairs with a teammate.

Alpha users skew to engineering

Pick alpha users by function, and report every finding by group.

Fallbacks for no-shows (the format works with 14), staging outages and privacy concerns are in the appendix.

HK workshop · plan

Decision

Can we go ahead?

Sign off the plan Back releasing 20 ICs for a day One of you opens the day (5 min) 30 minutes on Day +2
Detail follows in the appendix

Appendix

Session by session

For each session: what happens, what we ask, and what we come away with.

A1 · The day at a glance

A2 · Kickoff

A3 · Session 1 · First time on ai.com

A4 · Session 2 · Ask about your work

A5 · Session 3 · Their ideal ai.com

A6 · Session 4 · Your work, reimagined

A7 · Session 5 · Futures & vote

A8 · Session 5 · The cards they rate

A9 · Close, and what happens after

HK workshop · appendix

A1 · The day at a glance

The full day

TimeSessionWhereWhat happens
08:30Team setupAll roomsScreens, Wi-Fi, staging logins, Scribe test, materials. Huddle at 09:15.
09:30Arrival & breakfastMainBadges with first name, participant ID, breakout colour. Idea wall open all day.
10:00Kickoff · 15 minMain · JeetWelcome (sponsor, 5 min), "what you told us" from the survey, the rules. No product pitch.
10:151 · First time on ai.com · 45BreakoutsGroup A: 5 open goals, thinking aloud, one observer each. B and C: journey map (~20 min), then start session 2.
11:002 · Ask about your work · 45Breakouts5+ real questions each, every answer rated. A pairs with a teammate. 10 min group discussion.
11:45Break · 10 min
11:553 · Their ideal ai.com · 65BreakoutsSketch the home screen they'd open every morning. Optional: first react to two early concepts (keep / cut / steal).
13:00Lunch · 50 min. Team huddle at 13:05: anything to change for the afternoon?
13:504 · Your work, reimagined · 50Main · Jeet4 mixed tables of 5. "How might we…" prompts. Alone → pairs → table. 2–3 idea cards per table.
14:40Break · 10 min
14:505 · Futures & vote · 55Main · JeetRate six future scenarios. 10 tokens each on capability and idea cards. Home-screen sketch pitches, then dot vote.
15:45Close · 15Main · Jeet"What didn't we ask?", exit survey by QR code, what happens next.
16:00Team debrief · 30Main · JeetTop 3 from each of us, bug triage (P0 owners), check everything's collected.
HK workshop · appendix
K

A2 · 10:00 · 15 min · Main room · Jeet

Kickoff

Set the tone: we're building ai.com with them, not presenting it to them.

2Welcome6What you told us4Rules3The day
What happens
  • A sponsor welcomes everyone, then leaves
  • 3 slides of their own sign-up answers: AI tools and comfort, biggest time drains (anonymous quotes), what they'd hand to an AI agent
  • The rules, and the shape of the day

We don't pitch a vision, show a roadmap, or say who designed what.

What we say

"Today we're building it with you, not presenting it to you."

"Harsh feedback is the most useful kind. You can't break anything."

"There are no wrong answers. If something's confusing, that's on us."

"Notes use your participant ID, never your name."

What we get
  • A room that feels heard: the first thing they see is their own answers
  • Permission to be critical, set before anyone touches the product

Note: the sponsor's welcome (up to 5 min) comes out of this block; we shorten "What you told us" to fit.

HK workshop · appendix · sessions
1

A3 · 10:15 · 45 min · Two breakout rooms

First time on ai.com

Test

Watch real first use, before anyone has seen anything of ours.

3Intro355 open goals, thinking aloud, one observer each7Debrief
20"My week with ai.com" journey map25Early start on session 2: real questions, rated
What happens
  • 5 new users each work on their own laptop with their own observer, who stays quiet and logs ▲ delight, ▼ frustration, ? confusion and every assist, with the time
  • The other 15 work quietly at another table: they map their experience so far, then start asking it questions
What we ask new users

"Sign in and look around. What do you think this is for?"

"Get it set up so it could help with your work."

"Ask it something about your work."

"Send a message to someone in this room."

"Find one thing you'd come back for tomorrow."

Everyone else: "Mark the most useful moment and the most annoying one. If you stopped using it, what made you stop?"

What we get
  • 5 timed first-run logs: where each person stalled, and every time they needed help
  • Before and after "what is ai.com for?", from sign-up vs. now
  • 15 journey maps, including why alpha users stopped
First runNeeds
HK workshop · appendix · sessions
2

A4 · 11:00 · 45 min · Two breakout rooms

Ask about your work

Test

Test today's product on their real work, with their own data, not a demo.

2Intro30Ask at least 5 real questions each, rate every answer10Discuss as a group3Buffer
What happens
  • Everyone asks ai.com real questions, the kind they'd normally ask a colleague
  • New users pair with a teammate whose account has data
  • Each answer is rated on a sheet: useful? correct? showed its source? surprising?
  • The moderator walks the room; when someone reacts, "What just happened?"
What we ask

The 5 questions: two of their own · one from the table's list (built from the survey) · one they'd normally ask a colleague · one about last week

"Show us your best answer, and your worst."

"What made an answer useful, or useless?"

"What would you ask it every day?"

What we get
  • ~100 real work questions, word for word, each rated for usefulness and correctness
  • Best and worst answers, with reasons
  • "What I'd ask every day", word for word: a direct input to answer quality
NeedsFirst run
HK workshop · appendix · sessions
3

A5 · 11:55 · 65 min · Two breakout rooms

Their ideal ai.com

Test + create

See what they'd want to open every morning, in their own sketches.

2Intro20Optional: early concept A20Optional: early concept B4Alone6Share12One group sketch1
What happens
  • Core: each person sketches, shares, then the room makes one group sketch on A3 and picks someone to pitch it in the afternoon
  • Optional, if we run it: first, a quick look at two early concepts, unlabelled, 20 min each, shown in opposite order in each room. No vote on which is better.
What we ask

"Design the home screen you'd want to open every morning. Stick figures are fine."

If we run the concept review:

"What do you remember? What is it for?" (after 5 seconds)

"Where would you start your day?"

"What on this screen would you never use?"

What we get
  • 2 group home-screen sketches and ~20 individual ones
  • Patterns: what everyone puts first, and what nobody draws
  • If run: keep / cut / steal counts for each concept
Daily homeNew ideas

Without the concept review, this session is about 25 min. How to use the spare time is still to decide.

HK workshop · appendix · sessions
4

A6 · 13:50 · 50 min · Main room, 4 mixed tables of 5

Your work, reimagined

Create

Ideas from their own pain points, before they see any of ours.

5Alone, silent10Pairs15Table picks its best 215Round 2: swap prompts, add 15Buffer
What happens
  • Each table gets a "How might we…" prompt, written from the survey's top time drains
  • Silent writing first, so quiet voices count, then pairs, then the table picks its best
  • Round 2: swap prompts with the next table for fresh eyes
What we ask (example prompts)

"How might we never chase people for the same information twice?"

"How might we find out why something is the way it is, without knowing who to ask?"

"As many ideas as you can. Wild is fine, and so is small."

Read aloud: "Ideas don't have to be AI, or ai.com."

What we get
  • 12 idea cards (3 per table): the problem, the idea, who it helps, what today's tools fail at
  • The story behind each ("because last month…"), captured by note-takers
  • The cards go straight into session 5's vote
New ideasNeeds
HK workshop · appendix · sessions
5

A7 · 14:50 · 55 min · Main room, same tables

Futures & vote

Create

Force real priorities: their ideas and ours compete for the same budget.

15Rate six futures15Spend 10 tokens10Pitches10Dot vote5Buffer
What happens
  • Futures: six short, unlabelled stories of "a Tuesday" in different roles, rated alone, then discussed
  • Tokens: 10 each, spread across 12 capability cards, their own idea cards and blanks
  • Pitches: each room pitches its home-screen sketch, each table its top idea
  • Dots: 3 each for anything on the wall, at most one on their own work
What we ask

"Would I want this in my work? Tomorrow / Maybe / No, because…"

"Which would you fight for? Which would you refuse?"

"Put your tokens on what would make the biggest difference to your work."

"Why did the biggest pile get so big?"

What we get
  • 120 scenario ratings, each with a "because"
  • 200 tokens, photographed by table: a ranked list where their ideas and our capabilities compete
  • The dot vote: what the room most wants built
FuturesPriorities
HK workshop · appendix · sessions · the cards they rate are on the next slide

A8 · Session 5: the cards they rate

Six future scenarios and 12 capabilities

CardRoleThe Tuesday it describes
KCompliance analystIt chases three teams, collects sections and drafts the monthly report. Nothing goes out until approved.
PMobile developerIt explains why an old decision was made, with sources, and who knows the most.
RProduct managerNo status meeting: progress appears from the work itself, visible to everyone.
TData engineerBefore the day starts, it flags an upstream change, what it breaks and who made it.
WSupport leadThey describe a tool in a few sentences; ten minutes later the team has it.
ZProduct designerBack from leave, a 5-minute catch-up: what happened, what was decided, who to see.

Capability cards

C01 Chat with your team

C07 AI that does tasks for you

C02 Calls that take notes

C08 Automate repeated work

C03 Ask about your work

C09 Build a tool by describing it

C04 A morning briefing

C10 Meeting prep

C05 Connect your other tools

C11 Who knows what

C06 Search everything at once

C12 A self-updating team page

Built to be fair: scenarios are unlabelled, spread across functions and shuffled per table. Before Fri 2 Oct, Jeet and Aaron check that none reads as more polished, or more "ours", than the others.

HK workshop · appendix
✓

A9 · 15:45 · 15 min · Main room · Jeet

Close, and what happens after

Catch what we missed, and promise them we'll share back.

5"What didn't we ask?"5Exit survey (QR)5What happens next16:00 · 30Team debrief, bug triage
What we ask

"What should we have asked you today, and didn't?"

Exit survey, 5 min:

"How likely are you to use ai.com at work next week? Why?"

"What's the one thing we should build first?"

"I'd stop using ___ if ai.com could ___."

What we get, and when
On the day

Every P0 bug gets an owner

Day +1

Themes per question, counted by group

Day +2

~4-page readout to you both

+2 weeks

Participants hear what changed

PrioritiesFirst runNew ideas
HK workshop · appendix · sessions