1.7.0 - Jean Grey - AI Workshop Setup & New Onboarding Flow
Three things changed this month.
An assistant can now build a whole workshop for you while you watch it happen. First-run setup became five questions that actually configure your first session instead of just greeting you. And a lot of plumbing underneath both, which you will never see but which is the reason they work.
New
An assistant that sets up your workshop
Describe the workshop you need, in plain language, in a chat. It creates the workshop, the session, all five stages, the scenario for each stage, the brief and the participant invitations.
The chat sits on the left and a preview on the right fills in as it works, so you watch the workshop being made rather than waiting for a result.
It is part of a subscription. Buying a single report does not include it.
Everyone gets one free setup. It only counts once a workshop is genuinely created. If it times out, errors, or you stop it, you have not spent it, and it tells you this before it starts.
Anything that emails a person stops and asks you first. Inviting participants or adding someone to a workshop waits for your approval, and it will still be waiting if you close the tab and come back.
The assistant cannot delete anything. There is no delete or remove function in it at all, and we have a test that stops one being added by mistake later.
It has its own button: a brick in the corner of the app that blinks when idle and bounces while the assistant is thinking. It opens either the assistant or support chat.
Setting up became five questions
The old two-card role screen is gone. Instead: your role, how experienced you are with LEGO® SERIOUS PLAY®, what the workshop is for, how many people, and an optional step to invite them.
The answers do real work. Your purpose picks scenarios for each stage. Your group size suggests how many rooms you will need. Your experience level shortens the guided tours if you already know the method, and adds a link explaining it if you do not.
Invitations you enter here wait quietly until your first session exists, then send themselves.
Group size now tells you we recommend a maximum of eight participants, rather than letting you find out the hard way.
Elsewhere in the app
Anyone can open the example workshop. It used to be an admin tool. Now any signed-in user gets one: five stages of finished models, the rooms, and what each participant said. It is clearly badged so nobody mistakes it for real work.
Your photo appears in the header on workshop and session pages.
Delete confirmations have a copy button. One dialogue showed the phrase in capitals but expected lowercase, so typing exactly what you saw failed. You can now copy the exact phrase in one click.
My Designs is split into your personal designs and your workshop designs, and the workshop section explains plainly why those cards have no delete button.
There is a team page, linked from the footer.
Changed
Skipping a tutorial now counts as skipping it. No tick, no celebration, no credit at the end. The card shows a quiet "Skipped" label and an arrow to go again if you change your mind. It used to quietly re-offer skipped tutorials, which was dishonest.
The workshop tour starts on your first visit, just as the session tour did.
Your answers follow you between devices; your tutorial progress does not. Sign in on a new machine, and you get the tours again without being asked the five questions again.
Setup never advertises the assistant. We removed it from the first-run screens. Your first ten minutes should not be a sales pitch.
We corrected our own copy about "free". Two places on the site still promise to be free forever: the support card and the open-source heading on the home page. Both are fixed, and the check that was supposed to catch them has been widened so it cannot miss them again.
The help page now lists AI-assisted setup as a paid service and explains that the cost covers the AI credits and processing time.
The privacy page explains the AI processing: what gets sent, whose account it runs under, that it is not used to train anything, how long it is kept and where it goes.
The terms now cover AI features, including labelling AI-generated content and a caveat that it can be wrong.
Fixed
The assistant broke completely on launch day, twice. Both faults were invisible to our tests because our tests use a fake AI that never touches the real one. One was a rule we did not know about that rejected the limit we had set on how many people could be invited at once. The other was a database query that failed silently, making every action look like it was happening in the wrong workshop. Both are fixed, both now have tests, and we have written down a rule: try one real request against the live AI before shipping any change to it.
Our emails looked broken in Apple Mail. The cream background stopped partway down, leaving a white block under the footer.
A button in the new design dialogue looked dead, even though it was working.
Security
A flaw in the new workshop creation function could, in theory, have created a workshop owned by the wrong person when no user was signed in. Found in review, fixed before release, with a test that would catch it coming back.
The assistant can only touch the workshop it is talking about. This is enforced in code, not by politely asking the AI. We tested it by driving every one of its functions from an account with no access, including using forged details. Nothing got through.
Limits on how often the assistant can be called, per account and per network address.
