Case study
Floaa
An AI that turns the goals you care about into real time on your calendar.
Role
Product Designer, sole designer
Team
Design manager, PO, PM, developers
Duration
About four weeks
Platform
iOS & Android
Status
In investment phase
The gap
Most of us can name the things that matter to us. More time with family. Getting fit. Finally learning the thing we keep putting off. Then we open our calendar and none of it is there.
Calendars are good at managing time we've already promised to someone else. They do nothing for the time we keep meaning to protect for ourselves. That gap between what we say matters and where our week actually goes is where Floaa lives.
The decision
Two ways to build the same goal.
The brief
Set a goal, then open the calendar and block out time for it yourself. Clear and buildable, but it leaves the hardest part with the person.
The proposal
Tell an assistant what you want. The AI works out how much time the goal needs and books it into your week.
Direction taken
I proposed putting the AI at the centre. Before it went anywhere near the client, we talked it through with the design manager, PO, PM and developers. Engineering pushed back on feasibility, and fairly, since it was a much bigger build than the brief. My case was about the person using it: the original flow left them doing the hard part themselves, and the whole promise of the product was that they wouldn't have to.
Engineering went away to research what it would take. We presented the idea to the client, they loved it, and what started as one chatbot screen became the way the whole product works.
You describe the goal. The AI works out the time and books it.
How much the AI is allowed to do
An assistant that asks permission for every slot rebuilds the manual work the product was meant to remove. One that does whatever it likes is one nobody will trust with their week. So I designed three layers of control.
Autonomy by default
The AI books goal time straight away, with no approval step. This is what makes the product feel effortless.
Conversation to correct it
If a slot doesn't suit you, say so in plain language. Ask it to move your family time to Sunday morning, and the AI makes the change.
Manual control as the escape hatch
You can always open the calendar and move things yourself, without going back to the AI at all.
Control moves from before the action to after it. The person stays in charge, without doing the admin.
The AI has full authority over the time it creates, and none over the time you created.
The rule that made an autonomous assistant safe to trust
The boundary
Goal blocks belong to the AI. It can place them, move them and rearrange them freely.
Your meetings, your plans, anything you added yourself or imported from Google or Apple Calendar, are off limits.
When a goal runs into one of those, the AI stops and asks whether you'd like the goal somewhere else, or would rather rearrange your own plans. It acts freely inside its own lane and asks at the edge of it.
Setting a goal
Goals are created in conversation, not in a form. But an open chat box has a well known problem: people freeze because they don't know what to type.
So the AI leads. It asks, and offers multiple choice answers the user can tap, while the whole thing still reads as a conversation.
You get the ease of a form with the feel of talking to someone.
The AI asks, you tap. No blank box.
What I'd change
Move onboarding into the conversation
Right now onboarding is its own flow, and the chat takes over once you're in. I'd bring everything the app needs to know about the person into the conversation, so the very first interaction is the AI doing its job and people feel the value before they've finished setting up.
I'd keep system steps like sign in, permissions and calendar connection as native interface, since a tap is faster than a conversation there. The rule: the conversation handles everything about you, the interface handles everything about the device.
Show why the AI chose a time
The AI books time without asking, and correcting it only works if people understand its reasoning. There was no explanation of that anywhere in the design.
I'd add one line to each goal block, like "Tuesday evening, because it's your only free stretch before the weekend." It completes the boundary rule. One tells you what the AI won't touch, the other tells you why it did what it did.
How this project ran
This was a four week engagement at Zazz. I was the only designer, working with a design manager, a PO, a PM and the development team. The work began from a fixed requirements document; the AI direction came out of the internal review before the client presentation.
Google and Apple Calendar sync was designed but went on hold over feasibility, and I left before a final decision was made. The product is still in its investment phase and hasn't shipped, so there's no usage data to show here, and I'm not going to attach numbers I can't stand behind.
Next project
Designed by Dileep Kumar M