Dashboard
Period: YTD
Revenue $237,600 +25%
RevPAR $260 +25%
Dan
Concept · Autumn · pricing agent oversight
Two people. The one who catches an automated decision going wrong with a guest in front of them, and the one who later decides whether it should change a rule.
The problemThe fix happens off-platform in thirty seconds, and the software never learns it happened.
Why nowAutumn updates 47 prices a night and logs a reason for each one. Nobody reviews 47 reasons. The only workable signal is a human marking the one that was wrong.
Why meI founded and exited a multi-venue hospitality group. I have been the person who fixed it at the desk and then had to remember to say something about it.
User storyAs the person who caught it, I need to write why it was wrong while it is still in front of me, so that the person who decides is reading what actually happened rather than what I could still recall a day later.
SuccessRules change because of things that actually went wrong, instead of because someone remembered to complain.
Sven Manzanarez — an unsolicited concept, made for my application to the Product Designer role.
Your posting asks for workflows covering “how users delegate work, review outputs, maintain control, and recover when systems fail.”
Delegation you have already designed. This is the other end: what the person at the desk does when the agent gets one wrong, how that reaches the person who can act on it, and how it becomes a rule instead of a story nobody wrote down. Control is the constraint throughout — nothing here stops, pauses or reverts the agent.
Your own number, from the Dynamic Pricing page.2 It is also the argument. No hotelier reads 47 changes. The reasons are already written — what is missing is the one bit only a person can supply: that reason was wrong.
Onboarding already asks them to write their goals and their brand voice. They will type for you. You have just never asked them at the moment something broke.
Written before the screens. Where two conflict, the lower number wins.
No pause, no revert, no dialog, no approval queue. The agent keeps running. The only thing a flag costs is the twenty seconds in the next principle — and set-it-and-forget-it stays the promise.
The field opens with the flag, in place, already focused. Asking the next morning is cheaper for the person and worthless for the record: the urgency is gone, the protocol is broken, and the detail that made it a rule is the first thing to go.
Why a decision was wrong is never on a menu. A picker returns a category; the reviewer needs the sentence. You already believe this — onboarding asks for their goals, their brand voice and their hotel in prose. This is the same ask, at the moment it matters.
Not governance theatre — they hold different information. The desk knows why it was wrong. The owner knows whether it is worth a rule.
The system aggregates by shape and shows the number. It never decides when a count is enough. One occurrence can change a rule if a person says so.
Drive it.
Redrawn from their dashboard and marketing screens.3 Calendar View and Settings are inferred — I have not seen them.
I have not used Autumn. Three of your own screens are below, unretouched, next to the three surfaces I invented — so nobody has to guess which is which.3 Tap any of them to enlarge.
What I saw
Yours. Reproduced as given, credited to you.
Eight items deep in the nav — including AI Reasoning, which is a surface you already ship for explaining agent decisions. And density: real hoteliers are reading tables, not cards.
YoursThis is the one my prototype redraws, down to the greeting, the YTD strip, the bell and Emma’s name. Where the live product is dense, this is calm — and calm is the promise the concept has to keep.
YoursThe one that changed my mind. Your goals. Your brand voice. Tell us about your hotel. You already ask the operator to write in their own words — once, at setup, before anything has gone wrong.
YoursWhere the 47 and the 12 come from, and the fact that a reason is already logged per change. Drawn from memory of the page rather than shown — the premise of this concept sits here.
Cited, not shownWhat I made
Mine. Three surfaces that do not exist.
New. Opens with the flag, inside the card it belongs to, in the red you already use on the calendar. Nothing around it moves.
InventedNew. You describe a report read in two minutes; I have not seen it, so this is a proposal for where the review lands, not a redraw.
InventedNew. Settings is in both navs; I have never seen inside it. The provenance line under the rule is the part that matters.
Inferred + inventedYours — Autumn’s own screens, shown as given Inferred — exists, contents unknown to me Invented — mine, and the part worth arguing about
Swipe sideways →
Everything else I would instrument — share of flags that carry a written reason, time from flag to approval, proposals accepted — exists to move that ratio or defend that gate.
Hotel revenue management is not my domain. Four questions that would change this:
The nearest thing I have built is a lead-qualification platform in production, where the scoring rubric version is pinned to each record at capture — so a decision can still be explained six months later. Same provenance problem, smaller stakes.
About five years designing, the last three focused on product. Production front-end in Next.js and TypeScript. This page is working code, not a picture of one.