Asking a demo workspace three questions, with receipts

The demo workspace on Crossroad is a fictional booking platform for salons, gyms and clinics called Slotwise. It holds 12 sources: two customer calls, a discovery call and an RFP from a gym-chain prospect, a support chat, a support team's monthly sync, a bug ticket, an email from a customer, a set of app store reviews captured as a screenshot, a market report, an internal roadmap review and an activation memo. Out of those came 41 insights and eight themes.

Everything below is real output from the product on that sample data. I have only cleaned up formatting. You can repeat every question yourself; the demo is open without an account.

Question 1: which integrations do customers mention?

Calendar sync. Two-way calendar sync per staff member (Google Calendar, Outlook), requested by Bella Hair Studio and the prospect FitLab. [1] Practice management. Sync bookings with practice management systems to eliminate double data entry, required by NordPhysio Clinic. [2] Public API and webhooks. API and webhooks to sync bookings into the customer's own app, required by FitLab, and appearing in both their discovery call and their formal RFP. [3] Access control. Gym access control so booked members can badge in, required by FitLab. [4] Note: SSO and EU data residency are also FitLab requirements, though they are security and compliance items rather than integrations.

What I like about this one is the last line. The model was asked about integrations, noticed two adjacent requirements that a person might lump in, and flagged them instead of either padding the list or dropping them.

Question 2: do customers complain about pricing?

There are no insights about customers complaining about pricing. No mentions of pricing being too high, unexpected charges, or pricing model dissatisfaction. The closest related topics are payment flow transparency, where Studio Verde found the no-show fee setup confusing and wanted to preview charges before enabling them [1] [2], and payout timing, where Bella Hair Studio complained about weekly payouts straining cash flow [3]. Neither is a complaint about Slotwise's own pricing.

This is the answer that convinced me the approach works. A system that cannot say “no” is worthless for decisions, because you can never trust its “yes”. This one said no, then offered the nearest evidence and labelled it as nearest, not as an answer.

Question 3: which accounts mention payouts?

Bella Hair Studio is the account that mentions payouts. They raised the pain point of weekly payouts straining cash flow, requesting daily settlement or a clear payout landing date. This was mentioned three times, across a quarterly review call, a support sync and a direct email. [1] Payouts also appear as a recurring support theme in an internal observation noting that support volume concentrates on rescheduling and payouts month after month, though no specific customer account is named there. [2]

“Mentioned three times across a call, a sync and an email” is the sentence you want in a roadmap review. Nobody can argue with it, and if they try, the three sources are one click away.

What the themes found on their own

Themes are computed across all insights without anyone tagging anything. On this workspace they came out as:

• Reminders and no-shows
• Consumer booking experience
• Onboarding and activation
• Market and enterprise readiness
• Payments and payouts
• Integrations and calendar sync
• Staff scheduling and reliability
• Reporting and visibility

Each theme shows who is raising it. “Reminders and no-shows” is four customer sources plus a market report, which is a different conversation than four internal observations would be.

What a citation actually buys you

A cited answer is three things stacked:

1. The claim, in the answer.
2. The insight behind it: the type, the summary, and the exact quote.
3. The source: which call, which email, which review, on which date, from which party.

Click a number in any answer and you see all three. This is what makes the answer usable in a room where someone disagrees. You are not relaying what an AI thinks; you are pointing at what a customer said, on a date, in a document.

Limits, honestly

It only knows what was captured. The demo has 12 sources; a real workspace after three months has a few hundred, and the answers get correspondingly better. It will not find the thing you never wrote down. And it is not a survey tool, a recorder or a roadmap: it is the memory between those.

Try the three questions above on the demo, then try one of your own.