A coworking space's chatbot: day passes, memberships and meeting rooms

A coworking space sells something people decide on quickly. Someone in town for two days needs a desk tomorrow. A small team wants a meeting room on Thursday. A freelancer compares three spaces on price. They ask in the chat, and if the answer is slow or vague, they book somewhere else.

That makes coworking a good fit for an AI assistant, and the answers are mostly on the website already: plans, prices, opening hours, what’s included. The difficulty is that almost every answer depends on something: the plan, the location, the time of day, the number of people.

What people ask

Question	Depends on	Source
“How much is a day pass?”	Location; weekday or weekend	Pricing page
“Is there space tomorrow?”	Live availability	Booking system or hand-over
“What’s included in the monthly plan?”	Plan	Plans page
“Is it 24/7?”	Plan; location	Plans page, access section
“Can I book a meeting room for six?”	Room size; plan credits	Meeting rooms page
“How much is a meeting room per hour?”	Room; member or not	Meeting rooms page
“Can I use it as a business address?”	Plan; extra fee	Plans page or virtual office page
“Is coffee included? Is there parking?”	Location	Location pages
“Can I bring a guest?”	Plan; guest fee	House rules
“How do I cancel my membership?”	Plan; notice period	Terms

Where answers go wrong

The member price given to a non-member. Meeting room prices often have two columns: members and visitors. An assistant that reads the first column quotes the member rate to someone who has never been in.

24/7 access applied to every plan. “24/7 access” is usually a feature of the higher plans. Flexible-desk members often have office hours only. The assistant reads “24/7 access” on the plans page and answers yes.

One location’s facts for another. Multi-location spaces have different hours, parking and prices per site. The assistant answers from the first location page it finds.

Credits. Many plans include meeting room hours or credits. “Is the meeting room free for me?” depends on how many credits are left, which the website doesn’t know.

A made-up example

A coworking operator has two locations. Its plans page says: “Flex: €190/month, access weekdays 8:00–19:00. Fixed desk: €320/month, 24/7 access. Both include 4 hours of meeting rooms per month.” Its meeting rooms page says: “Room for 6: €25/hour for members, €40/hour for visitors.”

A visitor asks: “I’m thinking about the Flex plan. Could I come in on Sunday evenings to work, and how much would a room for six cost me for two hours?”

The assistant replies: “Yes, members have 24/7 access. A room for six costs €25 per hour, so €50 for two hours, and your plan includes 4 free hours, so it would be free.”

The first part is wrong for Flex: no weekend access. The second is right only once they are a member, and free only if the credits weren’t used. The visitor signs up, comes on a Sunday and can’t get in.

The fix: write each plan as one block with every condition in it: “Flex (€190/month): weekdays 8:00–19:00 only, no weekend access; 4 meeting room hours a month.” Label prices clearly as member or visitor rates, and give the assistant an instruction to ask which plan and location the person means before answering access or price questions.

Setting up the assistant

1. One block per plan, with price, access hours, what’s included and the notice period.
2. One page per location, with hours, address, parking and anything that differs.
3. Member and visitor prices on separate lines, each labelled.
4. A rule for availability: the assistant doesn’t guess whether there’s space; it links to the booking page or hands over.
5. A rule for credits: “Your plan includes 4 meeting room hours a month; you can check what’s left in the app.”

Testing it

Ask every question in the table twice: once as a visitor, once as a member of the cheapest plan. Then ask three questions about the second location. Most mistakes appear in the second and third rounds, because the first answer the assistant finds is usually about the top plan and the main location.

The same pattern, a price or rule that depends on plan and place, shows up in gyms, hotels and language schools. For the reason behind it, see how an AI assistant answers; for the cost of getting it wrong, our worked example.

If you run a space and want an assistant set up on your own pages, Assistant Setup does that, and Assistant Care checks the answers every week. More examples on our local services page. Platform guidance on testing, such as Zendesk’s advice to test with people who didn’t build the setup, applies here too.¹ Intercom’s guide on content for Fin has useful advice on writing plan and price pages an AI can read.²

Footnotes

1. Zendesk, “Best practices for testing AI agents”, support.zendesk.com. ↩

2. Intercom Help, “Optimizing content for Fin”, intercom.com/help. ↩