TL;DR:
It’s 11 pm, and you’re hosting a webinar tomorrow. Your internet has just died, and you need a desk somewhere by morning.
That happened to Crispin Piney 3 weeks into his job at OfficeRnD. He found MyWorkSpot, the nearest coworking space, but nobody there would see his problem until the next day. He got lucky: Will Ballantyne, MyWorkSpot’s co-founder, got into the office early.
Crispin told that story while hosting our latest webinar, Sales, Support & Beyond: What’s Next for OfficeRnD AI Hub. If MyWorkSpot’s AI agent had been running that night, he said, it would have been a lot easier.
That’s the problem the session was about. Enquiries arrive in the evening, on weekends, and during a busy afternoon when the team is with members. The operators who answer first get the booking.
Crispin, Managing Director, UK & I at OfficeRnD, hosted on his birthday and admitted he came to AI with some skepticism. Joining him were Momchil Blaskov, Principal Product Manager at OfficeRnD, who leads AI Hub, and Will, whose 2-site operation in Maidenhead, just west of London, has been open for about 7 years.
Momchil described OfficeRnD Flex today as a system of record. Your team goes in, creates a member, issues an invoice, and the software stores what happened.
The plan is a set of agents that act inside the platform across the member journey: bookings, memberships, renewals. They’ll run day and night, and you’ll approve what they do. Momchil said the transition has started and will speed up in Q4 and early next year.
The first agent is already live.
The AI Sales Agent picks up form, email, and web chat enquiries around the clock. It answers from your own inventory, plans, locations, and FAQs. It asks about budget and team size, then points the prospect to a tour or a direct purchase, such as a day pass or virtual office.
Recent releases let it suggest another location when one is full and make bookings for prospects. Tour booking is coming.
Setup got faster too. A new pre-fill step populates tone, style, guidance and your product list, so you can start testing sooner. The recording shows it in a short video.
Before the agent, every enquiry was a manual pickup. Front of house handled hot desks. A senior team member took larger enquiries, including those from property agents. Everyone wore several hats, so time spent on a low-intent lead was time taken from existing members.
Web chat was the weak spot. A chat would pop up, someone got a notification, and by the time they reached it the visitor had often left. Will said it got bad enough that they questioned keeping chat on the website at all.
MyWorkSpot had already built its own agent with a general AI tool. That work helped, because they already had the data OfficeRnD asked for during onboarding. Momchil noted their homemade attempt didn’t work smoothly with their system. With the agent inside OfficeRnD, it sends prospects straight to MyWorkSpot’s own booking links.
That pattern shows up in wider research too. MIT’s 2025 study, reported by Fortune, found that AI tools bought from specialized vendors succeed about 67% of the time, while internal builds succeed only a third as often. Our breakdown of building versus buying coworking software runs the numbers for flex.
Will’s biggest worry was handing customer conversations to something whose answers he couldn’t predict. MyWorkSpot prides itself on knowing the people in its space.
2 features settled it. He could test the agent in the background before putting it live. And he could choose which products it answers for, such as hot desking, while private office queries still go to the team.
When testing showed the agent elaborating on a topic, they added a rule telling it to hand those questions over.
Will estimates about 65% of their conversations are now handled all the way through by the agent, across web chat and forms. He doesn’t track conversion separately, but he sees a good share of those enquiries convert without the team stepping in.
A lot of hot desk enquiries are now no-touch. People turn up knowing what they’ve signed up for, without having spoken to anyone.
The agent also covers the evenings. Someone at home decides they need a desk tomorrow, gets an answer straight away, and is sent to book. Before, that email sat unread until morning, and Will thinks they may have lost some of those people.
It follows up with prospects who go quiet, too. His team used to note those down and chase them by hand.
Did it make the business feel less personal? When the first agent-handled customer arrived, Will went to talk to them in person. They had no idea they hadn’t been talking to someone from his team. How MyWorkSpot decided whether to tell prospects they’re talking to AI, and the name they gave the agent, is one of the best moments in the recording.
Momchil demoed the next agent live. The concierge will sit in the member portal and mobile app, and it already knows who’s asking.
In a few messages, he checked whether he could bring a guest, confirmed printing was included in his plan, and booked a meeting room for 8 people next Monday at 10 am. No calendar, no filters. He opened the calendar, and the booking was there.
Momchil said it will expand to credit allowances, invoices, and possibly contracts. The goal is for answers to match member type, so a drop-in and a monthly member see different information. The beta is starting now.
He also previewed an analytics agent, coming soon, that lets you ask questions about your revenue, occupancy, invoicing, and pricing data in plain language. He left space on the roadmap slide for more agents and asked operators which time-consuming jobs they’d want handed over next.
Every Sales Agent conversation is logged in its own tab. The agents are designed to follow your existing roles and permissions: if an admin can’t make a booking, an agent acting for them won’t be able to either. Momchil said future agent actions will appear in the audit log along with who approved them.
The gap between AI promises and results in flex is wide. OfficeRnD’s 2026 survey of 334 operators found AI use fell from 26% to 18% in a year, and the report links that drop to how poorly connected most operators’ tools are.
The ceiling is high for the operators who get it right. Gartner predicts that by 2029, agentic AI will autonomously resolve 80% of common customer service issues without human help, cutting operational costs by 30%. For where to start, read our guide to practical AI for coworking and the 2026 flex space tech stack report.
Attendees asked how long setup takes. Some operators went live in 2 weeks, and the longest so far was a 15-location operator at about 2 months.
They also asked about pricing and free trials, whether the agent works with CRMs like HubSpot, Microsoft Dynamics and monday.com, whether enquiries from property agents can go to the team, and whether you can opt out.
The answers are in the recording.