Industries / Construction and property / Real estate / Coworking
Community
This is the work of turning a room of strangers into members who know and help each other: hosting, events, introductions, and the research on whether any of it works.
7
Milestones achieved, from 2005 to 2019.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
2005
Hosts who introduce people
In early 2005 Jonathan Robinson and colleagues opened the first Hub, a members' workspace near Angel station in London. Its founders made hosting a job: a host welcomed people and connected members who should meet. Maria Glauser, the first host of Hub London, later led the hosting practice for the whole network, and nine Hubs were open by 2008.
LessonGrowth strained trust, and in 2010 the local Hubs moved to a co-owned association with one vote per Hub.
Credit
Jonathan Robinson and colleagues; Maria Glauser
2006
Coworking without a building
In March 2006 Amit Gupta and his former roommate Luke Crawford started Jelly: every other week, freelancers worked side by side in Gupta's Manhattan apartment, free of charge. By mid-2007 Wired reported similar gatherings from London to Austin to Orlando.
LessonA recurring gathering can show local demand before anyone signs a lease.
Credit
Amit Gupta; Luke Crawford
2007
Community before the lease
Alex Hillman and his business partner Geoff DiMasi worked alongside a group of freelancers in Philadelphia for months before opening Indy Hall on 1 September 2007. Deskmag credits Indy Hall with building a space "without having a budget, but by creating a community first." Hillman later wrote up coworking's five shared values, collaboration, openness, community, accessibility and sustainability, and says they "originated with CitizenSpace" in San Francisco.
LessonIn Hillman's words, "It's not a community until it has people in it.".
dangerouslyawesome.comdeskmag.comdangerouslyawesome.comdangerouslyawesome.com
Credit
Alex Hillman; Geoff DiMasi; Citizen Space
2010
Members asked what changed
The trade magazine Deskmag ran the first Global Coworking Survey and published results in December 2010. Of 661 people in 24 countries, 88% said they had better interaction with other people and 85% said they were more motivated. The survey became a series.
LessonReports from people who chose to join show satisfaction, not cause.
Credit
Deskmag
2010
The trade starts meeting
The desk marketplace Loosecubes turned meetups in Austin into "the first ever Coworking Unconference" in 2010, a meeting whose sessions attendees propose on the day. Liz Elam took it on as GCUC, which drew 250 attendees in March 2012. In Europe the first Coworking Europe conference met at the Hub Brussels in November 2010, where its organiser, Jean-Yves Huwart, "expected around 50 people but instead 150 showed up."
Credit
Loosecubes; Liz Elam; Jean-Yves Huwart
2015
Researchers test the claims
University researchers began measuring what members reported. In Harvard Business Review in 2015, Gretchen Spreitzer, Peter Bacevice and Lyndon Garrett of the University of Michigan reported that members' "thriving" averaged close to 6 on a 7-point scale. A 2016 study in Frontiers in Psychology by Cornelia Gerdenitsch, Tabea Scheel, Julia Andorfer and Christian Korunka found that support from fellow coworkers went with higher satisfaction with one's own work, and a 2019 survey of 112 coworkers by Robelski and colleagues found 92.2% rated social interaction better at the space than at home. All three were one-off surveys, and the 2016 and 2019 authors said causes could not be drawn from them and asked for studies that follow people over time.
Credit
Spreitzer, Bacevice and Garrett; Gerdenitsch, Scheel, Andorfer and Korunka; Robelski, Keller, Harth and Mache
2019
Community as pitch, then cost
WeWork's S-1 began "We are a community company" and counted "over 2,500 trained community managers". It postponed the share offering on 30 September 2019. In June 2020 it made over half of its UK community managers and community leads redundant, about 82 people by reports, and planned to spread the remaining managers across several buildings each.
LessonA community team is a running cost, and under pressure WeWork chose fewer hosts per building.
Credit
WeWork
Watch for
AI drafts community posts
Spacebring, a coworking software company whose platform it says is used by more than 500 locations, launched an AI agent called Lem as a test version in March 2026 to draft posts about events and member benefits, and says this should raise participation. Impact Hub Ljubljana said it removed "the need for manual drafting." Lem left testing as Spacebring Agent on 22 October 2026. Spacebring publishes no figures.
Achieved whenAn operator or researcher outside Spacebring publishes event attendance, or staff hours spent drafting, before and after.
Claimed by
Spacebring, March and October 2026
Scoring who is drifting away
Nexudus says its tools sort members into "high, medium, low or no engagement" and predict who is likely to cancel "within the next 30 days", with models updated monthly, so a host can reach members before they leave. A 2025 survey of 230 operators by another software seller, OfficeRnD, found 26% "already experimenting" with AI.
Achieved whenAn operator or researcher outside Nexudus publishes how accurate the predictions were.
Claimed by
Nexudus, product page read in November 2026
Open
01
Loneliness measured over time
A study that measures loneliness when people join a coworking space and again at 3 and 6 months, against similar remote workers who do not join, with the plan posted before anyone enrols (preregistration), spaces independent of the funder, and the anonymised data published.
The briefCredit · unclaimed02
Introductions with a tally
A tool that reads short, opt-in member profiles ("what I'm working on", "what I can help with", "what I need") and proposes a few introductions a week for a host to approve, judged against introductions the host picks and pairs picked at random.
The briefCredit · unclaimed
Open 01 · the brief
Loneliness measured over time
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A study that measures loneliness when people join a coworking space and again at 3 and 6 months, against similar remote workers who do not join, with the plan posted before anyone enrols (preregistration), spaces independent of the funder, and the anonymised data published.
Why it is still open
A 2023 review of 35 papers found research on coworking and wellbeing "remains limited" and mainly qualitative. The 2016 and 2019 surveys above were one-off and called for longitudinal work. The best-known loneliness figure, that 83% of members feel less lonely since joining, is members' own report from a survey whose aims "were not focused on the social side of coworking".
Built with what exists
OSF Registries for the preregistration; the three-item loneliness scale drawn from the Revised UCLA Loneliness Scale by Hughes and colleagues; LimeSurvey; R. No new AI is needed here: the obstacle is design and independence.
First step
Draft the preregistration on OSF and approach three independently owned spaces and a university ethics board. Participants give informed consent and can withdraw, the study is reviewed for ethics before recruiting, and it stays a survey, with a list of support services offered to everyone who takes part.
Done looks like
A preregistration dated before the first person enrols; at least 300 new members from at least 6 spaces in at least 2 countries and a comparison group of at least 300 remote workers; measurements at 0, 3 and 6 months; data and code posted; results published whichever way they fall.
Open 02 · the brief
Introductions with a tally
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A tool that reads short, opt-in member profiles ("what I'm working on", "what I can help with", "what I need") and proposes a few introductions a week for a host to approve, judged against introductions the host picks and pairs picked at random.
Why it is still open
Software sellers offer member directories and event tools but no member matching with published results. Nexudus offers a "CRM & member directory", and Spacebring's agent describes no introduction feature. The 2023 review found hosts' role "fundamental and highly valued by members", which makes the host the comparison to beat.
Built with what exists
A consented export of the member directory; Meta's Muse Ember to pull offers and needs out of free text (see the Note "The jobs agents repeat"); Gemini Decision 1 to score up to 1,000 candidate pairs as yes-or-no questions in one pass (see the Note "A thousand questions, one pass"). One matching agent for the whole space, not one per member, since members' attention and the host's time are shared: Anthropic Fellows simulations found one agent serving a group did better over a shared resource (see the Note "Many agents, one budget").
First step
In one space, invite 30 members to opt in; the host makes 15 pairs by hand and the tool 15, and both groups are surveyed two weeks later. Use only opt-in data, show members why they were paired, and seek ethics review if the results will be published as research.
Done looks like
A public tally over 8 weeks in at least 2 spaces of at least 200 suggested introductions, giving the share accepted, the share that led to a meeting and the share rated 4 or 5 out of 5 for usefulness two weeks later, each for the tool, the host and the random pairs.