After raising more than $22,000 on Kickstarter to make a storytelling card game called True Story, Daniel Steinbock, a designer and musician who started the project with three friends, had a simple plan for fulfilling the 750 orders. He would hire freelance designers and writers to make the game.
His plans became more interesting when a group of researchers from Stanford University’s computer and management science departments offered to build an on-demand organization for the project.
Steinbock agreed to use the beta version of the game as a test for their “flash organization” software, which automatically hires freelancers into roles, teams, and hierarchies that allow them to work together on a goal. The goal, as one of the paper’s authors, Michael Bernstein, describes it, was to build a tool that made it feel “like if anyone with an internet connection could with a click convene an organization around themselves to complete a project.”
Already many companies employ what you might describe as “on-demand hiring.” Almost all jobs created since 2005 have been in non-traditional work such as freelancing, temping, and on-call labor, and the “gig economy” has made hiring both Uber drivers and skilled professionals feel as easy as shopping on Amazon. The Stanford researchers, like many management scholars, imagine these trends will lead to organizations that are “fluidly assembled and re-assembled from globally networked labor markets.” Flash organizations are a model for how these systems might work.
Flash organizations built on “flash teams,” a similar system the same research group had developed a few years before. Those projects worked like an assembly line: When one step was completed, the system automatically hired a freelancer for the next step, on-boarded him or her, and handed off the project.
Flash teams successfully turned napkin sketches into functional web applications and recruited users to test them—all within a single day.
But the system was rigid. It could only handle projects for which there were a pre-defined set of steps. “You could assemble a car, but you couldn’t design a car,” explains Bernstein.
True Story, the card game, was based on a popular podcast that its creators started in 2012. Making it would involve conceiving a design and logo from scratch, as well as writing short poems to go with around 160 story prompts. Like the other projects the Stanford team selected to test flash organizations—making an app for EMTs and an online training portal for a large client services company—it was too complicated for flash teams.
To solve this problem, Bernstein and the research group, which also included Melissa Valentine, a Stanford management scholar, studied how organizations operate in the analog world. They decided to use the organizational structure of movie production crews and disaster response teams as their model. “[People on a movie set] might not now each other, but they’re very familiar with each other’s roles,” Bernstein says. “They know what actors and grips do and how they work with directors.”
Here’s how the “flash organizations” worked: Team leaders started by creating a blueprint organization, with roles, teams, and hierarchy. The software automatically filled those positions by pinging the job to qualified workers. The first to accept were automatically on-boarded to the organization: They were told for what tasks they were responsible, which roles in the organization depended on their work, and which roles they depended upon.
They were also given instructions on how to change the organizational structure when needed by editing the blueprint or “source code” of the organization. If someone working on a task decided they needed to hire someone with a different sort of expertise in order to complete it, for instance, they then added that role to the organization. If the change were approved through the chain of hierarchy, it was made and merged into the project, the right people were automatically hired, and the project continued.
For the True Story game, the first “flash organization” hires were 12 poets who wrote short poems for each story prompt card. They were the first 12 who responded to an automatic message alerting them about the job within a curated pool of Upwork freelancers, and their poetry ranged wildly in quality and tone. “A lot of them were just not great,” Steinbock says.
For the prompt, “Past Life” the freelancer wrote:
Formally known as or Before now
Opposite of life after death
Reincarnation can only come from this
The “Getting Hitched” prompt elicited:
Look at that rock
He said it was love
Rude awakenings
And another poet, writing on “Childhood,” submitted:
Naïve innocence
Wide eyes, bright minds
Scars, doors, deaf and mute
The researchers decided to hire a “head poet” to review and “integrate” the set. They also added teams to design the front and back of the cards, the packaging, and the logo. When those teams had finished their work, the True Story team hired a testing lead, who proposed hiring other people to test the game, submitted this new branch of the organization, and then activated the on-demand hiring of a team. Twelve teams worked on the project. All together, the system hired 29 people to work on the project, on average within 13 minutes of posting a job. They completed the project, plus an Android app added mostly to use extra time and budget, within six weeks.
Though flash organizations were a research project, some companies—including software development Gigster,website builder B12, and business outsourcing startup Konsus—have taken similar approaches to Managing teams.. Typically, they have higher barriers of entry—coding tests, for instance—than do freelance marketplaces like Upwork.
“Flash organizations” were intended to prove that complex projects could be accomplished through a crowd of online freelancers—which Bernstein believes could make work on platforms such as freelance marketplace Upwork and Amazon’s outsourcing site Mechanical Turk better for their workers. “Our goal is to bootstrap a better gig economy by creating an opportunity through growth through a complex project,” Bernstein says. “If all I can engage in is tasks, that will never happen. If I can engage in long projects that last a long time, then I think we have a prayer.”
To Steinbock, though, the process felt like it had the opposite effect. “I imagined it as a bunch of freelancers all working on a project,” he says. “In practice, it was much more like Mechanical Turk in that it was a process that treats human beings as an assembly line, micro-outsourcing model.”
He spent his time on the project filtering work from a bunch of freelancers and asking those who demonstrated some skill to continue iterating on their work. “It felt a little weird,” Steinbock says. “to go through all these many submissions from people who were these faceless, nameless, nationless submitters, and then make these quick judgments.” In the end, he wasn’t impressed with the results. “We couldn’t outsource our product vision to someone who grabbed a quick gig and didn’t have a deep understanding of the project,” he says. The intention was always to make a beta version, but Steinbock ended up hiring freelancers to remake the project entirely.
Bernstein concedes the system might work better for projects that value time efficiency and cost over quality. But the idea that the project turned human work into an assembly line was almost opposite of what the project intended. “It’s interesting to reflect that it was possible for them to lead an organization without getting to know workers closely-it feels like it indicates an area for future designs to improve upon,” he says. “It’s not necessarily the design we were going for.”