David Sacks walked into Dolores Park Café in San Francisco to talk to Lars Dalgaard, a venture capitalist, about what he should do with his life. Or rather, his money.

Sacks, who is 43 and has thick gray hair and blue, protruding eyes, made his first fortune as an early executive at PayPal, then a second as the co-founder of Yammer, a social network for businesses, which he sold to Microsoft in 2012 for $1.2 billion. He played in poker tournaments, produced the film Thank You for Smoking, and became an early investor in Uber and SpaceX. But by the fall of 2014, he was sick of jumping from hobby to hobby. He wanted in on a startup again.

Sacks and Dalgaard were business acquaintances—Dalgaard, a general partner at Andreessen Horowitz, had once tried to buy Yammer for $300 million. After the requisite industry gossip, they got down to business. Dalgaard urged Sacks to take a look at Zenefits, a new company in which Andreessen Horowitz had recently invested.

Zenefits makes online software that automates health insurance, payroll, and other essential office drudgery—kind of a human resources version of TurboTax. It’s not a sexy idea, but with 6 million small businesses in the U.S., it’s enormously useful.

The company was founded in 2013 by Parker Conrad, who realized he could streamline small businesses’ managerial needs, saving them hundreds of hours of mind-numbing paperwork—not to mention the cost of staffing an HR department—by putting everything online. Conrad was known to be a little frenzied and disorganized but fiercely intelligent.

“From an investment philosophy … we look for the magnitude of the genius, as opposed to the lack of issues,” says Andreessen’s founding partner Ben Horowitz. “And in a way, [Conrad] was like the prototype.” Conrad had no background in health insurance but quickly learned the intricacies of the business as well as any veteran. “If you’re an insurance broker,” he said at the TechCrunch Disrupt conference in 2013, “we’re going to drink your milkshake.”

As Sacks saw it, the most compelling part of Zenefits was that it gave its software away, making most of its money through brokerage commissions that insurance companies paid when a client bought one of their plans. Those commissions recurred annually: Once a business signed up, Zenefits would profit for years to come. “It’s a glaring omission that no one had created [something like Zenefits] before,” Sacks says. He and Dalgaard talked at the cafe for three hours, stopping only when the waiters started putting chairs on the tables. A few days later, Sacks met with Conrad. He joined the company in December 2014 as chief operating officer.

Zenefits was everything Silicon Valley loved wrapped up in one company. It had a visionary founder. It tackled a stodgy industry ripe for disruption. The recurring commissions gave it a steady stream of revenue from the start. And Zenefits was the first in the health insurance software space, the Uber to its future competitors’ Lyft.

The potential for greatness is what allowed Zenefits to expand from 15 employees to 1,600 in three years; raise $580 million in three fundraising rounds; and become one of Andreessen Horowitz’s biggest investments. Last year it was valued at $4.5 billion, which made it, in Valley parlance, a “unicorn” several times over.

“You’re looking for disruptive companies,” Conrad said in that same TechCrunch Disrupt speech. “We are going to mess stuff up.” He was right, although not in the way he meant.

Conrad is 36, with a round face, a hefty build, and closely cropped, reddish-blond hair usually accompanied by a day’s worth of stubble. He grew up in New York City, the son of an environmental activist and a corporate lawyer. He went to, flunked out of, reenrolled in, and ultimately graduated from Harvard. People who’ve worked with him say he’s prone to emotional highs and lows that are amplified under pressure. In 2011 he and a former college roommate launched the financial management company SigFig. Conrad’s partner pushed him out after a year.

According to interviews and talks he gave during happier days, Conrad came up with the idea for Zenefits in 2012 while trying to offer health insurance to SigFig’s few dozen employees. He’d become familiar with the intricacies of the U.S. health-care system while battling testicular cancer during his early 20s, and he even found navigating SigFig’s health-care options cumbersome. “It was an area that was a real pain point for me personally,” he told TechCrunch last year.

Before a company in the U.S. offers health insurance to its employees, it has to comb through thousands of plans offered by hundreds of insurers, and then select the two or three options that best suit its employees, based on age, marital status, and location. The smaller the business, the less likely it is to have someone with enough time to sort all that out. “We have 10 employees. We’re not big enough to have an HR department,” says Scott Yates, founder and chief executive officer of BlogMutt, a content service startup based in Boulder, Colo. “We can’t do this on our own.”

Enter the health insurance broker, who helps companies figure out what they need. Brokers are paid by insurance companies each time they sell an insurer’s plan. Sometimes they’re paid a flat rate, sometimes a cut of the plan’s premium. BlogMutt once used a broker. “We had an actual living, breathing human come talk to us, and it was great,” Yates says. “But he wasn’t always available. The promise of Zenefits has a lot of allure for a company like us.” BlogMutt signed up a year and a half ago.

Zenefits launched in April 2013 through startup incubator Y Combinator. Conrad had recruited one of SigFig’s top engineers, Laks Srini, to be his co-founder, and he expanded Zenefits’ scope to include payroll, 401(k)s, Cobra, offer letters, and other HR-related chores. Within eight months the startup was on track to make about $1 million in recurring annual revenue and had landed the investment from Andreessen Horowitz. (Srini declined to comment. Bloomberg LP, which owns Bloomberg Businessweek, is an investor in Andreessen Horowitz.)

That gave Zenefits—and Conrad—legitimacy. Andreessen Horowitz is one of the top VC firms in Silicon Valley, the big checkbook behind dozens of success stories, from Airbnb to Skype. Conrad was the type of person the firm was looking for. He could deliver a lecture on the inner workings of the insurance industry as easily as he could tell you what type of company Zenefits could become.

“I’m excited about meeting incredible human beings that change the world that we walk on, literally, and how we walk on it,” Dalgaard says. “I don’t think I’ve ever seen anyone—and I’ve seen a lot of people—think as comprehensively as [Conrad] did about the market he was going after and how to build his product.”

Andreessen Horowitz led two fundraising rounds totaling $82 million for Zenefits in 2014. Dalgaard joined the newly formed board. His job was to help Conrad develop Zenefits into a mature corporation. When Conrad set the company’s 2014 recurring revenue goal at $10 million, Dalgaard doubled it. “Lars sat there in his very Lars fashion and was like, ‘Why are you guys so f---ing bush league?” Conrad said at a software conference last year. Then, in Conrad’s telling at that conference, Dalgaard told him to add at least 100 sales reps to make it happen.

Zenefits started growing. It moved into a highrise in San Francisco’s SoMa district and expanded from fewer than 20 people to roughly 500, many on the sales side. As far as Silicon Valley perks go, Zenefits was pretty sparse—no personal chef, no meeting-room-cum-ball-pit. Conrad had the offices painted gray and orange and named each conference room after a character from Star Wars. By the end of 2014, Zenefits hit Dalgaard’s $20 million recurring revenue goal and was making good on its promises to shake up the brokerage business.

According to former employees, Conrad oscillated between a conviction that Zenefits was conquering the world—at the time, about 2,000 businesses used its services—and a near-constant fear that it would never make its numbers and he’d be fired. Employees were afraid that any mistake would stop the company’s growth. At a talk last year organized by Khosla Ventures, Conrad said, “There’s a low-level panic that suffuses the organization, a constant pressure to keep moving faster and faster and faster.”

Soon, Zenefits ran into snags. The Utah Insurance Department banned it from operating in the state because the agency considered the free HR software to be in violation of a law against offering rebates to customers. Zenefits was also finding that some insurance companies weren’t technologically advanced enough to integrate with its system.

“Most of these insurance transactions are still done on paper or with e-mailed PDFs,” says Sabrina Corlette, a research professor at Georgetown University’s Health Policy Institute. This increased the odds of errors—and Zenefits’ employees were often moving too fast to catch mistakes. “My dentist said they couldn’t run my insurance claim through, and it took three months to get someone at Zenefits to fix it,” says Brittany Keppel, a former office manager in Brooklyn who also had trouble getting Zenefits to sign her up for Cobra when she was let go from her job in January 2015. BlogMutt’s Yates says he deals with Zenefits problems every few weeks: “Names appear in records incorrectly. Dates are wrong. I don’t know where the breakdown is, but the biggest problem is there’s no quality check.”

Zenefits spokeswoman Jessica Hoffman responds: “Our focus on remediation over the last three months has included a quality initiative that has dramatically reduced error rates.”

This was the mess Sacks encountered when he joined Zenefits in December 2014. He showed up on his first day to find that he didn’t have a desk or a computer. Head count was approaching 500 employees, but the office management company didn’t have an office manager. There wasn’t even an IT person to fix the printers.

Conrad hadn’t forgotten to fill these positions. The way he saw it, Zenefits was full of gifted engineers; if their computers weren’t working, they should be able to fix them. Of course, time spent futzing with the Wi-Fi was time not spent building the product or dealing with customers. To Conrad, that just meant the engineers needed to work harder. “I guess the way I saw the chaos at that point was that it was undesirable,” Sacks says, “but if there was none of this, then there wouldn’t be a need for me to be at the company.”

Sacks says one of his earliest battles with Conrad was over the idea of hiring a receptionist. Sacks won—a friendly young woman now greets visitors from behind a sleek, white desk—but lost a battle to get Conrad to quit micromanaging the company’s HR decisions. Zenefits uses its own product to manage its employees, and Conrad controlled the account, which meant he personally approved every benefits change or vacation request for hundreds of (later, more than 1,000) employees. “We have people in HR now, but they actually don’t have access to the HR system,” Conrad said in an interview at TechCrunch Disrupt last year. “I do all of it myself. I’m a little crazy.”

The problems were much deeper than a disorganized office. Insurance brokers must pass a state licensing exam before they can legally sell or advise people on insurance. Each state has a different exam and training requirements. In California, brokers had to spend at least 52 hours on an online training course. Zenefits says Conrad created a Google Chrome browser extension that allowed people to bypass the 52-hour rule by making it appear as if they were working on the course when they weren’t. The extention was called “the macro.” (Conrad declined to comment on the Chrome extension.)

Every state has a different brokerage exam. A person licensed only in California can’t legally sell insurance to a company in Kentucky unless he applies for a reciprocal license. Zenefits sold its product all over the country, and its managerial chaos seems to have kept it from properly tracking who passed their state exams or got reciprocal licenses.

Conrad worked closely with Sam Blond, Zenefits’ trim, tanned vice president for sales, who liked to party. When their team closed a big account, Blond and Conrad poured shots for everyone in the office. Zenefits had kegs in its office. In an onstage interview at a tech conference in 2015, Blond told a story about how he and Conrad once got so drunk at a San Francisco steakhouse that they wrestled each other on the restaurant floor.

“Half an hour ago we were in the greenroom, and we saw two bottles of unopened drinks,” Blond said at the same tech conference while sitting onstage with Conrad. He smiled. “They’re no longer unopened.” (Blond declined to comment for this article.)

Last year the management company that runs Zenefits’ office space complained about cups of beer and used condoms found in the stairwells. “Do not use the stairwells to smoke, drink, eat, or have sex,” Emily Agin, Zenefits’ director of real estate and workplace services, reminded employees in a companywide memo.

Sacks wasn’t averse to a little debauchery. A few years ago, according to multiple news reports, he hired Snoop Dogg to perform at his Marie Antoinette-themed 40th birthday party. Instead of invitations, Sacks sent little cakes. He showed up in period dress. But at Zenefits, he was different. Employees say he didn’t socialize. He stared at his phone in meetings. He sometimes even escaped to an office in a nearby building for alone time.

Most of Zenefits’ problems remained internal. Investors couldn’t see them. Those they could see seemed fixable—as demonstrated when Sacks flew to Salt Lake City and stood behind the Utah governor while he signed a law allowing Zenefits to operate in the state. In May 2015, six months after Sacks joined the company, it raised $500 million in a third fundraising round, at a valuation of $4.5 billion.

Zenefits also said it would quintuple annual recurring revenue, to $100 million, by the end of 2015. “They used to say they were the No. 1 supplier to Anthem for small-group business [companies with 50 or fewer employees]. It wasn’t true,” says Darrel Ng, a spokesman for Anthem Blue Cross. Zenefits says it regrets any factual exaggerations that occurred under prior leadership.

Still, Zenefits kept growing. By the middle of 2015 it was serving 14,000 businesses and in the process of hiring more than 1,000 additional employees. “They told me that they were expanding so fast that once a month had gone by, I was guaranteed to move up,” says Aaron Semaan, 28, who joined as an entry-level sales rep in April 2015, making $35,000 a year.

The headquarters became a hodgepodge of disconnected office space sprawled over two, then three, then four floors of the highrise. Zenefits also opened a satellite office in Arizona that it packed with low-level sales reps who cold-called prospects. One manager in Arizona says he interviewed and hired people so quickly that when they showed up for work, he’d often forgotten who they were.

Semaan was one of the few who came to Zenefits already licensed as an insurance broker. And yet, for some reason—he still doesn’t know why—he wasn’t allowed to sell insurance. He was a coldcaller: When he found an interested company, he passed it off to the broker assigned to close the deal. But, as the company has publicly acknowledged, not all Zenefits brokers had passed their tests.

“I was giving [clients] to people who didn’t have their insurance license,” Semaan says. “They were selling insurance illegally. Like, it was really illegal.” He says that one woman worked at Zenefits as an account executive for four months without passing her exam. “I had multiple talks with my boss about it. I would tell him, ‘I don’t think it’s fair or right for me to be setting people up to make these sales when they don’t even have licenses.’ ” His boss didn’t respond to interview requests.

“The company culture of pressuring and bullying employees to cut corners and do the wrong thing is over,” says Sacks.

Zenefits began courting what it calls enterprise companies—those with more than 100 employees. Unlike small businesses, bigger companies have entire HR departments that enjoy long-standing relationships with insurance brokers. Sacks says he thought it was unrealistic for Zenefits to pursue them and that Conrad disagreed. When sales started to flatline, Sacks says he pushed for a hiring freeze and that Conrad thought such a move was unnecessary.

By the fall of 2015, it was clear Zenefits would miss its $100 million recurring revenue goal (the reality wound up being closer to $63 million). Some salaries were cut; Semaan says his dropped to $30,000. Conrad instituted a vacation ban, but Sacks ignored it and headed to the Caribbean. He also talked to Dalgaard about wanting to leave the company.

Dalgaard played intermediary between Conrad and Sacks. “I was trying to help wherever I could,” he says. “I spent more time than is normal just hoping that [Conrad] would change, because this could be one of the biggest companies we’ve ever heard of in Silicon Valley. And I still think it can.”

Sacks knew there was a problem as soon as he walked into the Jessika Pava room, named after an X-wing pilot who fought against the First Order in the battle of Starkiller Base. Five lawyers, including a few he’d never met, sat grimly in orange swivel chairs. It was late afternoon on Jan. 25, and they’d come to Zenefits to present the results of a two-month investigation the company had initiated. Sacks knew it doesn’t take five lawyers to deliver good news.

Zenefits had hired Cooley, a law firm, along with consulting firm PwC, after BuzzFeed published a series of articles about Zenefits’ use of unlicensed brokers. Washington state’s insurance commissioner had already opened a formal investigation into the company; according to BuzzFeed, 83 percent of all Zenefits sales in the state were made by unlicensed brokers. The company could be fined as much as $20,000 for each violation. If this were true in other states, Zenefits was looking at millions of dollars in fines.

The lawyers laid out a more damning situation. In California, they found, some of the sales team used Conrad’s macro to systematically cheat on the state’s training course, which included a section on ethics. “As far as a company doing what Zenefits has done, I don’t know that we have seen this before,” says Nancy Kincaid, press secretary for the California Department of Insurance, which has also opened an investigation. In March, Massachusetts’ division of insurance opened a third. Zenefits confirms that other states have since followed but won’t say which ones or even how many.

Sacks says he knew of the macro but didn’t know its significance or about Conrad’s involvement until the lawyers explained it to him in January. Cooley confirms Sacks’s account in a memo obtained by Bloomberg Businessweek. The memo also says that during its investigation, “Mr. Conrad acknowledged that he had authored the macro.” Meanwhile, a person close to Conrad claims he had told one of Zenefits’ lawyers about the macro more than a year earlier. Some of the Zenefits employees who weren’t properly licensed worked as benefits advisers, not salespeople. They reported up the chain to Sacks. Sacks says he wasn’t aware of their licensing problems until everything else came to light.

On Feb. 1, Zenefits held an emergency board meeting. The licensing problems and the macro were discussed. Dalgaard suggested to Conrad that, as the person who created the program, he needed to leave. Conrad resigned from Zenefits on Feb. 8; since then he’s spent part of his time at home binge-watching Star Wars: The Clone Wars, according to a friend. A person close to Conrad says he regrets resigning and is already working on a new company.

Sacks became CEO and is guiding Zenefits through its crisis cleanup. He has banned alcohol at the office and changed the company motto from “Ready. Fire. Aim.” to “Operate With Integrity.” In February the company laid off 250 employees, including the enterprise team. Sales Vice President Blond, Semaan’s boss, and any executive or manager known to have helped disseminate the macro are also gone. Zenefits says it has self-reported the findings of its internal investigation to all 50 states and is working with those that have opened formal inquiries.

Fidelity Investments, which owns a stake, has slashed its valuation of Zenefits from $4.5 billion to less than $2 billion. There are rows of empty desks at the San Francisco office; the company plans to downsize from four floors to three. The Star Wars-themed conference rooms will soon be renamed after inspirational entrepreneurs. Kegs have been replaced with cold-brew coffee. The stairwells are condom-free.

And yet, despite his downfall, Conrad is still a coveted name in Silicon Valley. People want to meet the man who created a $60 million company in just three years and made good on his promise to shake up the insurance industry. Dalgaard says he’s been getting e-mails from people eager to work with Conrad since the day his resignation went public.

Zenefits might also survive for the one reason that made its product so appealing to business owners in the first place: Shopping for health insurance remains really frustrating. The company says it now has 20,000 accounts. “As long as their problems don’t affect our company, we’ll stay,” says Todd Harmond, vice president for finance and operations of the e-book service Scribd, which uses Zenefits to offer Kaiser Permanente and Anthem health insurance plans to its 85 employees.

“Unless something else goes really wrong with Zenefits, we’ll stick with them for a while,” says BlogMutt’s Yates. “It’s too much of a hassle to switch.”

Copyright 2018 Bloomberg. All rights reserved. This material may not be published, broadcast, rewritten, or redistributed.