Summer field trip 2.3: Is Google too big to succeed?

Welcome to part three of my summer field trip into Google’s corporate culture. Part one talked about leadership myths, using former ads chief Sridhar Ramaswamy as an example. Part two told tales of strategic planning drama and farce. Today, I share how I went from a workaholic in tiny startups to a disgruntled cog in a 200,000-person leviathan. This one is long, so hoist the sails and join me on the wine-dark sea one last time.

I spent most of my early career at dot-com startups. I was never a founder, but helped them grow to dozens of people before they inevitably died and fell out of human memory. (If you ever visit the startup underworld like a venture-capital Odysseus, give my regards to Visual Radio, Wolff New Media, NextPlanetOver, and Eazel.)

Small companies need flexible, multi-talented employees. There’s too much work and not enough people for someone to say anything is “not their job.” We were like a planarian flatworm, where any chunk of its body is capable of regrowing into a whole new animal. Besides writing software, I took the bus to Ikea to buy and assemble my own desk; arranged a crosstown office move; and manned a booth at San Diego Comic-Con.

Startups were a great place for me in my 20s, because I had the freedom to try new things without worrying about whether it was my job or not, and it was easy to share ideas with the company’s leadership. I ended up spending almost all of my time at the office. At one startup, I spent a long weekend teaching myself how to build a (really poor) search engine, for fun, which ended up being part of the pitch that won a big investment. 

Being small also made us vulnerable, because there was no slack. I often worked 80-hour weeks because there was nobody else to cover for me. This also meant one bad egg could put the whole company at risk. Worst of all, sometimes the bad egg was the CEO. Michael Wolff (now famous for writing tell-all books about the Trump administration) convinced his employees to defer parts of their salaries, and then when he lost control of the company, he resigned, emptied the payroll account, and took his family on vacation to Italy with the proceeds. Young employees who had been living on credit cards because they trusted him were left crying and calling their parents to help make rent. Not as bad as getting eaten by a cyclops, but still pretty bad.

When I joined DoubleClick in 2004 after business school, it had around 1,500 employees. I was in heaven. My team was big enough to do meaningful projects and cover for each other, but the company was small enough that I still had access to senior leadership. I was able to shape a strategy that led to my product doubling its revenue to over $150 million by the time the Google deal closed in 2008, and I got to travel all over the US and Europe meeting with customers.

Company culture at this scale brought new challenges. Leaders didn’t have time to meet with everyone personally, so had to rely more on regular all-hands meetings to keep people informed and on track. Internal politics also got more complicated. Instead of settling arguments between individuals on a small team, I needed to negotiate with functions and product teams that each had their own visions and plans, and were often contending for the same limited resources.

One of the ways DoubleClick tried to address this was standardizing how we worked, to ensure everyone’s gears would continue to mesh as we grew. We started requiring every team to use the same software development methodology (Scrum), which meant doing planning at the same time every month, and using the same tools for tracking projects.

I liked Scrum, but it exposed how for some people, even a thousand-person company was too much structure. I vividly remember an all-hands meeting stopping dead while one person stood up and repeatedly told one of the founders of Scrum that they were doing software development wrong. It might have only been ten or fifteen minutes of filibustering, but for the hundreds of us sitting awkwardly in the audience, it felt like being trapped on Calypso’s island for seven years, if Charlize Theron was an irate software engineer.

Google had 17,000 employees when the DoubleClick deal closed. I hated it at first. I had trouble accepting that the jump in scale made corporate politics even more important. I naively thought I could still influence other organizations by pleading my case effectively. I didn’t understand that other teams in a company this big had completely separate power bases and patrons, which meant they could (and would) ignore me if my goals didn’t fit into their agenda.

Fortunately, Google’s chaotic heritage gave me enough autonomy to carve out a niche. People used to joke that Google at that time was less of a company than a collection of loosely affiliated startups. This meant that as long as I stayed in my lane within Display ads, I could still have a lot of influence. The variety also meant I could find people in other teams who saw things my way. I remember meeting Google salespeople who were thrilled that, after years of dealing with Google product managers who preferred data to people, I actually wanted to meet with customers.

I also felt like I had allies at the top. Larry and Sergey were still around and trying to live up to their 2004 IPO prospectus promise that “Google is not a conventional company.” They led a weekly all-hands meeting where they took unfiltered employee questions and gave candid answers. This was a company that would still try big things with personality. When Chrome launched in 2008, they got one of the world’s most prestigious comics artists, Scott McCloud, to write and draw an introduction to its technology.

By the end of my Google career in 2023, it was a fundamentally different beast, with almost 200,000 full-time employees, plus another 150,000 contractors. If startups were flatworms, Google in 2023 was a blue whale – a massive, complex organism full of specialized organs and structures all designed to keep it alive, swimming, and consuming vast quantities of food (i.e., data and money).

Becoming a blue whale also meant a huge proportion of the company’s work was internal. Math nerds will know the square-cube law, which shows that the volume inside a solid expands much faster than its surface area. Most of the cells of a tiny flatworm are exposed to the outside world, but in a blue whale, the vast majority of its cells are buried so deep inside the body that they probably don’t even know they’re underwater.

Keeping the guts of a leviathan in good working order required predictable employees who were happy to perform their hyper-specific duties without making waves. Being flexible and entrepreneurial would threaten that equilibrium. This had some logic. We have a word for a cell that tries to stop being part of its designated organ in order to become something else: cancer. 

I tried to carve out space for originality without breaking the system. I made a point of hiring at least one independent-minded, lateral thinker with strong connections to the outside world on several of my teams. They brought new ideas to the table and helped keep us from getting stale. But it took a lot of time and effort to integrate them into Google. And we lost a lot of battles too, less on the merits of our case and more because other teams would criticize us for not spending enough time getting them on board up front.

This was another factor in my decision to leave Google. I couldn’t get people to understand how the focus on internal consensus was crushing anything new and different. I felt more pressure to fit in than to have good ideas or to do good work. I don’t mean to say that I don’t believe in dialogue or compromise. But when the opposing side in every debate is “don’t rock the boat,” and you can’t proceed until everyone is satisfied, the results are often neutered and boring.

I was gone before the AI era kicked off in earnest, but I was not surprised that Google almost blew it. It’s been widely reported that Google actually invented the technology underlying large language models and built a chatbot before Open AI, but didn’t launch it, due to the risk aversion I described above. And even when they did start moving, their initial response (a chatbot called Bard and an awkward presentation in Paris) was widely panned. Google had spent years selecting leaders based on their ability to be predictable and agreeable deep inside the whale. These would never be the right people to innovate and get the outside world excited.

Google got lucky because it had an innovation Trojan horse. Larry and Sergey bought the AI company DeepMind in 2014, when it was only 75 people, and left it alone. DeepMind kept their headquarters and most of their team in London, far from Silicon Valley, and didn’t even report to Google CEO Sundar Pichai for their first six years. When chatbot mania hit, the DeepMind team came pouring out of London like Greek soldiers out of a horse, but instead of burning everything down, they launched Gemini, and helped Google get back into the AI race.

Blue-whale Google would never have been able to build a team like DeepMind on their own. Google has tried to build internal innovation teams repeatedly, such as Google X and Area 120, but they’ve never been able to make them stick. Innovative people are either filtered out by a hiring process focused on finding predictable people who fit in, or crushed quickly after hiring by the internal pressure for safe consistency.  [Waymo is a rare exception, because like DeepMind it’s been kept almost completely clear of Google.]

Sadly, recent reports suggest that Google might be reverting to form, even with AI. DeepMind CEO Demis Hassabis has stepped away, and the rest of the team is being integrated more tightly into Google. Some news reports say this was because Demis wanted to push ahead to new forms of AI technology like world models, but Google wants to stay focused on improving Gemini and rolling it out to more existing products. “Google, they said, was preoccupied with protecting its search traffic and beating OpenAI on chat, while rivals were running ahead on agentic and coding capabilities.”

Gotta keep that big blue whale swimming, after all.

Ideas? Feedback? Criticism? I want to hear it, because I am sure that I am going to get a lot of things wrong along the way. I will share what I learn with the community as we go. Reach out any time at [email protected].