Coinbase restructure is a huge bet on AI productivity
Cryptocurrency platform Coinbase announced this week that it would lay off 14 percent of its workforce – an estimated 700 people. Cofounder and CEO Brian Armstrong took to X to frame the move as part of an effort to “rebuild Coinbase to be lean, fast, and AI-native”. The company sees its future as “an intelligence, with humans around the edge aligning it”, he explained.
In his reasoning, Armstrong pointed to a volatile market in a down period, with Coinbase needing to adjust its cost structure to prepare for its “next phase of growth”. But the main theme of his post was AI, which he said has already changed the way Coinbase works: “Non-technical teams are now shipping production code and many of our workflows are being automated.”
Checks and balances
Understandably, there were a few eyebrows raised over this remark. Coinbase is at its core a financial company and holds customers' cryptocurrency – it’s the last platform you’d want to be vibe coded. Armstrong was quick to follow up: “It goes without saying that all AI generated code has rigorous human reviews. No one is vibe coding directly to production. We're increasing speed of shipping and innovation, while continuing to raise the bar on security.” However, we have no further information on who performs these reviews or exactly what the process is.
The timing raised questions – this was Coinbase’s third headcount reduction in recent years, announced days before a Q1 earnings report that came in below expectations on both revenue and losses. Critics argued that the AI framing was cover for underperformance. Maybe. But the more interesting question is the AI-centric way Coinbase plans to operate – and whether the organisational model Armstrong described holds together in practice.
A flatter structure
The first of Coinbase’s changes is the least controversial. It intends to simplify its organisational structure to boost the speed and efficiency of its decision-making, with a maximum of five layers below the CEO and COO.
“Layers slow things down and create coordination tax. The future is small, high context teams that can move quickly. Leaders will own much more, with as many as 15+ direct reports. Fewer layers also means a leaner cost structure that is built to perform through all market cycles.”
The logic here makes sense. Businesses with many layers take a long time to get things done, because any impactful decision must move up the hierarchy until it reaches somebody with the right level of authority, and outcomes must flow down through the layers to reach the people doing the work. As with photocopies, the message loses a little of its clarity with each transfer.
But there is a practical trade-off. Between day-to-day people management, performance reviews, and so on, managing 15 or more direct reports is a full-time job in itself. Leaders of bigger teams are going to spend more time managing their people, and will have less time to focus on the work.
The end of pure management
Unfortunately, the next point in Armstrong’s post is that Coinbase wants managers to spend less time managing. It wants to put an end to “pure managers”, with everybody rolling up their sleeves to get the work done.
“Every leader at Coinbase must also be a strong and active individual contributor. Managers should be like player-coaches, getting their hands dirty alongside their teams.”
This contradicts the previous point. In a flatter structure, managers become more important. They have more people to take care of and keep aligned with the company’s overall goals. It’s hard to focus on writing code when you have so many people seeking your guidance and direction on a daily basis.
Tony Fadell, who built teams at Apple and Nest, is clear on this. The early chapters of his book Build focus on his transition from individual contributor to manager, and the change in mindset required to adapt to the new role:
“Once you’re a manager, you’re no longer an accountant. Or a designer. Or a fisherman. Or an artist. Or whatever it is you really enjoyed doing… You now lead a team of people doing what you used to be good at. So at least 85 percent of your time should be spent managing. If it’s not, you aren’t doing it right. Managing is the job. And managing is hard.”
To Armstrong’s first point, Fadell says managers must constantly look “up and out” – up for senior leadership’s direction, and out at other teams’ work. That’s how teams across the organisation stay aligned, and it’s hard for a manager to maintain that visibility if they’re too focused on the doing.
The AI-centric future
Lastly, Coinbase is recentring its workforce around AI. Armstrong imagines a future where individuals will have greater impact by utilising large numbers of AI agents to complete work previously done by entire teams.
“We’ll be concentrating around AI-native talent who can manage fleets of agents to drive outsized impact. We’ll also be experimenting with reduced pod sizes, including ‘one person teams’ with engineers, designers, and product managers all in one role.”
It all sounds very cutting-edge, but these are all specialist roles. AI is useful as an accelerator, but it still requires oversight by specialists who can make informed decisions – that’s the role of those engineers and designers. People who can both review production-ready code and also design a user interface to a level good enough for a commercial financial service are unicorns.
There’s also the issue of AI lock-in. A couple of weeks ago I wrote about how AI prices will continue to rise, and an organisation that relies so heavily on the technology for productivity is especially vulnerable. Coinbase’s changes are framed as moves that will enable the business to weather volatility, but such AI dependence leaves them exposed to one of the biggest cost risks of modern times: that one day we’ll all have to pay what AI costs to run.
