From Terrified Engineer to DeepMind Director: What Really Makes Technical Leadership Work
I recently had a fascinating conversation with Ben Coppin, who spent 13 years at Google DeepMind watching it grow from a tiny startup to 6,000 people. What caught my attention wasn’t just his impressive resume, it was his honest reflection on the challenging, human side of technical leadership.
As data and AI leaders, we’re often so focused on model performance and infrastructure scaling that we forget the hardest problems aren’t technical at all. They’re about people. Ben’s journey from self-taught programmer to engineering director earned him some insights that might help others.
A Personal Transformation
Ben shared a story of a personal transformation. Early in his career, he was terrified of his boss—”a real monster” who had a habit of walking into rooms and yelling at everyone. But when this boss came after Ben’s small team, something shifted. Despite his fear, Ben found himself marching into the boss’s office to have that difficult conversation.
“I had this immediate reaction, which is, I’m gonna go and talk to him about this. This was not okay,” Ben recalled. “Afterwards I was like, whoa. What, where did that come from? How did that happen? I can’t talk to him.”
What struck me about this: leadership often emerges not from ambition, but from a sense of responsibility for others.
Why this matters now: In our field, we promote people based on technical excellence, but the best leaders I know got there because they genuinely cared about their teams’ success. If you’re considering a leadership role, ask yourself: are you doing it for career advancement, or because you want to see others succeed?
What you can do: Before your next team meeting, spend a moment thinking about each person’s growth and challenges rather than just the technical deliverables. Notice how that shifts your perspective on the conversation.
The Dangerous Habit of Having All the Answers
Ben actively tells new technical leaders to go into meetings and not say anything. “Don’t feel like you need to talk,” he advises. “You don’t have to have the last word. You don’t have to have all the answers.”
This goes against everything we’re conditioned to believe about leadership, especially in technical roles where expertise is currency.
Why this matters now: I see this constantly in data science and AI teams. The person who becomes a manager often got there because they had the best technical answers. But suddenly their job isn’t to have the answers, it’s to help their team find them.
What you can do: In your next team discussion about a technical challenge, resist the urge to jump in with your solution. Instead, ask questions that help your team think through the problem. You might be amazed at what they come up with when you’re not filling all the silence.
The Feedback Paradox
Ben shared something that resonated deeply with my own leadership struggles: the challenge of getting honest feedback from your team. We talk endlessly about giving feedback, but rarely about receiving it as leaders.
“No matter how good the psychological safety, no matter how good the relationship,” Ben observed, “there’s always this thought in the person’s mind that, what if this somehow harms my career?”
Why this matters now: In fast-moving AI and data teams, leaders need rapid feedback loops to adapt. But the power dynamics make this incredibly difficult.
What you can do: Try Ben’s approach: tie feedback requests to a shared mission. Instead of “How am I doing?”, try “We’re trying to ship this model by quarter-end. What could I change about how I’m supporting the team to help us get there faster?” Make it about the work, not about you.
Clarity.
One of Ben’s most practical insights was about feedback clarity. He talked about the “shit sandwich” approach—positive, negative, positive—and why it usually backfires.
“It’s incredibly confusing,” he explained. “The person comes away having been told there’s this thing you need to work on, but don’t worry about it ‘cause actually you’re doing great.”
Why this matters now: In technical teams, unclear feedback is especially damaging because people are used to precise, logical communication. Fuzzy feedback feels wrong and gets ignored.
What you can do: Next time you need to give constructive feedback, focus on one thing: clarity. Be specific about what you observed, what needs to change, and how you’ll help. Skip the compliment padding, it just muddles the message.
AI as Your Practice Partner
Ben described using AI to practice coaching conversations, having it simulate an engineer with challenges and then provide feedback on his coaching approach.
“It gave me like five quite detailed bullet points of things I was doing well, and then five quite detailed bullet points of where I could do better,” he said. “It was genuinely helpful.”
Why this matters now: We’re all figuring out how to use AI effectively. This is a perfect example of using AI for what it’s actually good at: providing a safe space to practice human skills.
What you can do: Try it yourself. Ask an AI to roleplay a difficult conversation you need to have with a team member. Practice your approach, get feedback, iterate. It’s like having a coaching session without the performance anxiety.
Not the Smartest
What struck me most about Ben’s insights is how they all connect to one fundamental truth: the best technical leaders succeed by getting comfortable with not being the smartest person in the room about the technical stuff.
Instead, they become the smartest about creating conditions where their teams can do their best work.