Getting to the Other Side, Safely

I was fortunate to be at GitLab when Duo went from roadmap slide to shipped product, and feel even more fortunate that the experience helped shape my view and use of AI. The thing I find myself returning to, one that resonates strongly with how some of the leading voices in AI safety talk about this moment, is less about how fast we move and more about making it to the other side of the adoption curve safely. Speed is not the enemy. Lack of control is.

That's not an abstract principle for me. It's the foundation for my career: the fastest program isn't the best one if nobody checked whether the incentives, enablement, the measurement, and the guardrails were built in from the start to align with the customer lifecycle rather than bolted on after the fact. AI adoption is the same problem at a different scale, with safety rather than profitability as the risk factor.

As a long-time participant in the partner ecosystem game, I'd argue part of the solution to getting to the other side safely is partners.

Partner in the loop

Technology doesn't democratize itself. Every powerful capability I've watched reach real people at scale, cloud infrastructure, SaaS, now AI, got there through an ecosystem of partners who translated it, implemented it, and stood behind it in the room with the customer. Partners are the amplifying glue that turns a capability into something a business can actually trust and adopt. Take that layer away, and the technology stays powerful in a lab and stays a stranger to most of the people it could actually help.

Human in the loop is the discipline that keeps AI accountable at the point of use. Partner in the loop is the same discipline at the scale of an entire market: keeping a trusted, accountable relationship in the path between a powerful new capability and the business that has to decide whether to bet on it. Strip that layer out to move faster, and you get the same failure mode, more capability, less control, just at ecosystem scale instead of individual use.

That's the belief underneath everything else in this piece. AI is probably the biggest test of it I'll see in my career: whether the industry lets partner ecosystems do what they've always done, carry trust, context, and accountability into the last mile, or treats them as friction to route around in the name of speed. I know which side of that I want to help build.

Duo taught me the human-in-the-loop lesson from watching a product get built. The years before and since have taught me the partner-in-the-loop version of the same lesson, at a scale I've spent my whole career inside of. Both point to the same place: getting to the other side of this shift matters less than getting there with someone accountable still firmly in the loop, whether that someone is a person at a keyboard or a partner standing next to a customer.