Steve Woods joined Inovia Capital as Partner and CTO in 2021 and has worked in software development, research and technical leadership for more than 30 years.
As co-founder of three software and services companies based in the USA (Quack.com, Kinitos and NeoEdge Networks), he built strong, high-performing teams of product and technical experts based in both Silicon Valley and the Waterloo-Toronto tech corridor.
Steve also led the growth and leadership development of the Google Canada product and engineering teams from 20 to over 1,300 people across a range of impactful products between 2008 and 2021.
In this conversation, we’re glad to hear Steve’s perspective on how our industry talks about AI, what we tend to get wrong and his his perspective on founder-market fit in today’s environment.
M.R. Rangaswami: Everyone in venture is making AI bets right now. What is the most misunderstood aspect of AI today?
Steve Woods: What people get wrong frequently is treating AI models as deterministic systems. They’re not. These models are non-deterministic by nature, meaning the quality and content of what you get as an output is not fixed and will not be identical from one query to the next. As one example, I recently watched a founder discover his error rate at peak hours California time was meaningfully worse than at midnight because the model provider was quietly throttling compute to support training.
Most people still think of AI the way they think of a conventional computer: reliable and consistently producing the same result every time. Building at scale on top of these models means accepting that you are working with non-deterministic statistical systems and designing accordingly in answering questions, generating content and, of course, using generated code. One has to be obsessive about anticipating and constraining error appropriate to situations, measuring variance and building test coverage that tells you when something has drifted.
The key is accuracy appropriate to the situation, and that means applying the right approach for each use case. All software exhibits inherent errors, AI or otherwise. What is of consequence is whether the platform is suited to the problem you are solving.
Another misunderstood aspect is how much authority people assign to AI’s outputs. I understand why it happens. When you don’t know the answer, having something that sounds confident feels like a gift. But these models are producing statistically plausible responses, not verified ones, and that distinction is enormously important when you’re making decisions that affect real customers.
The best AI-native founders I back understand what they’re working with and how to drive value and effectiveness. They are curious and careful in equal measure, and they are never confused about who is responsible for the answer.
M.R.: You’ve built and scaled engineering teams at Google and now work closely with founders. How is AI fundamentally changing what a high-performing engineering organization looks like?
Steve: The question I get more than almost any other right now is some version of: how many people can I cut? That is entirely the wrong question. The way to excel in today’s market is to ask how to do things no one has done before with the people you have. When Patrick Pichette (former CFO of Google and now Partner at Inovia) arrived at Google, every engineer in the room was expecting to be constrained. But Patrick walked in and said, we were going to do the biggest things possible, and that he was there as part of the larger team to help ensure the engineers would succeed.
Companies need to deploy that approach even more now than they did then. AI amplifies your best people. If you understand that, you can build organizations that can do things that simply were not possible before.
What I’m watching emerge in the highest-performing teams is a deliberate layering of expertise. You need strong individual contributors who show others how things work and how to approach problems, a layer of people who can coordinate that expertise across the organization and someone at the top who can see the whole gameboard and find the next problem worth solving.
The return of apprenticeships is underway, with some modern nuances like overtly mapping skills growth to business goals. Performant teams are bringing people fully into their context, teaching them how to drive true value for the business and building a team of skilled professionals who can do more, better and faster together.
Not investing in building your culture and team is a path to nowhere. Underlying culture has not changed as much as people imagine. You still need an environment where people want to build things, rather than just complete tasks. What has changed is what becomes possible when they do, and that means the possibilities for what an engineering organization can achieve have truly never been greater.
M.R.: Founder–market fit is often cited as a key lens for investing. In your experience, where does that idea break down, and how do you identify founders who simply won’t stop?
Steve: Product-market fit gets a lot of attention, and deservedly so. But in a market moving this fast, I am spending increasingly more time evaluating founder-market fit: whether this particular person is built for this particular moment and this particular problem.
What I’m looking for is a customer obsession that runs deep enough to keep generating new solutions, not just defending the first one, and a genuine inability to stop building. For example, one of our portfolio companies has found a third and fourth way to create value for customers well past the original product.
Their market has evolved in ways we did not predict when we originally invested, and this is because the founder is wired to look at a customer problem and keep solving it. That kind of founder is rare, and, right now, is the most durable bet I know how to make.
That instinct to keep building shows up outside of work too, and it’s one of the first things I ask about. I want to know what founders are building outside of work. Every serious founder I know has a personal project such as a game engine, a tax tool or something even more niche. This tells me whether building is a mode of thinking or just a job description for them.
I also ask about early competitive experience, whether it be sports or music, because there is a meaningful correlation between succeeding at something demanding early in life and having the grit to stay with a company when things get hard.
M.R. Rangaswami is the Co-Founder of Sandhill.com