We recently sat down with Alok K. Agrawal, former Chief Strategy Officer at Celestica and Managing Director at Agrawal Capital, one of our Venture Partners at Silicon Foundry. With a career spanning strategy and finance across global markets, Alok brings a pragmatic perspective on how large enterprises scale and innovate while legacy businesses are still strong. In this conversation, he shares lessons from repositioning complex organizations to navigating the build-buy-partner continuum in a way that drives real, lasting impact.
Press play to listen to the conversation
TANYA PRIVÉ: We recently welcomed our Venture Partner, Alok K. Agrawal, to the team. Alok, welcome! We’re thrilled to have you on board!
ALOK K. AGRAWAL: Hi Tanya, thank you for having me!
TANYA: Let’s start with your experience as the Chief Strategy Officer at Celestica. You were a crucial part in leading its transformation as a major player in AI infrastructure and advanced hardware. What lessons can corporates draw from that experience about repositioning for the next decade of technology shifts?
ALOK: At Celestica, something crucial I learned was that repositioning isn’t a messaging exercise. You can see it in the budget and the organizational chart. The work is stepping back from areas where you can’t win long-term and putting real money into what customers will pay for, such as strong engineering, advanced manufacturing, and consistent execution. I’ve realized that small pilots won’t change a company on their own. Instead, what changes a company is leadership making hard calls while the legacy business is still profitable. If you wait until the old model stops, you don’t have many options left.
TANYA: That’s a fascinating insight, especially the idea that real transformation requires commitment while the legacy business is still strong. Many corporate M&A and venture efforts struggle to translate activity into sustained impact. From your work leading strategy, M&A, and Celestica Ventures, what separates high-impact corporate innovation models from the ones that stall?
ALOK: The difference is being clear about intent and being willing to act on what you learn. Strong programs know what each deal is meant to achieve: capability, speed, talent, customer access, or risk reduction, and they track whether that’s happening. Weaker programs hide behind “optionality” with no owner and no decision triggers, then count activity instead of outcomes. If the operating leaders don’t own the result, it stays on the side and never actually scales. And if you don’t shut down what isn’t working out and move resources, you end up with a portfolio that consumes attention without paying it back.
TANYA: Many of your investments focus on AI, space, defense, robotics, and automation – areas where corporates are trying to build capability quickly. Where do you see the most immediate opportunities for large organizations to partner or co-develop with startups?
ALOK: I would say the fastest opportunities are where you can put something into operations quickly and learn fast. You often find that in physical workflows—automation, inspection, logistics, maintenance, and field operations—where improvements show up in throughput, safety, reliability, or labor efficiency. Those environments let corporates and startups move beyond pilots and build something that runs day to day, not just in a lab.
However, on the AI side, a lot of near-term value is moving intelligence closer to where work happens. Power limits, latency, connectivity, and security often make “everything in the cloud” unrealistic. Edge approaches—perception, control, and decision-making integrated into existing systems—tend to pay back faster than broad platform bets. The common thread is partnering when speed and learning matter, and co-developing when the real-world environment is what makes the solution work.
TANYA: Shifting from how companies scale globally to how they access new capabilities, you’ve led both acquisitions and strategic partnerships. When leaders are evaluating emerging technologies, how should they think about the build-versus-buy-versus-partner decision?
ALOK: Build vs buy vs partner isn’t always the first decision. The better starting point is which companies matter to your strategy and what role they could play. Once that’s clear, the structure can evolve as you prove adoption and build conviction. Some of the best outcomes start with close collaboration to understand the technology and how it fits the business, then transforming it into a partnership, an investment, a commercial agreement, or an acquisition as the importance and confidence increase.
Where companies get stuck is picking a structure too early. For instance, they could treat partnerships as a low-effort option or push for an acquisition before they understand what integration will take. A practical pattern is to partner first to prove adoption, invest to deepen alignment, and acquire only when the capability is truly core and you’re prepared to integrate it.
TANYA: From your vantage point as a board member and investor, what are the early signals that a startup is ready — or not ready — for enterprise-scale collaboration?
ALOK: Some early signs that I look for are whether a startup is thinking beyond the initial sale. The ones that are ready often talk about integration, security, deployment, and reliability because they’ve already experienced and dealt with those realities. Furthermore, they have documentation, a plan for procurement and security review, and customer success or services that don’t depend on the founder stepping in every week. They understand that the hard part starts after the public offering, when the solution has to survive rollout and operate reliably at scale.
On the other hand, the not-ready pattern shows up quickly. They treat enterprises as bigger checks, underestimate compliance and data access, and don’t have a plan for what happens when the first deployment hits real-world friction. I’m often drawn to founders who came out of the industry because they’ve felt that friction firsthand. The demo usually isn’t the problem. Running in production is.
TANYA: Given your work building a venture syndicate and portfolio across critical infrastructure, what risks or blind spots should corporate leaders be paying closer attention to in 2025 to 2030?
ALOK: Leaders often underestimate how quickly assumptions change once plans meet reality. Supply and access can look stable until a single dependency—one supplier, one process step, one qualification—suddenly determines the entire roadmap. Moreover, geopolitical shifts can also force parallel tech stacks, duplicated compliance work, and separate operating models that drive cost and complexity much earlier.
Another blind spot is how messy “early progress” can become later. Rapid AI adoption can create scattered systems that are hard to govern and expensive to scale if the architecture isn’t thought through upfront. And the constraint that keeps showing up is people—not generic “AI talent,” but engineers and operators who know how to make complex systems run reliably in production. That capability takes time to build and becomes a limiting factor sooner than most teams plan for.
TANYA: Finally, looking ahead, where do you see the most compelling opportunities for you and Silicon Foundry to jointly support corporates in navigating frontier-tech ecosystems?
ALOK: Most corporates don’t struggle to find startups. The real challenge is turning early interest into something that actually runs at scale. The opportunity is helping teams move from pilot to production in a repeatable way, starting with a practical readiness check that separates a good demo from something that can operate reliably day to day.
That means being clear on partnership terms around IP and data, and having a real plan for what needs to change across IT, operations, and procurement to roll something out across sites. The goal isn’t just to get one project over the line, but to build muscle inside the company so it can do this again and again. That’s especially important in areas like AI infrastructure and industrial automation, where execution and reliability matter just as much as the underlying technology.
