clear: “If you’re not owning the infrastructure, you’re always going to be behind.” For Zendesk, this meant investing heavily in building their own data infrastructure that could handle the massive volume of customer interactions without relying on external vendors. This approach allowed Zendesk to tailor their systems specifically to their needs, enhancing efficiency and control over data processing.
LinkedIn, however, took a slightly different approach. While they built their own orchestration systems to manage agent workloads, they still leveraged open-source tools where feasible. Singh noted that open-source solutions provided a flexible foundation that could be customized to meet LinkedIn’s specific requirements. This balance allowed them to innovate internally while still benefiting from the rapid advancements happening in the open-source community.
Walmart’s strategy was to foster an internal culture that encouraged experimentation with open-source tools but within a controlled framework. This approach allowed them to harness the creativity of their “citizen developers” while maintaining oversight to prevent chaos. Gosby emphasized that governance was key, ensuring that open-source projects aligned with company objectives and didn’t become a distraction.
### Implications for Founders and Engineers
For founders and engineers, the message from VB Transform 2026 is clear: the challenge isn’t just building AI models but architecting the infrastructure they operate on. This is a call to reevaluate existing systems and consider whether they are truly equipped to handle the demands of AI agents. Legacy systems, while stable, might not be agile enough to support the fast-paced decision-making processes that AI agents require.
Engineers should anticipate the need for custom solutions, particularly in orchestration and data management. This means investing time and resources into understanding the limitations of current infrastructure and identifying areas where bespoke engineering can make a difference. The discussion also highlights the importance of governance in managing AI projects, ensuring that innovation doesn’t spiral into unmanageable complexity.
For investors, the insights shared by LinkedIn, Walmart, and Zendesk suggest that the next wave of AI advancement might not come from new models, but from the companies that successfully bridge the gap between existing infrastructure and AI capabilities. Startups that offer solutions to these infrastructure challenges or provide platforms that facilitate seamless integration of AI agents could be poised for growth.
### What’s Next?
As AI agents continue to evolve, the focus will likely remain on optimizing the infrastructure they rely on. Companies will need to balance the benefits of open-source tools with the need for tailored, proprietary systems that meet specific business needs. The journey from pilot to production will require a strategic approach to infrastructure, with a keen eye on both performance and governance.
For founders and engineers, the lesson is to prepare for a future where infrastructure agility is as critical as AI capability. This means building systems that can adapt quickly and efficiently, ensuring that AI agents are not just fast thinkers but also fast doers.