From Assistive AI to Authoritative Systems: How Wonderful is taking Enterprise AI out of Pilot Mode

H: There’s a lot of focus right now on models getting better - it seems like every week. More context, better reasoning, lower cost. You argue that model capability isn’t the bottleneck in enterprise AI. That’s a strong claim. Why?

B: AI doesn’t land in a vacuum - especially if we’re talking about the enterprise. It lands in legacy systems, conflicting definitions of truth, approval chains that happen over varying channels, and workflows that were never designed for software to execute end-to-end. Organizations underestimate how much integration, governance, and change management is required when software stops augmenting and starts acting.

AI can be transformative in these environments, but getting there takes serious investment, because you're changing how an organization operates, not just adopting a new tool. What we've seen is that it requires being on the ground, working closely with customers, being in their offices to make sure delivery actually works.

H: Most companies today use AI in an assistive way. Draft the email. Suggest the next step. Summarize the ticket. You’re helping enterprises move toward something more authoritative. Where does that line get crossed?

B: Assistive AI makes people more efficient, but it doesn’t change the operating model. Authoritative AI changes the workflow. It’s the difference between “AI suggests a discount” and “AI routes, approves, and executes the discount within defined guardrails.” The moment software is allowed to act, you’re forced to confront much harder questions: Which system is canonical? What happens when two data sources disagree? What level of error is acceptable compared to human baselines? How do you roll back decisions? Figuring out the answers to these questions and then reflecting them in technology takes work that few companies are staffed or even minded to do.

H: Most enterprise technology companies scale by centralizing - sell from a hub, deploy remotely, maximize margin. You’ve taken almost the opposite approach: hyper-local teams, embedded deeply in-region. Why?

B: Because integrating AI into the enterprise is hard, and can only be done successfully if you’re on the ground with customers, able to sit in their office, become expert in their business context, and partner with them on changing the way work gets done.

Early enterprise AI is closer to field operations than traditional SaaS. If you want agents to handle real workflows - in healthcare, financial services, telecom - you have to understand the specific local shape of those workflows, and you have to be able to be inside customer environments to get them to production. That requires proximity and a heavy investment in local presence.

H: That’s an interesting framing. A consumer-style land grab, but in the enterprise. Is that what this is?

B: In some ways, yes. I don’t think it’s ever been done before because it probably wasn’t possible before. Enterprises have traditionally moved more slowly, which made this kind of strategy difficult to execute. But the speed at which AI is moving means that there is a unique window of opportunity to become the strategic AI partner to enterprises in markets from LATAM to APAC to EMEA.

When your system is embedded deeply enough to route decisions, reconcile data across systems, and handle exceptions, you’re not just another vendor. You’re part of the operating layer. Switching costs in that environment aren’t contractual. They’re structural. The longer you’re in the workflow, the more precedent, context, and institutional memory accumulates inside the system.

H: You're one of the fastest-growing applied AI companies - you’ve expanded to 25 sites in the last 6 months and you keep adding more. What challenges does that bring, and how are you solving them?

Growth at that speed forces you to be extremely deliberate about talent. And there are very few shortcuts to making sure that you’re hiring for the right DNA and approach. We’ve optimized a lot of the process, building internal tools for sourcing and scoring candidates, but ultimately, it’s the highest leverage activity that I can be spending my time on, and I invest a lot of that time in making sure we’re getting the best people on board.