What if I told you that the next big tech battle isn't about who builds the smartest AI, but who controls the invisible machinery that makes it work? That's exactly what's happening in enterprise software development right now. Companies like Coinbase, Shopify, and Ramp are building their own coding agents, yet they're still paying Anthropic for its models. It's not a 'build vs. buy' dilemma—it's a deeper architectural revolution. And I think this shift is going to redefine how we think about AI in the workplace for years to come.
Let me break this down. These companies aren't trying to replace Anthropic, OpenAI, or Google's models. Instead, they're creating something far more strategic: the 'agent harness.' This is the glue that connects developers to the AI, handling everything from workflow orchestration to security checks. Think of it like building a custom car engine (the AI model) but designing your own transmission, brakes, and dashboard (the harness). The model is just the fuel. The real power lies in how you channel it.
What makes this fascinating is the pattern emerging across these companies. Coinbase's Forge, Shopify's River, and Ramp's Inspect all follow the same blueprint. They integrate with tools like Slack and GitHub, automate code validation, and even produce mobile builds for review. But here's the kicker: none of them are trying to outcompete Anthropic or OpenAI. They're using those models as interchangeable parts. That's a huge shift. It means the competitive edge isn't in the AI itself, but in how you structure the ecosystem around it.
Personally, I think this mirrors the cloud computing era. Back in the 2010s, companies didn't build their own data centers after AWS came along. Instead, they built platforms on top of the cloud—customizing deployment pipelines and security protocols. Now, we're seeing the same pattern with AI. The language model becomes just another infrastructure dependency, while the enterprise-owned harness determines how that intelligence is applied. This is where the real value lies.
But here's where things get tricky. Let's talk about cost. Coinbase recently managed to reduce its AI spending by optimizing routing policies and using lower-cost models. But this only works if you own the gateway between developers and the foundation models. Otherwise, you're stuck with vendor lock-in and unpredictable budgets. Walmart and Uber have already run into problems with AI costs spiraling out of control. And according to Stanford research, autonomous coding workflows can consume up to 1,000 times more tokens than interactive sessions. That's not just a technical challenge—it's a financial minefield.
So what does this mean for the average developer? Well, commercial tools like Claude Code and Codex still have a role. They're great for interactive work in editors and terminals. But for asynchronous tasks—like fixing bugs reported in Slack or verifying code against production telemetry—the internal agents are king. It's like having a personal assistant for your coding workflow, but one that understands your company's systems inside and out.
What many people don't realize is that this isn't just about efficiency. It's about control. When you own the harness, you dictate how AI is used across the organization. You can enforce security policies, inject company-specific context, and even experiment with open-weight models without disrupting your team. Smaller companies without dedicated platform engineering teams might still rely on commercial tools, but the trend is clear: the future belongs to those who can master the orchestration layer.
This raises a deeper question: What happens when the AI models themselves become commodities? If enterprises keep their policy and context layers proprietary while open protocols for agent runtimes mature, the competition will shift entirely to the platform layer. Model providers will be fighting to become the preferred 'reasoning engine' inside systems they don't control. That's a seismic shift in the AI landscape.
In my opinion, this is the defining architectural shift of enterprise AI. The strategic advantage isn't in the model itself, but in the platform that governs how, when, and where it's used. And if you're not thinking about this now, you might be left playing catch-up in a few years. The future of software development isn't just smarter AI—it's smarter systems that make AI work for you.
