Rearchitecting the Enterprise Stack

Kelly Sikkema

The old model had a clear shape: software on one side, services on the other, labor in between. Buyers knew what they were buying. Vendors knew what they were selling. The boundaries weren't perfect, but they were stable enough to build entire industries around. Software valuations reflected recurring revenue and switching costs. Services margins reflected labor arbitrage and relationship lock-in. The categories were distinct, even when the lines blurred at the edges.

AI is breaking that structure, and the disruption is not where most people are looking. The current discussion frames this as "software becoming services." A recent essay from Sequoia Capital argues that buyers will increasingly pay to get the job done, not just buy the tool. There's truth in that. But the framing is still incomplete. The bigger shift is that AI is collapsing the distinction between software, services, and labor entirely, which forces a more fundamental question: who owns the workflow?

When intelligence work can increasingly be executed by machines, form factor matters less. Buyers care less about whether capability comes as a SaaS product, an outsourced team, or an internal function, and more about whether the outcome is delivered with acceptable quality, speed, and accountability. That changes where value sits. Sell only a tool, and you're exposed to model-driven commoditization. Sell only labor, and you're holding a cost structure AI is about to compress. Enterprises face the same tension: some work moves back in-house with smaller AI-enabled teams, while other work moves to AI-native providers delivering outcomes directly.

I have been thinking ahead of our meeting of the Executive Technology Board next week, and the agenda discussion on the new AI stack and how to allocate capital - build, buy, partner. The question is not whether AI favors outsourcing or insourcing. It is who controls the workflow once AI can do much more of the execution. The focus has shifted from isolated use cases to which workflows need to be owned, which can be externalized, and how the economics of both are changing. That is a fundamentally different strategic question than "which AI tools should we deploy."

A recent piece from Andreessen Horowitz sharpens this from a capital markets lens: the middle is disappearing. Companies will need to choose. Either drive real growth through AI-native products that own more of the workflow or rebuild for structurally higher margins through AI-driven efficiency. Incrementalism is running out of room. The result is a more polar market. On one end: companies that move up the stack and own outcomes, capturing more of the workflow closer to the business result. On the other: companies that become dramatically leaner, faster, and structurally higher margin. The hardest place to operate is in between, because neither the growth story nor the efficiency story is compelling there.

This also has direct implications for how enterprises should think about vendor relationships. A software vendor who only delivers a tool is increasingly a commodity input. A partner who owns a workflow, encodes your institutional logic into it, and delivers measurable outcomes is something different. The governance question shifts accordingly: you are no longer just evaluating a product. You are evaluating who controls a critical operational capability and what happens to your business if that relationship changes.

Importantly, this does not mean everything becomes autonomous. In domains with complexity, judgment, and risk, the near-term disruption is not removing experts. It is compressing the cost and expanding the capacity of expert-led models. Smaller teams, higher throughput, faster turnaround, different pricing. Human judgment where it matters, machine execution where it does not. That may be the most immediate shift for most enterprises, and it is already visible in how the better-run services businesses are restructuring around it. The headcount math changes before the workflow does.

I see this from multiple vantage points: enterprise transformation governance, AI software builder, large process outsourcer, and the pattern is consistent: when unit economics shift, the boundary of the firm shifts with it. Transaction costs drop, and the logic of what to own versus what to externalize gets rewritten. AI is not the first force to do this. But it is doing it faster, across more domains, and with less warning than any prior shift.

The companies that win will not be the ones that add AI to what they already do. They will be the ones that decide, deliberately, where they sit in the workflow and reorganize around owning it. That decision is strategic, not technical. And in many organizations, it is likely not being made at the right level yet.