The first wave of enterprise AI was characterised by point solutions: a product that used machine learning to automate one specific task within one specific function. Invoice processing. Candidate screening. Churn prediction. Quality control image analysis. These products were genuinely useful — the best of them saved tens of thousands of hours of manual labour annually at their customers — but they were fundamentally bounded by the scope of the problem they were solving. A company that deployed ten such tools had ten disconnected point solutions, each with its own vendor relationship, data silo, and integration surface. The productivity gain was real but the architectural complexity it introduced was also real.
What we are now seeing — and what we are specifically thinking about for Fund II deployment — is a second wave characterised by platform dynamics: AI-native products that started as point solutions but are expanding into the workflow orchestration layer that connects multiple process steps, and in some cases multiple functions. This expansion is not primarily driven by sales motion, though it has obvious revenue implications. It is driven by a product logic: once you have built the data pipelines and integration infrastructure to automate one step in a business process, the marginal cost of extending to the adjacent steps in the same process is much lower than the original build cost. The companies that understand this are deliberately designing their initial point solution to be the insertion point for a broader workflow platform, not a standalone tool.
Taktile represents this pattern in financial services operations: the initial product is decision automation for credit risk and fraud detection at financial operations teams, but the platform logic extends to any decision process in a financial services operation that can be modelled as a structured decision tree with data inputs and policy constraints. Flip in employee communications started as a messaging automation product but the platform logic extends to the entire internal communications and workflow surface for frontline workers. These are very different domains but the architectural pattern is identical: build the point of maximum pain depth first, earn the integration trust and data access that comes with a successful initial deployment, then expand the workflow surface.
The transition from point solution to platform is where most AI-native B2B companies encounter their most significant strategic risk. The expansion requires a product investment that is large relative to seed-stage resources, a sales motion that needs to shift from product-led to enterprise sales, and a technical architecture that was designed for a narrow use case now needs to handle the complexity of an open platform. We have seen companies get this transition wrong in both directions: some expand too early, before the initial point solution has achieved the kind of customer depth that makes the platform extension credible; others stay too narrow too long, allowing a platform competitor to capture the expanded workflow surface while they are still defending the original point solution.
Our working thesis for the second wave is that the companies most likely to win are those where the founding team has a clear, explicit view of which workflow surface they are trying to own by year five, and is making every product and go-to-market decision in year one through the lens of whether it positions them correctly for that expansion. This requires a kind of strategic patience that is genuinely difficult to maintain under the growth pressure of early-stage company building. The temptation to optimise for the metric that secures the next funding round rather than the architecture that enables the eventual platform position is real and understandable. But the companies that have clarity on their five-year platform destination and are executing a deliberate path toward it are the ones we are most excited about in our current pipeline.