AI Stopped Being a Feature. It Became the Plumbing.

For a while, AI in your product was a bolt-on. A summary button here, a suggestion box there. That season is ending. A few deals this month show AI sliding underneath the product, into the layer that runs the workflow, not the layer you toggle on. Let me catch you up on what that means for your roadmap.

When the acquisition is the roadmap

Zoom didn't buy Common Room for a feature. It bought a data layer. Common Room built a graph of buying signals across a company's customers, and Zoom is wiring that into its Revenue Accelerator so reps know which accounts are in-market before a call even starts.

The tell is in how the founders framed it. Linda Lian said joining Zoom "connects our graph to the conversations sellers have every day where deals are actually won and to the AI that can act on it." Read that again. The value isn't the dashboard. It's the graph plus the action. That's infrastructure talk, not feature talk.

For you, this reframes a common call. When a capability is really a data layer, building a thin version of it in-house rarely closes the gap. The moat is the data and the plumbing, and that takes years to fill.

Six tools for one job is the problem to solve

F5 is stacking acquisitions to sell AI security as one platform instead of parts. It bought SurePath AI this month, after grabbing CalypsoAI last fall. CEO François Locoh-Donou put the pain plainly: "Having four, five, six different tools to discover, test and secure your AI is a nightmare."

The reason the pain is fresh is speed. As Locoh-Donou noted in the F5 at 30 conversation, "the more an enterprise adopts AI, the less visibility it has into what AI is crawling in the organization." Chief product officer Kunal Anand called it "shadow SaaS with a faster clock and a larger blast radius."

So the product being sold is consolidation. If your customers are cobbling together three tools to do one job around AI, that gap is your roadmap, and someone will fill it whether or not you do.

The AI is the workflow now

KredosAI raised $7M to chase late payments, and the AI isn't a helper on top of collections. It is the collections logic. The software decides the wording, timing, and channel of each overdue message based on account history, then acts through text, email, and AI voice agents.

The results are why it matters. Kredos reports cutting write-offs by 11.5% and lifting customer lifetime value by 13.6% against conventional collections, across more than 200 million interactions in two years. Co-founder Balaji Sridharan built it on a simple read: "The majority of consumers who go late on payment actually want to pay."

Note the design point. The engine explicitly steers clear of off-limits signals like age. When AI runs the workflow, what it ignores is as much a product decision as what it uses.

The deep cut

Watch F5's buy test, because it is the cleanest build-versus-buy rule going. Locoh-Donou said he weighs three things: whether F5 can build the tech fast enough itself, whether the deal serves customers, and whether the team fits the culture. He walked away from companies with great tech and brilliant people because the fit was wrong.

Here's the part that changes your Monday. "Fast enough" is the whole call now. When AI is the plumbing, the question isn't can we build it. It's can we build it before the gap closes on us. If a capability is really a data layer or a consolidation play, a thin in-house version buys you a demo, not a moat. Decide which of your roadmap items are features you can ship and which are infrastructure you're already too late to start from scratch.

Three questions for your team

1. Which items on our roadmap are actually data layers or consolidation plays, not features, and are we honest that we can't build those "fast enough" alone?
2. Where are our customers stitching together three or four tools to do one AI job, and is that stitched-together mess our next product?
3. If AI runs one of our core workflows, what signals have we decided it must not use, and can we defend that choice in a review?