Something recently shifted how I think about our product roadmap, and I think it applies to every SaaS company building right now. We're building the second version of our MCP at PandaDoc so our product works natively inside the leading foundational models. This is not a chatbot or some flash feature to market "talk to your documents." We're making PandaDoc callable by any AI agent, in any environment, so that it can create a document, pull a contract summary, search agreement history, or trigger a signing workflow without ever opening our app. This is a fundamental shift in how SaaS products get used that we're banking on. For 15 years, we've designed software for humans clicking through interfaces. The next wave is software that works for agents acting on behalf of humans. If your product isn't accessible through those agents, you're invisible to a growing share of how work actually gets done. Here's what that looks like in practice at PandaDoc. Our growth team is architecting contract review as a service layer that any surface can invoke: the editor, the signing experience, MCP, a Claude connector, or standalone micro-apps. Our design team published LLM-oriented documentation so any AI agent can generate production-ready prototypes directly from our component library. This isn't a side project. This is how we're designing the product going forward. The way I think about it: the companies that won the last era made great UIs. The companies that win the next era will make great APIs and agent interfaces. The best product won't be the one with the prettiest screen. It'll be the one that works best when no human is looking at a screen at all. We serve 70,000+ customers, most of them are small businesses. These are people wearing five hats at once. If an AI agent can draft their proposal, route it for approval, and collect payment while they focus on their actual work, that's the promise of what we're building. Small businesses are being asked to do more with less, and this is how we help them actually do it. The interface era isn't over, but the agent era has started. We'll be ready for it.
Strong point. The future is how systems connect, not just how they look.
In the era of AI, everything going forward is hyper-personalized. And so the obligation for the builders is towards feeding your tools to be hyper-personalizable. If your APIs aren't publicly callable and publicly documented, LLMs will never reference them. Building a growth engine to be AI-first requires a fundamental shift as you said from a UI first thinking.
Keith Rabkin Strong thesis. The biggest shift is not UI to AI, but workflow software becoming agent-callable infrastructure. Once contract review, approvals, and signing become reliable service layers instead of app-only flows, the product with the best agent interface wins. Great UI will still matter, but the next moat is being the system agents can trust to execute work end to end.
But don’t give up on pretty UI! Human won’t do the bulk of the work but they will still tinker, review, and approve. A few things will stay important, like easy ways to update pricing/docs. But build documents from scratch definitely won’t be.
Build for agents, not just interfaces.
💯agree. Our ethos right now is how do we delete the this UI
We're so ready 💚
This is the shift. From software people use to software agents use on their behalf.