A Prediction, One Month Later
This month Apollo confirmed a forecast I made a month ago. It shipped an MCP connector that lets you run its whole B2B prospecting platform from inside Perplexity, Claude, ChatGPT, or Cursor. Then it published the adoption numbers to show people prefer it that way. Tyler Phillips, Apollo's Director of Product Management, shared the case study on LinkedIn, so for once I can check a thesis against something real.
My forecast was that the user interface is collapsing into two surfaces. One is a REST API that people's agents talk to. The sites and dashboards we use today are mostly a barrier, and soon an LLM will sit between the user and the business for everyday work. Apollo just did exactly that. It took its own dense interface, the thing users had to learn, and made it run from a chat window someone else owns. That post was about direction. This one is about a product that already got there.
What Apollo Is, and Why Its Interface Was a Tax
Apollo is a B2B go-to-market platform. Its core asset is a database of over 230 million verified professionals, with search, contact enrichment, email verification, and outreach sequencing on top. Sales teams live in it.
It is also, like every powerful tool, an interface you have to learn. Filters, segments, saved searches, enrichment toggles, sequence builders, tabs on tabs. None of that is the job. The job is "find the right people and send them something relevant." The interface is the tax you pay to get there.
I can speak to this from experience. I've used Apollo on and off for years, and the interface always got in my way. Not because it's badly built. The product is just complex, and no amount of better buttons makes that go away. The power is real, so the surface is dense, and every few months I had to learn it again. Driving it through an LLM, that's gone. I say what I want, and the model handles the menus.
And the work spilled outside Apollo anyway. To write one good email, a rep jumped between LinkedIn, company sites, press releases, and plain search, then went back into Apollo to act on it. Slow, scattered, impossible to scale. It's the Microsoft Office problem from my last post, in sales form: the capability is there, but getting to it costs too much.
The Move: Stop Making People Learn It
Apollo's answer wasn't a redesign. It was an MCP connector, built on the Model Context Protocol, that opens the platform to AI tools. You connect it once with OAuth, no API keys. After that you run Apollo from inside Perplexity, Claude, ChatGPT, or Cursor. The window you already had open becomes the front door.
What that looks like in practice: you type
"find technology buyers in the Midwest with $10M-$50M ARR"
and you get a ready list back, enriched with company and tech signals, without leaving the chat. The same window does the rest in plain language: create or update contacts and accounts, enrich prospects, verify emails, pull fresh company data, add people to sequences, check campaign performance. Everything that used to be a tab is now a sentence.
Nobody opens Apollo to do this. They open the tool they already had open and describe the outcome they want.
The Numbers Say It's Not a Demo
Neat integrations are cheap. Adoption is the real test. Perplexity's Apollo case study puts the result right in its headline:
Apollo users drive 46% more booked meetings with Perplexity's API.
The volume behind that number is what shows people really switched: the case study reports 5.6 million enrichments a month running through Perplexity's Sonar models. Perplexity's Chief Business Officer, Dmitry Shevelenko, put it plainly: "Apollo helps revenue teams move from research to execution faster."
That enrichment figure is the tell. Five and a half million times a month, the easiest way into Apollo's data was a chat window, not Apollo's own screens. People didn't choose chat because it was new. They chose it because it was faster than the interface Apollo spent years polishing.
This Is the Reversal, In Production
Map it onto the two surfaces from last month. Apollo became the API. The agent, Perplexity or Claude or Cursor, owns the conversation and the context: what you want, what you've already pulled, what comes next. Apollo's job shrank to answering cleanly when the agent calls. The user never touches the menus.
The interface didn't get redesigned. It got bypassed, and Apollo leaned into being bypassed. That's the part worth sitting with. Most product teams try to make their UI stickier. Apollo did the opposite. It put its capability somewhere else, in a window it doesn't own, because that's where the user already was. Tyler Phillips, Apollo's Director of Product Management, framed the goal as killing context switching, not decorating it.
The UI Doesn't Die, It Demotes
None of this means Apollo's screens disappear. They become the power-user and debug surface, the way the command line outlived the GUI without either one winning. Someone building a complex sequence, checking why a record changed, or making a dashboard still opens the app. But that's now the exception, not the default. The default front door is a sentence in a chat window.
And the trend only gets stronger, because it runs on the most reliable force in product: people pick the path that costs them less. Learning Apollo costs attention. Typing what you want into a window you already trust costs almost nothing. When a faster path shows up, demand finds it. Apollo is just the first vendor to publish the receipts.
What It Means If You Build Products
The lesson isn't "add a chatbot." Apollo didn't bolt a chat widget onto its app. It exposed its capability through a protocol so any agent could drive it, then let the agents compete to be the front door. You integrate the capability, not the screen.
One question for your own product: can people use it without your UI? If the only way to do the valuable thing is to click through your screens, you're betting users will keep paying the interface tax. Apollo just showed that bet getting worse every quarter.
I wrote last month that the unit of design stops being a screen. Apollo is the first clean example: a real company, with real revenue at stake, that agreed and shipped it. The interface was the barrier. The chat window is the door. The pull toward it is plain convenience, and it isn't going to weaken.
P.S. And yes, this is me circling one idea from a new angle again, exactly as I warned I would.
Originally published: Apollo Moved Into the Chat Window — Alex Rezvov's Blog
