Read Intercom's surfaces and the pivot is obvious

Pick any company you compete with and open four tabs: pricing, changelog, careers, and the homepage. Don't read the blog. Don't wait for the keynote. Just read the plumbing.

Intercom is a good specimen because the plumbing changed loudly. A few years ago it was a customer-messaging tool that happened to have a support inbox. Read its public surfaces today and you're looking at a company that reorganized itself around a single bet — AI-handled resolutions — and left the evidence sitting in plain view. Here's the walk.

The pricing page stopped counting seats

The tell is the unit of measure.

Classic SaaS pricing counts seats: per agent, per month, tiered by feature access. When a pricing page starts charging for outcomes instead of access, the company has decided its value is no longer "software your team logs into" — it's "work the software does without your team." Intercom's move toward charging per resolution — a support conversation closed by its AI agent rather than a human — is that shift made explicit. You're no longer buying chairs for people. You're buying results the product produces on its own.

That single change reframes everything downstream. It means the roadmap is pointed at deflection, not collaboration features. It means the sales motion has to justify a variable, usage-based line item to a finance team. It means the competitive threat isn't "a cheaper inbox" — it's "a vendor whose price scales with how much work they take off your plate." None of that was announced on the pricing page. It was structured into it. A rival PMM who only reads the feature comparison chart misses the whole story; the pricing model itself is the story.

The changelog changed genre

A changelog is a company narrating itself in real time, and the genre it's written in tells you what it thinks it's building.

For years Intercom's changelog read like a product-ops diary: new inbox views, workflow tweaks, reporting filters. Read a recent stretch and the genre has shifted. Entries cluster around one protagonist — the AI agent — and its supporting cast: answer quality, guardrails, hand-off logic, analytics for how often the agent resolves versus escalates. The verbs changed too. Less "you can now configure," more "the agent now handles."

When the ratio of a changelog tilts this hard toward one surface area, that's a resourcing signal you can read without any internal access. Engineering time is finite. A changelog dominated by one theme means the org chart is dominated by it too. This is the same move we walked through in what Stripe broadcasts without announcing it — the release notes are a public map of where the headcount went.

The careers page names the bet out loud

If pricing and changelog imply the strategy, careers confirms it in the plainest language a company ever uses about itself.

Job descriptions are written to attract the exact people who will build the next twelve months of product, which makes them unusually honest. Scan Intercom's openings and the pattern isn't subtle: roles built around applied AI, agent evaluation, machine-learning infrastructure, and the go-to-market people who sell outcome-based products. A support-tooling company staffing up on eval engineers and AI product managers is not planning to remain a support-tooling company.

The useful trick is to read careers against the changelog. If the changelog is shipping AI features and the careers page is hiring the people to ship more of them, you have two independent surfaces agreeing. Two surfaces telling the same story is a much stronger signal than either one alone — it's the difference between a marketing claim and a commitment they're spending payroll on. We pulled the same thread in Figma's roadmap hiding in its Community tab: the roadmap you trust is the one three separate surfaces confirm.

What the teardown is really teaching

Intercom is just the worked example. The method is the point.

No single page told the whole story. The pricing model gave you the shape of the bet, the changelog gave you the pace, and the careers page gave you the commitment. Read in isolation, each is a data point. Read together, on the same afternoon, they converge into a thesis you could brief your exec team on — before Intercom itself framed it that way in a launch.

The catch is that surfaces don't change on your schedule. A pricing page gets quietly restructured on a random Tuesday. A changelog's genre shifts one entry at a time, so gradually that no single visit registers it. The convergence is only visible if you're comparing today's version against a version you remember — and human memory is a terrible diff tool. This is the narrow, honest thing Seeto does: it watches public surfaces like these continuously and surfaces the changes as discrete events, so a pricing-model rewrite or a careers-page tilt lands as a dated entry instead of something you half-notice six months late. It won't write the thesis for you — that's still your read. It just makes sure you're reading the current page, and that the quiet Tuesday edit didn't slip past. For which changes actually deserve that kind of flag, see the competitor changes worth an alert.

Pick your own competitor. Open the four tabs. The pivot is usually already on the page — someone just has to read it before the press release does.