Atlassian Says Its FDEs Aren't a Services Business. Put the Field Under the CPO.

Mike Cannon-Brookes drew a line on Atlassian's earnings call this week that every CPO should copy.

Matt Hedberg of RBC asked about the company's forward deployed engineer push. Cannon-Brookes said "all FDEs are not built the same." Atlassian's are AI and ML engineers "writing full production customer code," not "rebadged customer success staff." Then he said the quiet part plainly: "Our goal is not building a service business." And the reason the program is working, in his words from the transcript: "we are seeing great speedups in our ability to build product with FDEs," because it "helps us speed up the customer request to delivery."

Customer request to delivery. That's the number. And it's a product number.

The same week, two other companies gave their field a different home. HubSpot's Yamini Rangan said on the Q3 call that HubSpot brought sales and customer success under one organization for the upmarket segment, to smooth the handoff from trial to expansion, and that Customer Agent resolved over 2 million support issues in the quarter. On Thursday Genesys expanded its forward deployed engineering, and as Dom Nicastro reports at CMSWire, it sits alongside more than 2,000 CX professionals across advisory, deployment, and managed services.

Product, revenue, services. Each is a sensible home for something. Only one is built to shorten the time between what a customer said and what the product does.

This is part 7 of The CPO Transition, part of The Product OS. Part 6 argued that product, design, product marketing, and FDE report to the CPO because one loop needs one owner. This part takes the customer end of that loop.

The short version

• The field is the fastest product signal a company has, and in the AI era it reports to the CPO.
• Atlassian said this week its FDEs write production code at the customer, aren't a services business, and speed up customer request to delivery. That's the right shape.
• The same logic covers customer success, which AI splits in two: the commercial half, renewal and expansion, stays with the CRO, and the adoption half, onboarding, the support agent and its transcripts, and the engineers who fix deployments, moves under the CPO next to the FDEs.
• When a support agent resolves 2 million issues a quarter, as HubSpot's did, support is a product surface, and the conversations it can't resolve are the most honest spec backlog the company has.
• The test is a clock: hours from a field finding to an eval row, days to a fix. And because taking knowledge back from customers is the point, the terms for it are a product policy the CPO sets once, not a deal term.

The reporting line decides who reads the signal

An engineer inside a customer's workflow learns three things every week. A fix for this customer. A pattern across customers. And a fact about what the product should be.

The reporting line doesn't change what the engineer sees. It changes who reads it. Under services, the fix becomes a billable practice. Under revenue, the pattern becomes a pitch. Under product, the pattern becomes a spec, and the fix becomes a feature every customer gets. Part 15 of The FDE Transition made that case for the FDE org in full, so I won't restate it.

What's new this week is Cannon-Brookes's other line: Atlassian is doing "more and more core software engineering at the customer site." Read that as a product leader. When the build is cheap enough to happen at the customer, the customer site is the front of the build pipeline. And the front of the build pipeline reports to whoever owns the build decision. That's the CPO.

Customer success splits in two

The harder question is customer success, because it holds two jobs that used to fit in one seat.

One is commercial: the renewal, the expansion, the relationship with the buyer. HubSpot's move puts that with the people who carry the expansion number, and I think that's right. The other is adoption: getting the customer to the outcome, reading where they get stuck, and fixing it. Rangan touched that one too. She said trial results showed customers needed hands-on help to turn AI trials into paid adoption, and that it changed the trial experience. That's an adoption finding that became a product change.

AI pulls those two jobs apart. When an agent resolves 2 million support issues a quarter, support isn't a queue anymore. It's a product surface with its own prompts, its own knowledge, and its own evals, and every conversation is a signal. The conversations it can't resolve are the most honest spec backlog a company has, written by customers in their own words. Whoever owns that agent owns a real piece of the product. The knowledge it reads is the same living file The Product Marketing Transition argues product marketing should version and test.

2 million

So here's where I draw the line. The commercial half of customer success stays with the CRO. The adoption half, onboarding, the support agent and its transcripts, and the engineers who fix live deployments, reports to the CPO, next to the FDEs. The month-two handoff from FDE to CSM still happens. It just happens inside one org, against one outcome number.

Hours, not quarters

What changes when the field sits under the CPO isn't the chart. It's the clock.

Count the hops a field finding takes today. The engineer writes it up. A manager forwards it. It lands in a channel, the way Salesforce routes field feedback through a tagged Slack channel, as part 6 noted. A PM matches it to a theme. It goes to planning. Every hop that crosses a reporting line is a wait, and at AI build speed the wait is now longer than the build.

Put the field engineer, the PM, the designer, and whoever owns the support agent in one org, and a finding can become an eval row the same day and a fix the same week. That's what "customer request to delivery" means when a working prototype costs an afternoon.

Measure it. Tag every field-originated request by source, FDE, customer success, or the support agent, and report the median time from first heard to shipped. The method is the same one I use for signal-to-ship cycle time, with the field as its own lane.

The customer has to agree to the loop

One more thing makes this a CPO decision rather than an ops one.

In The Register, Lindsay Clark quoted Gartner's Mukul Saha on what vendors' FDEs do: "They're going to take back certain knowledge that makes the product better." His advice to buyers: "figure out the exit strategy." He's right on both counts. Taking knowledge back is the whole point of a field that reports to product, and buyers are going to ask exactly what comes back and on what terms.

That answer is a product policy. What generalizes into the product, what stays the customer's, how a pattern gets anonymized before it becomes a feature. Decide it once, publish it, and put it in the contract, the way the buyer's contract lays out. Left to a services or sales scorecard, it gets decided deal by deal, and the loop either stalls or loses trust.

What I expect

These are my predictions, not facts.

My prediction: by the end of 2027, some of the companies that talk about FDEs on earnings calls report a product number beside engagements, a version of Atlassian's customer request to delivery.

My prediction: within two years, customer success splits at AI-native companies. The commercial half keeps the name and reports to the CRO. The adoption half, with the support agent, moves under the CPO.

My prediction: by 2028, the support agent's unresolved-conversation queue is the most-read backlog in the product org, read weekly beside the repair rate that The Design Transition argues design should own.

Next in the series: the story is part of the product, product marketing under the CPO, and positioning as a living spec agents read.

One thing to try this week

Pull the last ten customer requests that came from the field and shipped. FDEs, customer success, support, any of them. For each, write three things: the date you first heard it, the date it shipped, and how many reporting lines it crossed on the way. Take the median gap and the median hop count. If the gap is measured in months and the hops are three or more, you've found the reorg conversation, and it fits on one page for your CEO.

Sources: Atlassian, Q1 fiscal 2027 earnings call, Mike Cannon-Brookes answering Matt Hedberg of RBC, automated transcript via TradingKey, November 6, 2026, quotes as transcribed. HubSpot, Q3 2026 earnings call, Yamini Rangan, transcript via The Motley Fool, November 5, 2026. Dom Nicastro, "Genesys Scales Forward-Deployed Engineers to Speed Agentic CX," CMSWire, November 5, 2026. Lindsay Clark, "Vendors asking forward deployed engineers to make good on trillion-dollar AI bet," The Register, October 22, 2026, with Mukul Saha of Gartner.

Part of the running argument on Product Leadership: the product org chart was calibrated to a build cost that collapsed, and in the AI era the whole product outcome, from customer signal to adoption, belongs to one function under the CPO.

Related answer: Should customer success report to the CPO?