Decagon Hit $100 Million Betting Against Forward Deployed Engineers: What Jesse Zhang's Bold Stance Actually Means for Your Career
Decagon Hit $100 Million Betting Against Forward Deployed Engineers: What Jesse Zhang's Bold Stance Actually Means for Your Career
By Richard Chen and James Won | TechCareers.io
The Headline That Stopped the FDE World
Two weeks ago, a newsletter headline landed that sent a shockwave through every forward deployed engineering community, Slack group, and LinkedIn feed in the enterprise AI world.
"Decagon Hit $100 Million Betting Against Forward Deployed Engineers."
The immediate reaction from most FDE candidates and professionals was some version of the same two questions.
Is this a threat to my career?
And what does this actually mean?
Both are the right questions. The answers are more nuanced, and more useful for your career, than the provocative headline suggests.
This blog gives you those answers. Not the surface-level reaction. The actual strategic implications for anyone who is building toward a forward deployed engineering career in 2026.
What Jesse Zhang Actually Said: The Precise Argument
Before reacting to the headline, understand exactly what Decagon's CEO argued and what he did not argue.
Jesse Zhang did not say that the work of forward deployed engineering is unnecessary. He did not say that the problems FDEs solve are not real. He did not say that enterprise AI deploys itself.
What he said was specific and surgical.
Long-term reliance on an embedded engineer is evidence the software is too hard to use, not proof that deployment requires one.
That is the precise claim. And the key word is long-term. Zhang's argument is not that you do not need technical expertise to get AI running inside a complex enterprise environment. His argument is that if the product is strong enough, clients should be able to operate and evolve it themselves after the initial deployment, without a permanent embedded engineer who becomes an unavoidable bottleneck for every change and every new workflow.
He backed this argument with a specific competitor comparison. Sierra, the customer service AI company led by Bret Taylor and backed by three times Decagon's funding, uses traditional forward deployed engineers. Zhang described a specific customer who spent a year with Sierra's forward deployed engineers and built three customer service workflows in that time. Every new workflow, every deeper look inside the conversations, required going back through the engineers. The system had become a black box that the client could not see into or evolve independently.
Zhang's case: that black box is a product failure disguised as a service offering. Decagon's product is designed so that clients can customize their own agents, understand what they built, and evolve it without routing everything through an engineer. The Agent Operating Procedures system, AOPs, exists specifically to give non-engineers the ability to configure agent behavior in plain English.
The $100 million in annualized revenue is Zhang's proof that this bet is working.
Why Zhang Is Right About One Thing, and What He is Not Saying About Another
Zhang is making a genuinely important point that every FDE candidate needs to understand. And he is also not making a point that the headline implies.
Where Zhang is right:
The traditional FDE model has a real scaling problem. If every change to every workflow requires an embedded engineer, the product has a structural dependency that limits how fast clients can move. For a company whose entire competitive pitch is speed, going live faster, iterating faster, hill climbing quality faster, that dependency is incompatible with the core value proposition.
Decagon's Agent Development model, the ADM who owns the customer relationship, the ADE who owns the technical build, is specifically designed to front-load the engineering work in a way that leaves clients able to operate independently afterward. The ADE builds systems that clients can understand and own a year later, rather than black boxes that require ongoing engineering intervention. That cuts the custom engineering work per agent by 80 percent relative to where Decagon started.
This is not a rejection of technical deployment expertise. It is a rejection of technical dependency as a permanent feature of the client relationship.
Where the headline overstates:
The PYMNTS summary of Zhang's argument is pointed: Decagon thinks the forward deployed engineer job should not exist. But this overstates Zhang's actual position when you read the full context.
Decagon has a CTO of Forward Deployed Engineering. They have an Agent Development team of nearly 100 people that includes Agent Deployment Engineers who write Python, build integrations, and own the technical agent build. They presented an 18-minute masterclass on how their forward-deployed motion has evolved at an a16z event in late July 2026.
Despite the headline, Decagon is not running its enterprise business without technical people embedded with clients. They are running it with a specific and deliberate model of technical embedding that is designed to compound over time rather than create permanent dependencies.
The distinction: Zhang is arguing against the permanent black-box FDE model where clients can never operate independently. He is not arguing that you can deploy enterprise AI without technical expertise and client-facing engineering work.
For FDE candidates, this distinction is everything.
What This Actually Means for the FDE Market in 2026
Here is the market-level implication that most people who read the headline miss because they stop at the headline.
The Decagon story is not evidence that FDE demand is declining. It is evidence that FDE demand is evolving, and that the professionals who understand the evolution will be dramatically more valuable than those who do not.
Look at what is happening across the market simultaneously.
Microsoft committed $2.5 billion and roughly 6,000 engineers, technical consultants, and industry specialists to a program embedding technical staff inside client organizations.
Amazon Web Services pledged $1 billion to a similar effort.
Salesforce publicly committed to building a team of 1,000 forward deployed engineers to drive Agentforce adoption.
Anthropic committed $100 million to its Claude Partner Network and said it would expand its partner-facing team fivefold, including Applied AI engineers supporting live customer deployments.
OpenAI acquired a deployment-focused startup that brought approximately 150 forward deployed engineers into the organization overnight.
FDE job postings grew more than 800 percent year over year through early 2026. More than 1,000 percent through some measurements. Palantir has 95 open FDE-family roles as of July 2026. The entire enterprise AI deployment industry is scaling forward deployed engineering teams simultaneously.
The Decagon story does not contradict this picture. It adds a specific and important dimension to it. The FDE model is not going away. The FDE model is being refined. And the companies that will win the enterprise AI deployment market over the next three to five years are the ones that figure out how to combine technical deployment expertise with product design that leaves clients empowered rather than dependent.
The forward deployed engineer who understands this evolution, who can build systems that compound and transfer rather than create permanent dependencies, is the one the market is aggressively looking for right now.
The Competitive Landscape That Explains Everything
To understand what Decagon's stance actually means, you need to understand the competitive environment it is being made in.
Decagon is at $100 million in annualized revenue and a $4.5 billion valuation, extraordinary numbers for a company less than three years old, but not first place in the enterprise AI customer support market.
Sierra, led by Bret Taylor, former co-CEO of Salesforce, now chair of OpenAI's board, and Clay Bavor, 18 years at Google, has surpassed $200 million in annualized revenue and $150 million ARR as of February 2026, with more than 40 percent of Fortune 50 as customers. Sierra uses traditional forward deployed engineers embedded with clients. Sierra's funding is three times Decagon's.
Salesforce Agentforce surpassed $1 billion in ARR and then doubled down by agreeing to buy Fin, the customer agent company formerly known as Intercom, for $3.6 billion in June 2026. Salesforce has committed to 1,000 FDEs for Agentforce adoption.
These are not small companies failing with the FDE model. These are the largest and best-funded players in the market using the traditional FDE approach and winning significant enterprise contracts with it.
Zhang's anti-FDE stance is not a neutral market observation. It is a competitive positioning move. Decagon is smaller than Sierra. It has less funding. It does not have Bret Taylor's Valley credibility or Salesforce's existing enterprise relationships. The argument that Decagon's product is so good that it does not need permanent FDE embedding is also the argument that justifies why clients should choose Decagon over Sierra even though Sierra has more engineers, more funding, and more market presence.
Understanding this context does not mean Zhang is wrong. It means the argument should be evaluated as both a genuine product philosophy and a competitive positioning strategy, not purely as a neutral market signal about the future of the FDE role.
The Three Career Implications That Actually Matter
Here is the practical read for a technical professional who is evaluating the FDE career path in light of this story.
Implication One: Systems thinking is now the differentiating FDE skill.
Whether you are targeting Palantir, Decagon, OpenAI, Anthropic, Salesforce, or any other company in the FDE ecosystem, the candidate who can build systems that compound, frameworks that make the next deployment faster, solutions that clients can operate and evolve independently, playbooks that generalize beyond the individual engagement, is dramatically more valuable than the candidate who can heroically solve one client's problem one at a time.
Katherine Xiao, Decagon's Director of Implementations, was explicit about this shift. The failure mode she described, agent builds that decay into a black box of prompts and patches no one can safely touch, is real and it is happening across the enterprise AI deployment market. The FDE who builds that black box is the FDE whose skill set Zhang is arguing against. The FDE who builds systems that clients can own a year later is the FDE who the entire market is desperately trying to hire.
This is not a Decagon-specific insight. It is a market-wide evolution. The AIP Bootcamp model at Palantir is the same instinct applied differently, build something working quickly and then ensure the client's team can operate and extend it, rather than creating a permanent dependency on Palantir FDE bandwidth.
Implication Two: Product understanding is now a required FDE competency.
Zhang's critique of the permanent FDE model is also a critique of FDEs who build custom solutions without pushing back on the product gaps that made the custom solutions necessary. Every piece of bespoke work an FDE does for a client that could have been self-serve if the product were better is both a short-term client win and a long-term product failure.
The FDE who understands this, who actively identifies the recurring custom work patterns, brings them to the product team, and helps turn one-time bespoke solutions into self-serve platform capabilities, is the FDE who is making the company more competitive. Decagon explicitly measures this: ADEs compound the value of every deployment by feeding learnings back to product and engineering.
At Palantir, the FDE has always been expected to feed insights from the field back to the platform team. The companies that are scaling FDE functions in 2026 are increasingly building this feedback loop as a formal expectation rather than an informal practice.
Implication Three: The Decagon model may be the template for what FDE looks like at scale.
Every company that is building an FDE function right now faces the same scaling challenge that Decagon faced. You can brute-force the first ten clients. You cannot brute-force the first thousand. The companies that figure out how to deliver the value of technical deployment expertise at scale, without permanent engineering dependency on every client, are the ones whose FDE functions become sustainable competitive advantages rather than cost centers that grow linearly with revenue.
Decagon's Agent Development model, the ADM and ADE pairing, the systematic conversion of bespoke work into platform capabilities, the focus on leaving clients able to operate independently, is one answer to that scaling challenge. It is not the only answer. But it is a genuinely useful framework for thinking about what sustainable FDE work looks like at scale.
The FDE candidate who understands this framework is the one who can speak intelligently about the future of the role in an interview, at Decagon, at Palantir, or at any other company that is building a forward deployed function in 2026.
What This Changes About How You Should Prepare
If you are actively preparing for FDE interviews, the Decagon story adds one specific preparation element that was not obvious before this headline.
Every FDE interview you have, at Palantir, at Decagon, at Scale AI, at Rippling, should now include a story about systems thinking and knowledge transfer. Not just a story about what you built. A story about how you built something in a way that made it possible for someone else to maintain, extend, and own it after you were done.
The FDE candidate who has a specific example of turning a one-time client solution into a reusable framework, a playbook, a documented pattern, a training session that left the client's team genuinely capable rather than dependent, is the candidate who demonstrates the most valuable FDE capability in 2026.
This is the capability that Zhang is arguing the market should demand. It is the capability that Decagon's ADE role is specifically designed to develop. And it is the capability that Palantir's most senior FDEs have always been evaluated on, the transition from fixing one account to building leverage for every account.
Practice telling that story before you walk into any FDE interview this year.
The Honest Bottom Line
The Decagon headline is provocative. The underlying argument is important and worth taking seriously. And the career implications are genuinely useful for anyone building toward this kind of role.
But the headline does not mean what it sounds like it means if you read it without context.
Decagon is not running without technical people embedded with clients. They are running with a more deliberate and systematized version of technical embedding. The enterprise AI deployment market as a whole is scaling FDE functions, not eliminating them. And the FDE who understands the evolution, systems thinking, product feedback loops, knowledge transfer, sustainable delivery models, is more valuable in this market than the one who does not.
That is the career opportunity inside the provocation.
Schedule Your Forward Deployed Engineering Strategy Session
If you want to understand how the evolution of the FDE market affects your specific preparation and career positioning, and what the fastest path looks like to the roles in this market that are most aligned with your background, book a strategy session with our team.
Book your session here: https://consultation.techcareers.io/o-discover-fde/about