Enterprise AI’s Hottest Job Just Found Its Biggest Skeptic
The fastest-growing job in enterprise AI is the forward-deployed engineer, the specialist companies hire to make the software actually work inside their walls. Decagon, a customer service artificial intelligence startup that crossed $100 million in annualized revenue three years after launching, thinks that job should not exist
Monthly job listings for the role rose more than 800% between January and September 2025, PYMNTS reported. Microsoft committed $2.5 billion and roughly 6,000 engineers, technical consultants and industry specialists to a program embedding technical staff inside client organizations, and Amazon Web Services pledged $1 billion to a similar effort, PYMNTS reported separately. That spending answers a complaint buyers keep making about themselves. Among executives ant companies with at least $1 billion in annual revenue, 71% blamed organizational readiness, not the technology, as the primary barrier to AI performance, PYMNTS Intelligence found. Only 11% blamed the model.
Decagon is betting that consensus is wrong. Long-term reliance on an embedded engineer is evidence the software is too hard to use, not proof that deployment requires one, CEO Jesse Zhang told Newcomer. The contrast he draws most often is with Sierra, the customer-service AI company led by Bret Taylor that has raised three times Decagon’s funding and reached $200 million in annualized revenue, according to equity research firm Sacra
Decagon’s Case Against Forward-Deployed Engineers
Zhang’s case rests on a single customer, whose experience he recounted in Newcomer. That customer spent a year with Sierra’s forward-deployed engineers and built three customer service workflows in that time. Zhang described the arrangement as a black box. Any new workflow, or any deeper look inside the conversations, meant going back through the engineers, who were eventually reassigned to other accounts
After switching to Decagon, Zhang said, the same customer built seven new workflows within about a month. He credits the product, which he said lets a client’s own staff, including non-technical employees, operate it directly instead of routing every change through an embedded engineer
Decagon runs its own forward-deployed engineers, a fact that complicates the pitch. Zhang has said they get deployments live in roughly six week. The company revenue to about $35 million by October 2025 and added more than 100 enterprise customers
The difference, by Decagon’s telling, is what those engineers leave behind. A product built for a client’s own staff to configure directly eventually breaks the link between adding customers and adding implementation headcount. Roughly 90% of Decagon’s workloads now run on fine-tuned open-re faster and cheaper on narrow, repeatable customer-service tasks
Salesforce Is Playing a Different Game
Salesforce is running a similar experiment at a far larger scale. Agentforce reached $1.2 billion in annualized recurring revenue in the first fiscal quarter of 2027, up 205% year over year. In June, the company agreed to buy Fin, the AI agent business formerly known as Intercom, for about $3.6 billion, it said in an announcement
Those numbers are not a like-for-like comparison. Salesforce’s installed base and sales infrastructure give it distribution advantage neither Decagon nor Sierra can replicate so its recurring revenue measures reach as much as product pull. It is a different measure of the same market, not a scoreboard against Decagon’s growth
What the Bet Means for Enterprise AI Budgets
The bigger question is whether enterprise AI can become self-service fast enough to change its labor economics. Forward-deployed engineers exist because today’s artificial intelligence systems still require substantial customization, integration and oversight to work inside large companies. Decagon is betting that this is a temporary stage rather than a permanent feature of the market. If buyers can eventually configure, expand and manage AI systems themselves, vendors can grow revenue without expanding implementation teams at the same rate. If they cannot, the engineer stays on the payroll, and someone keeps paying for software that needed one.
