New Relic: Forward Deployed Engineers Are Becoming Key to AI Agent Success
New Relic says forward deployed engineers (FDEs) are increasingly the people who make sure AI agents deliver in production. In a blog post, the company cites a Gartner forecast that more than half of AI tech and service providers will use FDEs by 2028, up from 10% in March 2026.
- Gartner predicts more than half of AI tech and service providers will use FDEs by 2028, up from 10% in March 2026.
- New Relic formalized its own FDE function for customers adopting AI and AI observability capabilities.
- New Relic reports 67% of leaders say AI writes most of their code weekly, and 78% report more production incidents.
- One in four organizations runs AI agents in production with no observability, according to New Relic.
- Gartner predicts 70% of software engineering teams will adopt eval-driven development by 2028, up from under 20% in 2026.
Why FDEs are gaining ground
The post, written by Stela Udovicic, Senior Director of Solutions Product Marketing at New Relic, draws on a conversation at the company's New Relic Now event. Camden Swita, who leads AI product management at New Relic, spoke with Kevin Bai from Anthropic's applied AI team. Bai helped build the FDE function at Rippling before joining Anthropic.
According to New Relic, FDEs create the most value when a highly technical platform meets a customer whose expertise lies in running their business. Customer engineers are often committed to other priorities, while new platform capabilities arrive quickly. An FDE pairs deep platform knowledge with the customer's business knowledge and carries much of the complexity. New Relic says it recently formalized its own FDE function, aimed at customers adopting its newer AI and AI observability capabilities.
The post also quotes the 2026 Gartner The Forward Deployed Engineer Handbook: "Forward deployed engineers bridge technical execution and business operations by embedding in customer teams to solve challenging technical problems and translate product capabilities into measurable client outcomes."
Context and evals for agents
New Relic argues that agents should be onboarded like a new hire. Models keep getting more powerful, but they still need specific knowledge of how a company works. Connectors are a big part of that, and the post points to the Model Context Protocol (MCP), an open standard introduced by Anthropic for connecting models to systems of record. Its rule of thumb: if a human teammate would need access to telemetry to finish a task, the agent needs it too.
The post gives two examples. An IT operations leader at a large benefits administration company said early bots read IT tickets but missed telemetry. A ticket only captures what someone reported, while telemetry shows what happened across systems. Customer Turo described an agent that checks New Relic when an alert fires, so a root cause analysis is waiting before engineers are online. In one midnight incident, according to the post, it found the root cause within a minute.
On evaluation, Bai is quoted as saying, "You cannot understand what you cannot measure, and you definitely can’t improve something that you can’t benchmark." The advice is to define what the agent is for, work backward to measurable factors such as tone, timeliness and shared resources, and keep running evals as applications, connectors and models change. Gartner predicts that by 2028, 70% of software engineering teams will adopt an eval-driven development approach, up from less than 20% in 2026.
Trust, observability and new products
New Relic cites figures from its 2026 State of AI Coding and State of Observability Forecast reports. According to the company, 67% of leaders say AI now writes most of their code in a typical week, and 78% report more production incidents. One in four organizations runs AI agents in production with no observability at all. Organizations that actively observe their agents are twice as likely to see a 3x return on their observability investment.
Chief Product Officer Brian Emerson described reliability as a spectrum from agentic coding through performance health, user experience, business impact, and cost and ROI. In the keynote demo, a refund agent made up its own refund policy. The team traced the problem to a single prompt change and built a custom evaluator from production data.
Products launched at New Relic Now include:
- AI evaluations, prompt ops, experiments and topic discovery: these bring evals into the same data model as other telemetry, with OpenTelemetry as a first-class citizen.
- New Relic Ground Truth: delivers New Relic intelligence into coding tools and customer-built agents, focused on accurate, fast and token-efficient answers.
- New Relic Autopilot: a set of SRE agents that investigate issues using unified context.
What it means for enterprise teams
The post's message to engineering and operations leaders is that AI strategy should be intentional across people, processes and agents, and that there is no universal answer on how to adopt it. Kevin and Camden compared the moment to the early days of mobile, when companies chose between web, mobile or both. The common question, according to New Relic, is how people use a platform today and how they should use it to succeed tomorrow.
Whether FDEs come from a vendor or an internal team, New Relic says what they build must keep running, improving and proving its value after the engagement ends. That requires visibility into every agent, the systems it touches, every prompt and every token spent. Note that the post is vendor content and the survey figures are New Relic's own.
New Relic: Do You Need Forward Deployed Engineers? 5 Signals from New Relic Now
