The Solutions Engineer Is Dead. Long Live the Forward-Deployed AI Engineer.

TL;DR#

The traditional Solutions Engineer role — pre-sales SE with a deck, a sandbox demo, and an RFP template — is structurally obsolete at AI-native companies, and the FDE role is what replaces it. Forward-Deployed Engineer postings grew roughly 800% between early 2024 and early 2026 across OpenAI, Anthropic, Cohere, Databricks, and a long tail of YC-backed AI startups, while net SE hiring at those same companies has trended flat-to-negative. The work that closed deals in 2020 — slide-driven discovery, sandbox demos, RFP cycles — doesn't close AI deals in 2026, because buyers want production code in their environment on day one. The FDE is a customer-embedded engineer who ships code and runs continuous discovery on the deployment. The lighter side of the SE job — qualification, scoping, first-touch discovery — is being absorbed by AI agents and AI-moderated conversations like the ones Perspective AI runs.

The Original Solutions Engineer Playbook Worked — Until AI Broke It#

The classic SE playbook was built for a software economy where the product was finished before the buyer arrived. An SE ran a four-step motion: discovery call, tailored deck, sandbox demo with seeded data, then an RFP and security-review cycle. That loop took 8–14 weeks and converted around 25–30% from qualified opportunity to closed-won at most B2B SaaS companies. It worked because the product had a stable surface area.

What's broken is that assumption. AI products in 2026 don't have a stable surface area. The same model writes SQL, summarizes sales calls, drafts a contract, or runs an underwriting interview depending on the system prompt, retrieval layer, and tools wired around it. A demo with seeded data is structurally a lie — a version of the product that doesn't exist in the buyer's stack. The deck doesn't survive contact with the buyer's eval set. The RFP answer about "how the model handles PII" is wrong six weeks later. The four-step SE motion borrowed trust against a stable artifact, and the artifact is no longer stable.

Customers figured this out before vendors did. By late 2024, the standard ask had shifted from "send a demo and a security questionnaire" to "deploy a working pilot in our environment, on our data, in two weeks." Vendors who could do that won.

What the FDE Does That an SE Never Could#

A Forward-Deployed Engineer is a customer-embedded engineer who ships production code into the customer's environment as the act of selling and onboarding. The job: write code in the customer's stack on day one, run continuous discovery on what's working in production, hand back a deployed system the customer's team can extend. The role originated at Palantir — covered in the Palantir forward-deployed engineering playbook every AI lab is now copying — and has been re-formed for the LLM era by OpenAI's customer-embedded engineering team, Anthropic's applied AI engineers, and Cohere's enterprise FDE function.

The mechanical difference is that an FDE produces working software inside the customer's account before the contract is signed. They sit in the customer's Slack, ship to the customer's repo, run evals on the customer's data. The "demo" is the actual deployment. The "discovery call" is watching the customer work for a day. The "RFP answer" is the eval suite they wrote on Tuesday.

The second difference is that FDEs run discovery continuously, not at the front of the funnel. Questions an old SE asked once in week one, the FDE asks weekly against the deployment: which workflows hit the new tool calls, where did the model refuse, which segment is bouncing off the conversational intake. That cadence is what teams run on Perspective AI to capture the qualitative "why" alongside production telemetry — see how forward-deployed engineers run customer discovery in 2026 for the playbook.

The Hiring Data: 800% FDE Growth, Flat SE Hiring at AI-Native Companies#

FDE postings on LinkedIn and YC's Work at a Startup board grew roughly 800% between January 2024 and February 2026, per tracking compiled in the rise of the forward-deployed engineer report. Growth concentrates at AI-native firms — OpenAI, Anthropic, Cohere, Databricks, Scale, Glean, Sierra, Decagon, and roughly 200 YC-backed AI startups — and bleeds into AI divisions of incumbents like Stripe, Notion, and Shopify under titles like "applied AI engineer."

Over the same window, net SE hiring at AI-native companies has been flat to declining. The 2026 AI research stack report found the AI-native cohort posted 3.4x more FDE/applied roles than pre-sales SE roles in 2025, while the non-AI-native cohort posted the inverse. The U.S. Bureau of Labor Statistics' sales engineer occupational data shows the headline category flat year-over-year — but that aggregates growth in legacy hardware sales with contraction inside AI software. Levels.fyi's tracking puts FDE total comp at top AI labs in the $250K–$450K band, materially above SE comp at the same companies for equivalent tenure.

OpenAI rebuilt its enterprise GTM around customer-embedded engineers in its 2025 push. Databricks scaled its "delivery solutions architect" function alongside its data-lakehouse expansion — see the Databricks FDE customer-research strategy. The hiring isn't speculative; it's already happened.

Who Still Needs a Solutions Engineer (And Who Shouldn't Hire One)#

Not every company should drop SEs. Three buyer profiles still need the SE motion: deterministic enterprise software with stable feature surfaces (billing, ERPs, security tooling) where the demo represents the product; regulated procurement environments (federal, large healthcare, large finance) where the process requires a deck-driven RFP cycle; and channel-led businesses where the SE supports a partner motion that can't tolerate vendor engineers in the customer's repo.

The trap is hiring SEs by default for AI products that fit none of those profiles. If you're an AI-native company selling to engineering or product orgs, hiring a traditional pre-sales SE in 2026 is hiring for the deal motion of 2020. The broader funnel shift is in the post-form era report and the conversion gap between forms and conversations.

The lighter parts of the SE job — first-touch qualification, top-of-funnel discovery, intake, scoping — are being absorbed by AI. Buyers don't want a discovery call to answer ten standard questions; they want a conversation that adapts. Teams running AI-moderated discovery on Perspective AI replace the first-call script with an asynchronous conversation that probes the "why" and routes the buyer to the right human at the right moment. That's the half of the SE job getting automated; the other half — code in the customer's stack — becomes the FDE.

Career Path: From Solutions Engineer to FDE in 12 Months#

The FDE role is reachable from where most SEs are, but the transition is real engineering work, not a title change.

Months 1–3: Get production-fluent in one LLM stack. Pick OpenAI, Anthropic, or an open-weights stack. Ship one project end-to-end with retrieval, tool use, and an eval harness. Read Anthropic's applied engineering case studies and the Twilio AI customer-engagement build-out.

Months 4–6: Rebuild your demo as a working integration in your friendliest customer's environment. Replace the deck with a pull request and the discovery doc with an eval suite.

Months 7–9: Run continuous discovery against your deployment. Capture qualitative signal from users. Patterns are in the continuous discovery habits guide and the always-on research playbook. Use a template like the user research interview to standardize cadence.

Months 10–12: Move internally or interview externally. AI-native companies hire FDEs from SE backgrounds — see why every AI startup needs a forward-deployed engineering function and the founder playbook. Three deployed projects, three eval suites, three customer-discovery write-ups is the new resume.

What Gets Lost in the Transition (And How to Keep It)#

The honest cost of the SE-to-FDE shift is real.

First-touch discovery becomes uneven. A good SE could map a stranger's org chart, pain, and budget cycle in 45 minutes. An FDE is worse at this on day one. The fix is structured asynchronous discovery in front of the engagement — the gap teams close with AI-moderated customer interviews and conversational intake.

Procurement still needs translation. Enterprise procurement runs on RFPs, SOC 2 questionnaires, and security reviews. FDEs hate this work. The answer is a small technical-account-management function that handles procurement artifacts on behalf of FDEs.

The trusted-advisor relationship has to be rebuilt. SEs at their best were lifetime accounts. FDEs are often project-scoped. Companies that get this right re-create the trusted-advisor layer with named FDEs on top accounts plus continuous discovery on top of the deployment.

Frequently Asked Questions#

What is the FDE role and how is it different from a Solutions Engineer?#

The FDE role is a customer-embedded engineer who ships production code into a customer's environment as the act of selling and onboarding, rather than running demos and decks. FDEs sit in the customer's stack, write code in their repo, and run continuous discovery against the deployment. The SE role is pre-sales and demo-driven; the FDE role is in-sales and code-driven. At AI-native companies, the FDE replaces 60–80% of what the pre-sales SE used to do.

Is the Solutions Engineer role really dead, or is this hype?#

The traditional pre-sales SE role is structurally obsolete at AI-native companies but remains alive at companies selling stable-surface enterprise software, in regulated procurement, and in channel-led businesses. The data is clear: roughly 800% growth in FDE postings between 2024 and 2026, with flat-to-declining net SE hiring at the same firms. "Dead" fits the AI cohort; "transformed" fits the rest of the market.

What is the difference between an applied AI engineer and a Solutions Engineer?#

An applied AI engineer is a customer-embedded software engineer who ships LLM systems into a customer's environment; a Solutions Engineer is a pre-sales role focused on demos, decks, and RFP responses. The applied AI engineer title is most common at Anthropic; "forward-deployed engineer" is the equivalent at OpenAI, Palantir-derived teams, and most AI startups. Both produce working software and stay engaged after the deal closes.

How much do Forward-Deployed Engineers make compared to Solutions Engineers?#

FDE total comp at top AI labs sits in the $250K–$450K band, materially above SE comp at the same companies for equivalent tenure. The gap reflects that FDEs do work the company can't ship without — production code in customer environments — while SE roles at AI-native firms have been absorbed by FDEs and AI agents handling discovery. Levels.fyi and Glassdoor publish current benchmarks.

Can a Solutions Engineer transition to a Forward-Deployed Engineer role?#

A Solutions Engineer can transition to an FDE role in roughly 12 months by building engineering reps on one LLM stack, rebuilding the demo motion as a working integration in a real customer environment, running continuous discovery against that deployment, and presenting three deployed projects plus eval suites as a portfolio. The transition requires real production code; a title change alone does not work.

Where does customer discovery fit in the FDE workflow?#

Customer discovery runs continuously, not just at the front of the funnel. FDEs run it at three layers: before the engagement (async AI-moderated interviews to brief the engineer), during it (weekly conversations with end-users), and after launch (always-on conversations that catch drift). The pattern replaces one-time discovery calls with ongoing conversational research that captures the "why" behind production telemetry.

The Manifesto#

If you're a founder still hiring pre-sales SEs into your AI-native company, stop. You're hiring for the deal motion of 2020 into a market that wants 2026 deliverables. If you're a Solutions Engineer reading this with the hair on the back of your neck up, that's the right signal — the role is being unbundled fast, and the half that survives is the half you can convert into FDE work in the next 12 months. Pick the stack. Ship the integration. Run the discovery loop.

What survives of the old SE function lives in two places: continuous AI-moderated customer discovery — the "why behind the why" layer Perspective AI was built to run — and customer-embedded engineering inside the deployment. Everything between is the dead middle. Long live the Forward-Deployed AI Engineer, because the artifact at the center of the FDE motion is the deployment itself. Start a research project on Perspective AI or see how teams use the platform.