By FDE Interview Prep, a SourceStrongAI company

In this post11 sections
  1. Pre-sales or post-sale FDE: what changes
  2. Do forward deployed engineers do pre-sales? What postings say
  3. Where it stops being FDE work
  4. Is a pre-sales FDE just a sales engineer?
  5. How FDE interviews test pre-sales skill
  6. A worked scoping answer for an undecided prospect
  7. Mistakes that sink pre-sales rounds
  8. Practice the call, not just the answers
  9. Questions people ask
  10. Keep reading
  11. More from the blog

You found an posting that says “pre-sales” and now wonder if it is a sales job with an engineer’s title. It isn’t, if you also build what gets sold: some forward deployed engineer (FDE) roles join after the contract is signed, and some help get it signed, like answering an account executive’s “can you join Thursday’s call and show them it works on their data?” For how the whole FDE job compares with sales engineering, read the FDE vs sales engineer guide.

The short answer: yes, some FDE roles include pre-sales work. OpenAI’s financial services FDE posting in New York asks the FDE to lead discovery and scoping “from pre-sales through post-production”. Source 1Forward Deployed Engineer (FDE), Financial Services- NYCPublisherOpenAI (Ashby)Source typecompany job posting LangChain’s Deployed Engineer posting asks the role to own the technical win in pre-sales by designing proofs of concept. Source 2Deployed Engineer (Bay Area) @ LangChainPublisherLangChain (Ashby job board)Source typecompany job posting What keeps it FDE work is that the same engineer then builds and runs what was sold. To prepare for these interviews, we train one skill above the rest: scoping a first version and demoing it without promising more than you can ship.

Pre-sales or post-sale FDE: what changes

Think of the sale as a line: discovery, demos and proofs of concept to its left; the build, the rollout and the on-call to its right.

A post-sale FDE meets the customer to the right of the line. The promises are already made, and part of the job is finding out which ones hold. A pre-sales FDE starts to the left of it and helps decide what gets promised in the first place.

That difference changes your interview prep. A post-sale FDE needs stories about inheriting a messy scope. A pre-sales FDE needs stories about shaping one, often with a salesperson in the room who wants the answer to be yes.

Do forward deployed engineers do pre-sales? What postings say

Here is what employers write, on both the AI-lab and enterprise sides:

  • Discovery that starts before the deal. OpenAI’s posting also asks the same FDE to own on-call readiness for the production AI systems it ships, so the person scoping the deal answers for it later. Source 1Forward Deployed Engineer (FDE), Financial Services- NYCPublisherOpenAI (Ashby)Source typecompany job posting
  • The technical win, plus the build. LangChain’s Bay Area posting lists designing proofs of concept, guiding evaluations and running demos, next to co-building production AI agents with customer teams. Source 2Deployed Engineer (Bay Area) @ LangChainPublisherLangChain (Ashby job board)Source typecompany job posting Its UK posting describes the team’s work as running “from pre-sales evaluations to post-deployment advisory work”. Source 3Deployed Engineer (UK) @ LangChainPublisherLangChain (Ashby job board)Source typecompany job posting
  • Sales engineering as one ingredient. Baseten’s FDE posting calls the job a hands-on coding role that also includes parts of product management, technical customer success and pre-sales solution engineering. Source 4Forward Deployed Engineer @ BasetenPublisherBaseten (Ashby)Source typecompany job posting
  • Shaping the deal inside a consultancy. Deloitte’s ConvergeHEALTH FDE posting describes a role that works with account and product teams pre-sale, to qualify and shape opportunities, and with client teams post-sale. Source 5ConvergeHEALTH - Forward Deployed Engineer, Data Studio - Innovation_Delivery_TransformationPublisherDeloitte USSource typecompany job posting
  • A whole program for the pre-sales stage. As of September 2026, Anthropic’s board lists a Pre-Sales Program Lead for Forward Deployed Engineering to take the FDE team from opportunity to a signed contract, beside a Post-Sale Program Lead who owns delivery from signature onward. Source 6Pre-Sales Program Lead, Forward Deployed EngineeringPublisherAnthropic (Greenhouse)Source typecompany job posting That posting describes the FDE team as having more demand than it can serve, and “no machine yet for choosing well and closing fast”. Source 6Pre-Sales Program Lead, Forward Deployed EngineeringPublisherAnthropic (Greenhouse)Source typecompany job posting

Outside the postings, The Pragmatic Engineer reported, in August 2025, that many startups employing FDEs use them to help win customers, offering an FDE when a customer is undecided Source 7Forward Deployed Engineers (Gergely Orosz)PublisherThe Pragmatic EngineerSource typenews report, and quoted a Ramp job ad asking FDEs to work with sales and go-to-market teams to close deals. Source 7Forward Deployed Engineers (Gergely Orosz)PublisherThe Pragmatic EngineerSource typenews report

Note “choosing well”: when a team can’t serve every prospect, pre-sales work is partly deciding which deals to take, and you can show that skill in an interview.

Where it stops being FDE work

Past a certain edge, pre-sales work is sales engineering with a different title.

Colin Jarvis, who leads forward deployed engineering at OpenAI, said on a South Park Commons panel in May 2026 that the FDE unit is a business unit separate from solutions engineering, which he called pre-sales for the wider market, and from solutions architecture, which he called post-sales. Source 8The Future of Forward Deployed Engineering | OpenAI, Ramp, Nominal, DatalandPublisherSouth Park Commons (YouTube)Source typerecorded talk or interview He also said the unit has to say no more often than it says yes. Source 8The Future of Forward Deployed Engineering | OpenAI, Ramp, Nominal, DatalandPublisherSouth Park Commons (YouTube)Source typerecorded talk or interview

Cursor’s FDE postings draw the line from the other side: “This is not a demo role. You are responsible for systems that work in the real world.” Source 9Forward Deployed EngineerPublisherCursor (Ashby job board)Source typecompany job board

Incentives are the third edge. Andreessen Horowitz (a16z), in its essay on forward deployed engineering, advises that forward deployed teams be closely aligned with account executives, and says it is rare to find implementation leads who want to carry a quota, which can also lead to counterproductive behavior. Source 10Trading Margin for Moat: Why the Forward Deployed Engineer Is the Hottest Job in StartupsPublisherAndreessen HorowitzSource typeindependent analysis Our reading: close to the deal, but not carrying its quota. If a posting lists , our post on why some FDE jobs pay OTE shows how to read it.

Our test, for any posting: does the role end at the demo, or at the running system? If someone else builds what you demoed, it is sales engineering. From the postings above, these phrases signal a real pre-sales FDE role when they sit next to “on-call” or “ship production”:

  • “own the technical win”
  • “designing POCs” or proofs of concept
  • “from pre-sales through post-production”

Coming from sales engineering? Our post on moving from sales engineer to FDE covers the interview gaps.

Is a pre-sales FDE just a sales engineer?

Both roles can scope a proof of concept for the same prospect. What differs is what the proof of concept is for. This table is our framing, not a rule any employer publishes:

QuestionSales engineerPre-sales FDE
After signatureHands off to a delivery teamBuilds and runs it
A shortcut becomesSomeone else’s backlogYour own backlog
Success meansThe deal closesThe system runs in production

The middle row is the one interviews probe. A pre-sales FDE’s shortcut becomes their own backlog, so a good one says out loud which parts of the demo are real.

How FDE interviews test pre-sales skill

No employer publishes how its pre-sales skills are scored. What some postings do publish is the format.

  • A demo. Roboflow’s FDE posting lists a 60-minute technical assessment and a 30-minute demo in which the candidate builds a project with Roboflow and shares the work. Source 11Forward Deployed Engineer @ RoboflowPublisherRoboflow (Ashby job posting)Source typecompany job posting
  • A case study. Attio’s Forward Deployed GTM Engineer posting lists a case study interview and panel interviews with relevant stakeholders, and does not describe the case study’s format. Source 12Forward Deployed GTM Engineer @ AttioPublisherAttio (Ashby job posting)Source typecompany job posting
  • A proof of concept presented to a “customer”. One candidate reported, in a repository created in June 2026, an FDE technical challenge, built on the HappyRobot platform, in which the candidate presents a proof of concept, an inbound voice for carrier sales, to a customer played by the interviewer. Source 13FDE Technical Challenge: Inbound Carrier Sales (PDF)PublisherIanDalton (GitHub)Source typecandidate’s take-home repository

That last one is a pre-sales call in miniature. Our post on the customer role-play interview covers the moves for when the interviewer plays the client.

Across all three, our method trains three skills (how we teach, not an employer’s rubric):

  1. Scoping a first version the prospect can judge. Small, on their problem, with a pass bar agreed before you build.
  2. Demoing without overpromising. Say which parts are live, which are stubbed, and what would change in production.
  3. Holding the line on commitments. Commercial terms belong to the account owner. Our free question on procurement asking you for a staffing promise drills exactly this.

The free lesson what customer rounds look like, and the core moves sets out the moves these rounds reward.

A worked scoping answer for an undecided prospect

Here is a prompt of our own. You are the FDE on a call with the director of revenue cycle at a regional hospital network, who is evaluating whether your AI platform can check insurance claims for errors before they are submitted. The account executive is on the call. The director opens: “Your sales team says this works. I’ve heard that before. Convince me.”

Step 1: Refuse the pitch, ask about the decision

Don’t open with features. Find out what she has to decide.

“Fair. Before I show you anything, can I ask what ‘works’ has to mean for you to recommend this? Who else signs off, and what did you try before that didn’t work?”

Suppose she answers: denied claims are the pain, her billing team fixes them by hand, and a previous vendor’s tool flagged so many claims that the team stopped reading the flags. The CFO signs.

Step 2: Play it back

“So the real risk isn’t that we miss errors. It’s that we flag so much your team ignores us, like last time. And your CFO needs to see fewer denials, not a clever model. Is that right?”

Wait for the correction: the playback is where she tells you what you got wrong.

Step 3: Propose a proof of concept small enough to judge

“Here’s what I’d suggest instead of a general demo. Pick one claim type your team corrects most often. Give us last month’s claims of that type, both the ones that were denied and the ones that paid, de-identified. We run our checks and show you two numbers: of the denials caused by errors a check could catch, how many we’d have flagged, and of the claims that paid, how many we’d have flagged by mistake. You set the bar for both before we start.”

This is the core of the answer: it runs on her data, measures her fear (too many flags), and ends in a decision she can take to her CFO.

Step 4: Say what the demo will not show

“To be clear about what this proves: it runs on a copy of your data, not inside your billing system. Connecting to your system, handling access and running it every day is the real project, and I’d be the one building it. I’ll write down what we’d need from your IT team so you can check it before you sign. If de-identifying strips fields our checks need, like dates of service, we’d need a business associate agreement before we touch the real claims, and that goes on the list for your compliance team.”

That is not a formality: the Safe Harbor method in the HIPAA Privacy Rule removes every element of a date tied to a patient except the year, so a date of service does not survive de-identification.

Step 5: Handle the promise question

She will test you: “Can you promise it will catch every error?”

“No, and you shouldn’t trust anyone who does. What I can promise is that you’ll see the misses and the false flags on your own claims before you decide. If the numbers don’t clear your bar, you should walk away, and I’ll tell the account team the same.”

That last sentence costs you something in front of the account executive, which is exactly why the director believes it. Agree it with the AE before the call. It also fits the Anthropic line quoted earlier: a team with more demand than it can serve has to choose well before it closes fast.

Step 6: Close with owners and dates

“So: you send the de-identified extract and the bar. I send the data request and a one-page plan by Friday. We review results together. Pricing and contract terms are for your account executive and you.”

If you want a structure for turning a call like this into a written proposal, the lesson from discovery call to technical proposal in one sitting walks through it. That lesson is in Pro, which starts with a 7-day free trial; see pricing.

The story to have ready

The behavioral half asks for a time you shaped a scope with sales. A fictional skeleton to swap for your own:

  1. Situation. An account executive told a prospect the proof of concept would include live sync with their CRM.
  2. Before the call. You told the AE that sync was a month of work and offered a smaller test you could stand behind.
  3. The smaller proof of concept. A nightly export of one team’s records, run through the model.
  4. The pass bar. The prospect’s ops lead set it in her own words: fewer misrouted records than her team’s manual process, checked by her team.
  5. What shipped. The test cleared the bar, the deal signed with sync in phase two, and you built it.

The question when sales promised what you cannot deliver is the full drill.

Mistakes that sink pre-sales rounds

Pitching before diagnosing. You hear “claims” and describe your model. Fix: ask about the decision and the last failed attempt first. Our lesson diagnose before you prescribe is built around this habit.

A happy-path demo. You show the one claim that works. Fix: show a miss on purpose, and say what you would do about it. In our method, that minute is where a skeptic starts to believe the rest.

Promising production results from a proof of concept. “It’ll be the same accuracy in production” is a promise you can’t keep. Fix: name what changes between the demo and production, and who does that work.

Making commercial commitments. Staffing, price and contract timelines are not yours to sign. Fix: acknowledge the ask, name the owner, and offer the assumptions you can write down.

Treating the account executive as the enemy, or as the boss. Both lose the room. Fix: agree the plan with them before the call, and correct a promise together, not in front of the customer.

A proof of concept with no pass bar. Without an agreed bar, a good result closes nothing. Fix: get the bar in the prospect’s words before you build.

Forgetting you will own it. A demo nobody could run in production only moves the problem to your future self. Fix: for every shortcut, say how the real version would work. Sometimes the right answer is that the prospect doesn’t need an embedded engineer at all; our question on when a customer does not need an FDE is that judgment call.

Before a pre-sales round

  • I can say, in one sentence, whether the role ends at the demo or at the running system.
  • I have a story about scoping a first version with a salesperson in the room.
  • I can say “no, and here’s what I can promise” without sounding defensive.
  • I name the stubbed parts of any demo I show.
  • I close with owners, dates and the decision the proof of concept enables.

Practice the call, not just the answers

Reading the moves is the easy part. Doing them while a skeptical customer interrupts is the hard part, so practice against someone who pushes back.

The free practice case at /try puts you in front of an AI customer: a city whose mayor has already promised an AI review tool, holding facts it reveals only when you ask the right question. Pitch a tool before you’ve asked where the time goes, and it pushes back. It is free, with a sign-in. Like the call above, someone has already been promised something, and your questions decide what gets built.

Start with the free practice case, and run it the way you would run Thursday’s call with the prospect on the fence.

GlossaryForward deployed engineerA software engineer who builds and ships production systems inside a customer’s problem and environment, accountable to that customer’s outcome.More on Forward deployed engineerGlossaryOn-target earningsBase pay plus variable pay if targets are met; a posting that states pay as OTE includes incentive pay in the figure.More on On-target earningsGlossaryAgentA system in which a model chooses steps and tool calls to complete a task, within limits the design sets.More on Agent

Questions people ask

Do forward deployed engineers work pre-sales?

Some do. OpenAI’s Financial Services FDE posting asks the FDE to lead discovery and scoping from pre-sales through post-production, and LangChain’s Deployed Engineer posting asks the role to own the technical win in pre-sales by designing proofs of concept.Source 1Forward Deployed Engineer (FDE), Financial Services- NYCPublisherOpenAI (Ashby)Source typecompany job postingSource 2Deployed Engineer (Bay Area) @ LangChainPublisherLangChain (Ashby job board)Source typecompany job posting

Is a pre-sales FDE the same as a sales engineer?

Not in the postings we read. OpenAI’s and LangChain’s FDE postings that include pre-sales work also have the FDE build and ship production systems. Colin Jarvis, who leads forward deployed engineering at OpenAI, has described its FDE unit as separate from its pre-sales solutions engineering.Source 1Forward Deployed Engineer (FDE), Financial Services- NYCPublisherOpenAI (Ashby)Source typecompany job postingSource 2Deployed Engineer (Bay Area) @ LangChainPublisherLangChain (Ashby job board)Source typecompany job postingSource 8The Future of Forward Deployed Engineering | OpenAI, Ramp, Nominal, DatalandPublisherSouth Park Commons (YouTube)Source typerecorded talk or interview

How do FDE interviews test pre-sales skill?

Some postings list a demo or a case study. Roboflow’s FDE posting lists a demo in which the candidate builds a project and shares the work, and Attio’s Forward Deployed GTM Engineer posting lists a case study interview without describing its format.Source 11Forward Deployed Engineer @ RoboflowPublisherRoboflow (Ashby job posting)Source typecompany job postingSource 12Forward Deployed GTM Engineer @ AttioPublisherAttio (Ashby job posting)Source typecompany job posting

Keep reading

Guides

Lessons

Questions

More in The FDE role, page 2