Who Gets an FDE, and Who Doesn’t: The Great B2B + AI Debate Right Now | Saa Str

Overview

Free e Books

	e Book: Hiring a Great VP of Sales
	e Book: Raising Capital
	e Book:  The First $1m ARR

University All Posts University Podcasts The Top CROs VC Fundraising Top Videos Q&A Best of Saa Str #1 Bestselling Book Search Everything Join the Community

Details

Free e Books

	e Book: Hiring a Great VP of Sales
	e Book: Raising Capital
	e Book:  The First $1m ARR

London 2025 Annual 2026 Events Overview Sponsors

	Event Sponsorship
	Media Sponsorship

Digital AI Day 2025 (Free) Speaker Submissions Speaker Requirements Overview

Who Gets an FDE, and Who Doesn’t: The Great B2B + AI Debate Right Now

by Jason Lemkin | Artificial Intelligence (AI), Blog Posts, Saa Str. Ai

I was catching up with one of our favorite AI agent vendors the other day. Strong product. Strong traction. One of the ones we recommend.

And they told me: going forward, only customers with 5,000 or more employees get a dedicated Forward Deployed Engineer.

Everyone smaller is on their own. They can self-train, read documentation, etc.

I get it. When you’re growing that fast, FDE capacity becomes your binding constraint almost overnight. You cannot hire and train them fast enough. So you triage. You put your limited humans on your biggest, most expensive accounts. That’s the rational call.

But here’s what it means in practice: a two-tier world is forming. Enterprises get the deployment help that actually makes agents work. Everyone else gets a knowledge base and a Loom video.

And if you’ve been paying attention to what actually drives results with AI agents, that should bother you.

We run 20+ AI agents at Saa Str. They generate over $1M in revenue. I’ve been at this for 18+ months across every part of our go-to-market.

Here is what I know for certain: the single biggest variable in whether an agent actually works is not the model, not the prompt, not even the vendor. It’s whether you get a real human from the vendor helping you deploy it.

Every agent we run that actually works had FDE involvement at launch. Every one.

Our outbound agent took weeks of ingesting a decade of attendee data. Our inbound qualification agent, which now books 130+ meetings automatically and runs 24/7, required someone getting the Salesforce routing right, the qualification logic right, the handoff right. None of that happens from a help doc.

Zendesk’s CEO said it outright in our workshop with G2 a few months back: enterprise customers who go through proper deployment hit 60-80% automation rates. Self-serve customers land around 20% when self-trained on docs from the customer’s website and FAQs, etc.

That’s not a small gap. That’s the difference between an agent that changes how your company operates and one you turn off after 90 days because it never quite worked.

The most interesting data point I have on this is Salesforce.

Agentforce hit $540M ARR. Real number. But only around 8% of their customer base has adopted. And they had to discount hard at first to get enterprises in the door, which if you know Salesforce, tells you something.

What got results for us was deployment. Salesforce put real FDE resources on us. They didn’t just send a deck. They got in the system with us. Configured it against our actual data. Got it working.

After the last 3-4 Saa Str AI Annuals we had roughly 1,000 people who’d filled out our sponsorship interest form and received zero follow-up. A rep had ghosted them. Revenue from those leads: zero.

72% open rate. Ten-plus percent response rate on contacts that had gone dark for six months. Deals closing from leads we’d written off entirely. It worked because the agent knew their full Salesforce history. Every event attended, every prior sponsorship, every interaction. The outreach felt like a conversation continuing, not a cold sequence from a company they barely remembered filling out a form for.

That’s what deployment actually produces. Not the discount.

Five AI Agent Vendors Asked Us to Try Their Product This Week. Just One Got Deployed. Here’s the Difference.

Five AI Agent Vendors Asked Us to Try Their Product This Week. Just One Got Deployed. Here’s the Difference.

The companies that most need AI agents to work are often not the enterprises. It’s the 200-person company running a three-person sales team. The 500-person company trying to handle 50,000 support tickets without 40 reps. The 800-person company trying to grow pipeline without growing headcount proportionally.

Those companies need the agents to actually work. Not as a productivity bump. As a structural shift in how they operate.

And they are exactly the customers who are no longer getting FDE support.

Palantir invented the FDE model. Before anyone else was calling it that, they were embedding engineers inside government agencies and banks because there was no other way to make the software work in real environments. Messy data, legacy systems, workflows nobody had ever documented. You couldn’t fix that remotely.

For years, Palantir’s “acquire” phase meant multi-month pilots where they absorbed the upfront FDE cost themselves. Expensive. Hard to scale. Their engineers were on-site at Airbus assembly lines, inside airgapped military networks, doing work that genuinely could not be done from headquarters.

But then they did something smart. They separated “prove value” from “full deployment.”

The AIP Bootcamp. Five days. The customer’s own real-world data. By the end, a working use case in production — not a demo environment, not mocked-up data. What used to be a traditional multi-month sales cycle got compressed into days. These bootcamps now carry a nearly 75% conversion rate.

The insight was structural. The bootcamp handles the “Acquire” phase. Then, once a customer sees value in the initial use case, Palantir deploys FDEs during the “Expand” phase — when the product is already working, when the customer is already bought in, when the FDE is expanding something proven rather than proving something unknown. That’s a completely different FDE job, and a far more efficient use of expensive talent.

Palantir doesn’t hand you the keys and walk away. It builds, deploys, and even operates the workflows in production, then helps the client stand up a Center of Excellence with defined roles so the organization can eventually run on its own. The intent is customer autonomy. The reality is that by the time a customer can run independently, they’ve usually found ten more things to build.

The result of all this: U. S. commercial revenue surged 137%, and operating margins hit a record 51% in late 2025, driven by the lower overhead of AIP deployments.

That’s not a coincidence. That’s what happens when you solve the deployment bottleneck structurally instead of just throwing more FDEs at it.

No other major AI vendor has done this yet. Most are still in Palantir’s 2015 mode — FDE-heavy from the first day of a new customer relationship, which means FDE capacity directly caps how many customers you can take on. The 5,000-employee threshold my vendor friend described is what happens when you haven’t found your version of the bootcamp yet.

What if the answer isn’t to triage by company size — but to build products that deliver FDE-quality deployment to every customer?

Think about what an FDE actually does. They ask the right questions to understand the workflow. They find where the data is broken. They configure for the specific use case. They test, iterate, catch the edge cases before go-live. They check in for the first 90 days.

That is a defined set of tasks. Hard, judgment-heavy, relationship-requiring. But defined.

Some vendors are starting to build toward this. Intelligent onboarding that adapts to how your data is structured. Agentic deployment assistants that walk you through configuration in real time. None of them are at Palantir-grade FDE quality yet. But the direction is obvious.  Sierra just announced something similar.

And if you just get close, you may be able to deploy with far less FDE time.  Some time, but less of it.

The vendors who close this gap first — who figures out how to deliver real deployment outcomes regardless of company size — has a very different competitive position than everyone else. Right now the FDE constraint is an artificial ceiling on growth. And the companies below 5,000 employees who are getting so-so results? When someone shows up offering them actual deployment support, they’ll switch.

If you’re an AI agent vendor: be honest about who’s actually getting what.

Don’t let sales oversell white-glove and then hand the account to a CS rep with 200 customers. Tell smaller companies exactly what they’re getting, what they’ll need to invest internally, and how long this realistically takes.

The two-tier world is real. Enterprises get FDEs. Everyone else is largely on their own. That’s not a criticism of any specific vendor — it’s just where the math has landed for now.

Forward Deployed Engineer: What It Takes to Make AI Work in B2B. But Do They Work for SMBs?

Forward Deployed Engineer: What It Takes to Make AI Work in B2B. But Do They Work for SMBs?

Dear Saa Str: How Long Does It Take to Deploy an AI SDR?

Dear Saa Str: How Long Does It Take to Deploy an AI SDR?

Industry News

Get from

Key Takeaways

• Free e Books

    e Book: Hiring a Great VP of Sales
    e Book: Raising Capital
    e Book:  The First $1m ARR
  

• University All Posts University Podcasts The Top CROs VC Fundraising Top Videos Q&A Best of Saa Str #1 Bestselling Book Search Everything Join the Community

• Free e Books

    e Book: Hiring a Great VP of Sales
    e Book: Raising Capital
    e Book:  The First $1m ARR
  

• London 2025 Annual 2026 Events Overview Sponsors

    Event Sponsorship
    Media Sponsorship
  

• Digital AI Day 2025 (Free) Speaker Submissions Speaker Requirements Overview