- The Angle
- Posts
- FDEs probably aren't the answer
FDEs probably aren’t the answer
David Peterson
I always get nervous when a tactic gets sexy enough to become “the playbook.”
That is what happened to product-led growth in the late 2010s. And I suspect that is what is happening to “forward deployed engineers” in AI now.
I lived the PLG version of this. I was running growth at Airtable and we were trying very hard to act like a standard product-led growth company with funnel optimization, clever activation flows, all the usual stuff.
But none of it worked.
The short version of a longer story I wrote about elsewhere is this: the PLG playbook assumed a legible funnel and a clear job to be done. Airtable had neither. But because all the “it” companies were PLG companies, we kept trying to make the tactic fit.
Eventually we gave up and started from first principles. And what we came up with was a super unique motion. We didn’t run a single growth experiment. Instead, we spent all our time watching what people built, identifying those emergent use cases that really clicked, and then backing into product and growth strategies that would serve those use cases. It might have looked like PLG on the surface, but underneath, it was anything but.
Fast forward to today and FDEs are having their PLG moment.
The line you hear is that AI is so horizontal, so probabilistic, that the only way to get it into the crevices of an organization is to embed people on site. Every AI startup needs FDEs. It is just what you do.
Palantir is the obvious reference point. They have been running a forward-deployed model for two decades, mostly with governments and mega-enterprises. In 2024, more than half of Palantir’s roughly $3 billion in revenue came from government customers, and the U.S. Army just awarded them an “up to $10 billion” enterprise software deal over ten years. Average account sizes are measured in the millions and some programs are worth hundreds of millions or more over their lifetime.
Most AI startups do not live in that universe. If your median deal is $100K and you are copying the GTM of a company doing multi-million-dollar deployments to the DoD, you are probably telling yourself a nice story while you quietly build a consulting firm.
When you’re building your product for the first time, you should of course embed deeply with customers. But the goal should be to build such a great product that it does not require you to forward deploy anybody for a customer to be successful. Relying heavily on the FDE motion brings its own dangers. As Villi from Category Ventures puts it, if you solve problems with people, you will default to throwing bodies at problems and will ultimately build an inferior product.
The real lesson from both PLG and FDE is boring and unsatisfying: there is no tactic that saves you from having to think.
Things only started to work at Airtable when we stopped asking “what works for the best PLG companies?” and started asking “what will work for us?” I suspect the same will be true for most AI companies. The winners will not be “the ones who forward deployed the most cracked engineers” or whatever. They will be the ones who understand their product and customer so well that the GTM almost feels obvious.
You should absolutely study any company that seems to have unlocked something magical with their product and with their go-to-market. Go research Facebook’s growth team and the early product-led growth thinkers. Go deep on Palantir and their FDE model. Go analyze Wiz’s unique enterprise motion that enabled such insane growth. But then close those tabs. Carefully consider your unique product and customer. And design something that fits. If it happens to rhyme with the tactic of the day, your VCs will probably love you for it (we’re dumb like that). But that is a coincidence, not a strategy.
FROM THE BLOG
The AI-Native Enterprise Playbook
Ten real-time observations on a rapidly evolving playing field
No More Painting by Numbers
It’s the end of the “SaaS playbook.
The Age of Artisanal Software May Finally be Over
Every wave of technological innovation has been catalyzed by the cost of something expensive trending to zero. Now that’s happening to software.
Founders as Experiment Designers
David on why founders should run everything as an experiment.
WORTH READING
ENTERPRISE/TECH NEWS
The all time high - As NVIDIA’s valuation hit $5 trillion, the first company to ever do so and just months after it reached $4 trillion Calcalist dives into the story of how Michael Kagan a 68 year old Israeli founder became the company’s CTO and his thoughts on the future of AI. Kagan, who was the founder of Mellanox, which was acquired by NVIDIA, is an electrical engineer by training, and has often compared AI to infrastructure. “That’s exactly the revolution. The story of AI is like our dependence on electricity, it will become infrastructure for everything. You’ll use it without thinking, just as you use electricity or water without wondering how it reaches you. When you’re driving and your car keeps you safely in your lane, that’s AI. Or in a smart home. Even with electricity, people didn’t initially understand how far its uses would go; they started with the light bulb, and now look, we have electric cars and smartphones. The same is true of artificial intelligence. We’re only at the beginning.”
The Gemini supremacy - As polymarket predicted, Google’s release of Gemini 3 has finally given them the top model in the field. Aakash Gupta, who had early access to the model, writes an effusive post about how it isn't just the top model, it's rewriting AI infrastructure. Google trained Gemini 3 entirely on custom TPUs, looping NVIDIA out of the supply chain. “But Gemini 3 proves something else. If you can build state-of-the-art models on your own chips, and you have the capital to manufacture them at scale, the game changes.Google has capital. Google has training data. Google has proprietary TPUs. And now Google has proven these TPUs can reach the frontier. Nvidia sells shovels. Google is building the mine.”
HOW TO STARTUP
The weekly email - Gokul Rajaram argues for a weekly or fortnightly CEO email to their whole company. While All-Hands are great for company culture and communication, his view is that they are neither personal or permanent enough and that these can include things which are top of mind for the CEO, performance metrics and miscellaneous notices. “The CEO email should give an update on how the organization is doing against these goals and initiatives, in an unvarnished way. In terms of format, I prefer a Red / Yellow / Green grade for each item, with commentary on progress since the last update. If you are unable to assign a grade to something, it needs to be rephrased.”
The problem you’re hiring for - Do you hire ‘people first’ (as has become trendy recently) or for specific roles? Neither as Gaurav Vohra writes in Growth by Gaurav. And the whole world of recruiters would agree with him, to focus on deliverables you need, the outcomes you’re looking for, rather than a paint by numbers approach to ‘needing a growth marketer’ without knowing what that is. Or hiring a plethora of unmoored high achievers. Vohra’s view on the latter is that it works for co-founders, first team hires and anyone in the top 10 (people not percent) in their field in the world. But even then, high achievers will ultimately want to know what challenges they need to be pointed at. One crucial point Vohra mentions is to write an MOC (what A16Z outlines in their hiring advice, or ‘Mission, Outcome, and Competency’) before a JD. Implicit is to never rip a JD from google or ask an AI to write it. GH Smart, recruiter guru, would call this a ‘scorecard approach’, similar but different vernacular - and he would suggest outlining deliverables, characteristics, skills and traits for each role is a useful approach.
HOW TO VENTURE
The road to Hel - US and European VC’s converged on Helsinki last week for the annual Slush conference, with a particularly pithy welcome sign ‘Still doubting Europe? Go to Hel’. As Pitchbook points out, everyone was talking about whether there is an AI bubble, the state of European tech (concerns over regulation, depth of talent and funding), and defence tech was still a hot topic. Spare a thought for the VCs who in between saunas and cold plunges needed to find 10 different ways to say ‘I don’t know’ about the AI bubble.
PORTFOLIO NEWS
Blue Energy featured on Bloombeg News over plans to develop a nuclear-data center campus in Texas.
PORTFOLIO JOBS
Reply