The future of Forward Deployed Engineering in relation to small businesses
An FDE for the Small Business
The future of Forward Deployed Engineering in relation to small businesses
People spend a lot of time arguing about what the new AI jobs should be called. One month it is a glue automation agency. The next it is a vertical AI agency, an in-house AI lead, or a Forward Deployed Engineer (Joe Schmidt, "Trading Margin for Moat: Why the Forward Deployed Engineer Is the Hottest Job in Startups," Andreessen Horowitz, 2025 https://a16z.com/services-led-growth/). I think the names are mostly a distraction. Whatever the market ends up calling it, my bet is that small businesses will need some version of a Forward Deployed Engineer of their own.
Joe Schmidt, "Trading Margin for Moat: Why the Forward Deployed Engineer Is the Hottest Job in Startups," Andreessen Horowitz, 2025 https://a16z.com/services-led-growth/That entity might be a full-time employee, a freelancer kept on retainer, or an agency that steps in to do the same work. The arrangement matters much less than the function. A small business needs a representative who can translate rapid progress in AI into measurable outcomes, specifically in sales, revenue, and profit. Those are the things a small business actually cares about. It wants to grow, sell more, and become more profitable. Anything that does not eventually show up there is hard to justify paying for.
We have seen this pattern before. Microsoft Enterprise grew into an ecosystem worth hundreds of billions of dollars because Microsoft built a network of certified people and sent them out into companies. Those people set up integrations, worked through local problems, and brought regular staff onto the new tools. Amazon did the same thing with AWS through certifications and partner programs. Good software rarely finishes the journey into a business by itself.
I do not think AI is going to change that, because the rate of technological progress almost never matches the rate of adoption (Arvind Narayanan and Sayash Kapoor, "AI as Normal Technology," Knight First Amendment Institute, 2025 https://kfai-documents.s3.amazonaws.com/documents/c3cac5a2a7/AI-as-Normal-Technology---Narayanan---Kapoor.pdf). It took decades for the internet to diffuse into ordinary life, and it still has not reached saturation (International Telecommunication Union, Measuring digital development: Facts and Figures 2025, ITU, 2025 https://www.itu.int/hub/publication/d-ind-ict_mdd-2025-3/). It is still spreading. The same goes for computing. The latest phones and computers have been around for years, yet large parts of the world and millions of people are still nowhere near them. Almost every breakthrough we now take for granted took a long, uneven time to travel from the people who built it to the people who needed it.
Arvind Narayanan and Sayash Kapoor, "AI as Normal Technology," Knight First Amendment Institute, 2025 https://kfai-documents.s3.amazonaws.com/documents/c3cac5a2a7/AI-as-Normal-Technology---Narayanan---Kapoor.pdfInternational Telecommunication Union, Measuring digital development: Facts and Figures 2025, ITU, 2025 https://www.itu.int/hub/publication/d-ind-ict_mdd-2025-3/AI hides this lag because the entertaining parts spread so quickly. Generating pictures, making short videos, and talking to chatbots traveled everywhere in months. Those use cases belong mostly to media and entertainment, where curiosity is enough to get someone started and nothing breaks when the output is slightly wrong. Moving the bottom line of a real business is a slower and harder problem. Getting a company more sales, steadier revenue, or wider margins means fitting the technology into how that company actually runs (Paul A. David, "The Dynamo and the Computer: An Historical Perspective on the Modern Productivity Paradox," American Economic Review, 1990 https://ideas.repec.org/a/aea/aecrev/v80y1990i2p355-61.html).
Paul A. David, "The Dynamo and the Computer: An Historical Perspective on the Modern Productivity Paradox," American Economic Review, 1990 https://ideas.repec.org/a/aea/aecrev/v80y1990i2p355-61.htmlA small business owner rarely has the room to figure that out alone. Not because they are incapable of learning, though keeping up with how fast models change is already a full-time job on its own. It is mostly because running a small business takes everything you have. You are chasing customers, managing staff, handling deliveries, and watching cash flow. You cannot pause the business for three months to study how a new model behaves, figure out where it fails quietly, and connect it to your daily operations without breaking what already pays the bills.
This is why I think the real work belongs to someone standing in the middle. The closer you stay to the models, the easier it becomes to understand how they work, what they can genuinely handle, where their limits are, and how to observe them once they are running. When you also understand how a business makes money, you can map one onto the other. You know both sides of the relationship well enough to tell when a new capability is actually useful and when it is just noise. Most owners have no desire to study models closely. They want someone who understands both sides well enough to connect them and produce a reliable result.
I know there is a real chance this eventually stops being true. As models grow in capacity, they may become smart enough to understand a business on their own, connect themselves to its operations, and make the person sitting between the two redundant. That is possible. The honest answer is that we do not know yet if models will reach that point, or how many years away it might be.
Until they do, the distance between what a model can do and what a small business actually captures will remain wide. Models will keep getting stronger, and small businesses will keep caring about sales, revenue, growth, and profit. For the foreseeable future, they will need an FDE in some form, whether that is an employee, a freelancer, or an agency. Someone still has to understand what the models can do, understand what the business needs, and stay responsible for turning one into growth for the other.
Trading Margin for Moat: Why the Forward Deployed Engineer Is the Hottest Job in Startups, a16z.com https://a16z.com/services-led-growth/
Diego Comin and Bart Hobijn, "Cross-Country Technology Adoption: Making the Theories Face the Facts," Federal Reserve Bank of New York Staff Report 169, 2003 https://www.newyorkfed.org/research/staff_reports/sr169.html
1Joe Schmidt, "Trading Margin for Moat: Why the Forward Deployed Engineer Is the Hottest Job in Startups," Andreessen Horowitz, 2025 https://a16z.com/services-led-growth/
2Arvind Narayanan and Sayash Kapoor, "AI as Normal Technology," Knight First Amendment Institute, 2025 https://kfai-documents.s3.amazonaws.com/documents/c3cac5a2a7/AI-as-Normal-Technology---Narayanan---Kapoor.pdf
3International Telecommunication Union, Measuring digital development: Facts and Figures 2025, ITU, 2025 https://www.itu.int/hub/publication/d-ind-ict_mdd-2025-3/
4Paul A. David, "The Dynamo and the Computer: An Historical Perspective on the Modern Productivity Paradox," American Economic Review, 1990 https://ideas.repec.org/a/aea/aecrev/v80y1990i2p355-61.html