AI is at an inflection point. Models are powerful, tools are plentiful, and pilots are easy to launch. Yet most organizations stall when moving from prototypes to meaningful transformation. The constraint isn’t intelligence; it’s execution.
The solution is not new. For decades, the best engineering outcomes have come when builders sit close to users. Palantir recently helped popularize the term “forward-deployed engineer,” but the practice long predates it. Embedding technical talent in the field, side by side with operators, has always been a reliable path to impact.
Lessons From the 1990s: Reengineering the Corporation
The 1990s marked the dawn of the information age. Client-server systems and the early web enabled companies to connect data in ways never possible before. ERP and CRM systems like SAP and PeopleSoft promised to rewire how entire enterprises worked.
But the challenge wasn’t the technology itself; it was reengineering business processes around it. Armies of consultants and engineers were embedded inside companies, redesigning workflows from the ground up. Michael Hammer and James Champy’s Reengineering the Corporation captured this moment: organizations had to rethink how work got done, enabled by new digital capabilities.
Forward-deployed engineers weren’t always called that, but they were there: sitting inside enterprises, customizing ERP rollouts, bridging between messy reality and new technology, and learning patterns that would later be generalized into products. Over time, what started as bespoke process reengineering matured into productized SaaS categories like CRM, ERP, and HRIS.
The New Parallel: The Intelligence Age
Today’s AI moment carries the same disruptive weight. LLMs and agents can augment or automate work in ways that seemed impossible only a few years ago. But realizing that potential requires embedding engineers directly with the business to figure out how to reengineer processes around this new capability.
Just as in the 1990s, the bottleneck isn’t the underlying technology. It’s the translation into the real work environment. APIs are undocumented, authentication is fragile, legacy systems don’t talk to each other. The models are brilliant, but integration with messy enterprise reality is the real test.
Why Forward Deployment Matters Now
Business users already know what’s broken. They live the inefficiencies every day. What they lack isn’t ideas; it’s the tools and engineering leverage to fix them. Forward-deployed engineers close that gap, turning user pain points into working systems quickly, and then generalizing the lessons into patterns that can scale.
This is how reengineering becomes transformation. First, solve the problem in one place. Then identify the repeatable pattern. Then scale. The path mirrors how ERP and CRM matured in the 90s, except today the cycle runs in months instead of years.
The AI Tiger Team: Assembling Your Crew for Success
A forward deployed engineer is your spearhead, but they don’t work alone. Success requires a small, agile “Tiger Team” with distinct roles:
The Forward Deployed Expert: This is your technical leader on the ground. Depending on your project’s stage, this could be a builder, a specialist, or an enabler. The key is their technical skill combined with a relentless focus on the customer’s mission.
The Business Analyst: This isn’t just someone who documents processes. The modern BA in an AI world needs a deep, boots-on-the-ground understanding of how work actually gets done. More importantly, they are now power users of the new AI tools themselves, able to quickly demonstrate what’s possible.
The Executive Sponsor: This role provides air cover for this new way of working. They must understand that real transformation takes months, not weeks, and protect the team from pressure for premature scaling. Their key metric isn’t a successful POC; it’s tangible, sustained adoption.
The End Users: These are your experts who understand the current state and who can evaluate the usefulness of the solution. They already know what’s broken; they just lack the tools to fix it. They are critical for the rapid feedback needed to iterate and refine solutions.
Principles for Forward Deployment
So, how do you put this into practice? It starts with a fundamentally different approach.
Start with Immersion, Not Requirements
Don’t start with solutions. Embed your team directly in the operational environment. Watch the actual work, including the informal processes and workarounds. Document the exceptions, not just the happy path. Identify where current tools fail and manual intervention begins. The objective is to understand the nuances of the work and identify the specific points of friction where technology can make a real difference.
Prototype in Hours, Not Weeks
The days of disappearing for three months and coming back with a finished product are over. With today’s tools, the forward deployed engineer and business analyst can build and show a working prototype in a matter of hours. The conversation shifts from an imaginary concept to a tangible tool that the user can touch and break. The question becomes, “Is this what you mean?”
Embrace the Unscalable
This is the hardest part for many organizations. Your first solution will not be elegant. It might involve an engineer manually triggering processes. That’s okay. The goal is to deliver value first. Solve the specific, messy problem for one team. Once you prove the value, you can then focus on generalizing the solution and scaling it. The most scalable solutions almost always start this way.
Your Next Move
Successfully implementing AI means closing the distance between technical capability and business reality through small, empowered teams embedded directly where work happens.
Start with one broken process that everyone acknowledges needs fixing. Assemble a tiger team combining technical excellence with business intimacy. Give them permission to build imperfect solutions that deliver real value. This won’t happen overnight, but starting small means you can show value within weeks while building toward transformation.
In an environment where capabilities evolve monthly but business complexity persists, this human-centered, field-first approach is the difference between AI that demos well and AI that actually works.
