Short answer
Forward-deployed engineering puts a vendor's engineers inside a client's team to build and ship software against the client's real problem; Palantir created the role in the early 2010s. Pivotal Labs, founded in 1989, embedded its engineers with client teams too, pairing on every line of code so the client could keep building after Pivotal left. A Pivotal Labs-inspired AI consultancy combines the two: engineers who ship against a business number, and pairing when the client wants its own team to own the result.
Key takeaways
- A forward-deployed engineer works inside the client's team and writes and ships the code, rather than advising from outside.
- Palantir's model leaves behind a working deployment of the vendor's product; Pivotal's left behind software plus a client team that could keep building it.
- Pivotal Labs' shorthand for capability transfer was "I do, we do, you do."
- Cosmic delivers the outcome by default and pairs with the client's team when the client wants to own the work after the engagement.
Forward-deployed engineering is back in fashion. AI companies are hiring forward-deployed engineers to sit inside customer teams and get models working on real problems, and The Pragmatic Engineer reported in 2025 that OpenAI had built an FDE team from two engineers to more than ten. The idea is older than the current wave. Palantir named the role in the early 2010s, and Pivotal Labs had been embedding its engineers with client teams for years before that. Cosmic comes from Pivotal Labs, so this post explains what each model is, where they differ, and how we use both in AI work.
Where the forward-deployed engineer came from
Palantir created the role in the early 2010s and called it "Delta" internally. Its software went into government agencies and large enterprises where it couldn't be installed and handed off, so Palantir sent engineers to work on site and build against the customer's actual operations. According to The Pragmatic Engineer, until about 2016 Palantir had more forward-deployed engineers than ordinary software engineers.
The goal was adoption of Palantir's platform. The engineer's job was to make the product work inside the customer's world, and to feed what they learned back to the product team.
What Pivotal Labs did
Pivotal Labs was founded in San Francisco in 1989 and became known for extreme programming: pair programming on every line of code, test-driven development and small, frequent releases. Client engineers, designers and product managers came to work alongside Pivotal's people, paired with them all day, and learned the practice by doing it. Pivotal Labs later became part of Pivotal Software and then VMware Tanzu Labs, which shut down in January 2025.
Every Pivotal engagement balanced people, process and technology. The software mattered, and so did whether the client's team could keep building it the same way after Pivotal left. The shorthand for how that transfer happened was "I do, we do, you do": first we do the work while you watch, then we do it together, then you do it while we watch.
The difference between the two
Both models put engineers inside the client's team and judge them by shipped software. They differ in what the client keeps at the end.
A Palantir-style forward-deployed engineer leaves behind a working deployment of the vendor's product. A Pivotal-style engagement leaves behind working software plus a client team that can run it and keep building it. One serves product adoption. The other serves capability transfer.
How Cosmic applies both to AI implementation
We start where the forward-deployed model starts: inside your business, on one number you care about (see revenue-focused AI consulting), with engineers who build and ship. Discovery with your executive sponsor sets the number and maps the system that produces it. Then we build the AI that moves it and report each release against the baseline.
Being a boutique AI consultancy headquartered in McLean, Virginia, just outside Washington, DC, we can do this in person for clients across DC, Maryland and Virginia. Pairing is in our DNA, and it isn't always the right call. Most clients want the outcome first, so by default Cosmic does the work and drives the result with the sponsor. When a client wants its own team to own the work after we leave, we find the right people on their side and pair with them, I do, we do, you do, until they can run it without us.
The part that's new since Pivotal is who else is on the team. AI agents now do the repetitive build work, running in harnesses we built ourselves, such as an agent work coordinator that lets one person run several agents on one backlog. The pairing discipline carries over: a person reviews everything an agent builds before it ships.
What to ask a Pivotal Labs-inspired AI consultancy
These sit alongside the broader checklist in how to choose an AI consultancy, and the questions on how AI consulting is priced.
- Who from their team will be in your standups, and are they the people who build?
- What number will the work be judged by, and how will they measure the baseline?
- If you want your team to own the result, how does the handoff work week by week?
- Who reviews what the AI agents build, and what has to pass before it ships?
FAQ
What is a forward-deployed engineer?
A software engineer who works inside a client's team, on the client's real problem, and writes and ships the code that solves it. Palantir created the role in the early 2010s, and AI companies such as OpenAI now hire forward-deployed engineers to get their models working inside customer businesses.
Did Pivotal Labs invent forward-deployed engineering?
No. Palantir named the forward-deployed engineer. Pivotal Labs practiced a related model earlier, embedding and pairing its engineers with client teams, with the added goal of leaving the client's team able to keep building on its own.
What does "I do, we do, you do" mean?
It's the paired enablement model Pivotal Labs used to transfer practice to client teams. First the consultant does the work while the client's people watch, then they do it together, then the client's people do it while the consultant watches.
What happened to Pivotal Labs?
Pivotal Labs became part of Pivotal Software, which VMware acquired in 2019 and renamed Tanzu Labs. After Broadcom acquired VMware, Tanzu Labs shut down in January 2025.
Sources
Founder of Cosmic. Nearly 15 years in strategy, management and sales as a digital strategist, director of sales and product manager. He closed over $25M in new revenue at VMware Pivotal Labs, launched digital products at Capital One, and helped Viget grow from 30 to 75 people. He started Cosmic and built the Revenue Design® process after advising friends whose companies struggled to close deals and keep revenue steady.