// PROOF
Dismantling Jira for a Forward Deployed Engineer model, at 10x output
Replaced traditional ticket-driven project management with a Palantir-style Forward Deployed Engineer model, embedding engineers directly in client problem spaces and framing high-velocity pods under The Harvest Methodology.
METHODMeasured as delivered scope per pod per cycle against the previous ticket-driven baseline, on comparable engagements. The figure describes throughput of shipped client value, not lines of code or ticket counts — the latter deliberately went down.
Ticket systems do not describe how AI work actually happens. They describe how work happened in 2011, when scope was knowable in advance and the expensive part was implementation.
In AI delivery the expensive part is finding out what the customer actually needs, which cannot be specified into a ticket because nobody knows it yet — including the customer. A backlog full of confident tickets about an unexplored problem is a document about wishes.
The Forward Deployed Engineer model puts the engineer inside the client’s problem instead of downstream of a description of it. The person who will build the thing is the person who watched the operations team work around the broken process for two days. That shortens the loop from observation to working software from weeks to hours, and it eliminates the most expensive artifact in professional services: a perfectly-executed build of the wrong thing.
Pods are how that scales without becoming chaos. Each has clear ownership, naming discipline, and a rapid feedback loop, and each selects its work through ICE scoring rather than through whoever argued most persuasively in planning.
Naming discipline sounds like the trivial part and is not. When every pod names its artifacts the same way, the output of one engagement is legible to the next. That is the precondition for reuse — and reuse is the only thing that makes the second engagement cheaper than the first.
Every enterprise AI programme I have been handed was a slow ascent. Eighteen months of planning produces a very well-documented lump of graphite.
Forward Deployed Engineer modelDeployment PodsICE scoringThe Harvest Methodology