The other day a colleague heard Pablo Dominguez of Insight Partners on Bloomberg. He was talking about their new Forward Deployed Practice, and she sent me a note saying something clicked with her.
It clicked for me too.
In this case, a forward deployed engineer (FDE) sits inside the customer’s environment, but was employed with an AI vendor. Their job is to move AI out of the pilot and into real workflows. Insight launched the practice on October 5 with an advisory council that includes leaders from Microsoft, Salesforce, SAP, ServiceNow and Databricks. So the big vendors are paying attention.
The role is growing fast. FDE job postings grew 1,165% between the first ten months of 2024 and the same stretch of 2025. That’s growth off a small base, and it still tells you where vendors are putting their money.
Why now? Insight’s announcement cites an IDC number: 88% of AI proofs of concept never reach broad rollout.
That number comes from IDC research with Lenovo, where only four of every 33 proofs of concept reached production. IDC’s explanation was “the low level of organizational readiness.”
And I get the appeal. The vendor sends in someone who knows the product cold to work alongside the customer’s team.
But…What Happens When the FDE Leaves?
Here’s the thing. The FDE is on loan. At some point they roll off to the next customer.
Insight’s own look at four FDE teams describes it this way: “The FDE comes in for a specific problem and moves on; the CSM and sales rep provide continuity.”
Both of those roles work for the vendor. So who provides continuity on the customer side?
Then what? Who owns the workflow when the process changes? Who notices that the team next door has the same pain point? Who tells the vendor what’s working?
Like many organizations I talk with, the customer often ends up with a great deployment in one corner of the business. This points to impact without spread, where one team gets the win and everyone else keeps working the old way.
Gartner put a number on this in late September. They predicted that 70% of enterprises will abandon agentic AI systems built with vendor help by 2028. One reason they give is that customers don’t gain the knowledge and control to maintain the systems after the vendor leaves.
We learned this exact lesson with business process re-engineering a few decades ago. BPR taught us that reinventing processes without insider help leads to failed implementations.
An FDE brings deep product and technical skill. The organizational context lives with the customer: which process hurts the most, and who people go to when they’re stuck.
Pablo makes a related point in Optimize first, automate second: map the process and remove the waste before you apply AI. That’s Lean thinking, which is where I started, studying the Toyota Production System in Tokyo. The people who can map that process work for the customer.
The AI Lead as the Inside Partner
In the Hyperadaptive Model, AI Leads are peer champions. They sit inside a department and know its work, so their colleagues come to them.
LEARN MORE: AI Lead Accelerator
Take Greenfield Savings Bank. I shared an Uber back from a conference with their CIO, and she told me how she approached it. She picked one to two people per department to find the worst pain points and match one AI solution to each.
A biweekly accounting process went from eight hours to three to five minutes. (Eight hours!) As champions emerged, 50 of the bank’s 200 employees were pulled into the pilot. The full-bank rollout moved up six months because people were asking for it.
Greenfield did this with their own people. My hunch is that the FDE and the AI Lead belong together.
The FDE builds the solution with the AI Lead beside them. Then the AI Lead is the one who says, “Hey, here’s what we built in finance, and I think it fits what operations is fighting with.” When the FDE rolls off, the knowledge stays with the AI Lead and keeps spreading.
The people doing this work are already pointing here. On an Insight panel, Rajkumar Irudayaraj of Alteryx said the FDE should be “a permanent learning loop for the company.” I agree. A learning loop needs a named person on the customer side, and that’s the AI Lead.
Questions to Answer Before the FDE Arrives
If you’re the customer:
Who is the named AI Lead in the department where the FDE will work?
How will what they learn reach the other departments?
Who owns the workflow after the FDE rolls off?
If you run an FDE team:
Do you ask the customer for an inside partner before you start?
What do you hand that person on your way out?
The vendors are writing the FDE playbook right now. I’d love to see the inside partner written into it from the start.
Where to get the tough questions answered
If you want to work through questions like these with peers, the next cohort of Running Hyperadaptive Organizations meets December 8 and 15.
If you lead an FDE team, I’d like to compare notes. When your engineer rolls off, who on the customer side keeps it alive?
Sources & Further Reading
Demystifying the forward deployed engineer, Insight Partners
A day in the life: How four companies run FDE motions, Insight Partners
Optimize first, automate second, Pablo Dominguez
7 in 10 enterprises expected to abandon vendor-built agentic AI by 2028, The Register