"Stop sending me engineers": a CIO.com op-ed argues enterprises need backward-deployed domain experts, not more FDEs
In a CIO.com opinion piece, Praveen Jonnala argues the labs' FDE push is built to convert pilots to production, not to build lasting AI capability inside customers. His test for any engagement: when it ends, can the company find the next problem itself?
A CIO.com opinion piece published on October 13 by Praveen Jonnala takes the buyer's side against the forward deployed engineering boom. He grants the labs are right about the problem, citing low pilot-to-profit rates and the money now going into deployment arms at OpenAI, Anthropic, Microsoft and AWS. His argument is that those FDE programs sit on the commercial side of the vendors and are measured on pilots converted to production, which is different from getting AI to spread through the economy.
His example is a demand-forecasting agent that improves accuracy by double digits yet frees no cash, because the value only shows up when it changes decisions like safety stock, supplier commitments and build schedules. Outside engineers, he writes, can deliver local, siloed wins measured in weeks, but they cannot own trade-offs that cross functions and show up in cash two quarters later. He also warns that the learning runs one way: the vendor learns the customer's business while the customer learns the vendor's product.
For FDEs, the piece is worth reading as a picture of the skeptical buyer you will meet. It lines up with other critiques fde.me has covered, from Gartner's prediction about abandoned vendor-built agents to calls for contractual handover. His proposed check for any engagement is whether the client is better at finding the next problem itself once the engineer leaves. FDEs who leave behind domain context, evals and internal ownership, not just a working workflow, are the ones who pass it.
Source: www.cio.com