The 17-point gap shaping AI confidence

October 5, 2026

AI confidence goes up the further you sit from delivery. 60% of C-suite leaders say they’re very confident in their organization’s AI readiness. For VPs? Only 49%. And for the directors and managers shipping the agents, it goes down to 43%. A 17-point gap that’s consistent across the board.  

That’s because no one’s looking at those programs through the same lense. Leadership is looking at where the technology is heading, while those building it are focused on that failure the team spent last night patching.  

Shipping an AI agent is hard, but production is where things get uncomfortable. That’s where these agents meet the reality of customer data: it’s messy, and full of gaps and inputs that no existing prompt is ready to address. Those are the challenges hiding in the 74% rollback rate Sinch’s AI Production Paradox study reported last May.  

And when those rollbacks happen, they hit AI teams harder than they hit their leaders. It’s the late-night call, the race to trace down the root cause, the work to fix it. By then, the roadmap everyone felt confident about is set, way before production could teach anyone anything.  

Both sides have the numbers, yet they’re looking at them from different perspectives. The C-suite’s confidence is high, but they’re not the ones rebuilding the agent after a rollback. Proximity to the consequences lowers confidence. That’s how the confidence gap gets baked in. It’s not something a new dashboard can fix.  

In June, AWS invested $1 billion into forward deployed engineering (FDE), the hottest new role in tech. Microsoft made the same move in July: $2.5 billion and 6,000 industry and engineering experts went into its Frontier Company. 

The role is in high demand, and it’s not limited to tech giants. The number of job postings for FDE roles is exploding, and some companies have even started building the role internally. Uber’s CTO had 30 of its top AI engineers sit in cross-departmental “Agentic pods”, so they could get closer to the processes they were aiming to automate. 

The idea underneath the hiring is simple: Someone who can ground the leadership view in the reality of production and also support the 2 a.m. rebuild when something breaks. Someone to force both perspectives into one room. 

Is part of the FDE craze just froth meant to justify AI capex? Probably. But people make people feel comfortable. Having a person in the middle, someone to have the uncomfortable conversations the AI failure rate forces, doesn’t sound like a bad idea. 

Whether brought in or in-house, what a shiny job title alone can’t earn is trust. Trust gets built from both sides believing that person is actually working toward the same goal. The moment the builders suspect the FDE is just there to grow the vendor’s bill or sell the board on something new, that trust breaks.  

And that person alone can’t make up for broken data, infrastructure that lets AI agents leak personal information, or rollbacks that can’t get diagnosed because there was no audit trail.  

Infrastructure is the thing you have to fix regardless. That’s what defines what your AI can do next week, or next year. But closing a confidence gap is something only a human can own, no matter how they’re labeled. Right now, in most organizations, nobody does.

In Production is a monthly editorial newsletter on AI in customer communications, written and published by Sinch. Read past issues on Sinch.com and catch the newsletter on LinkedIn.