Trent Gillespie
July 09, 2026
A CEO asked me something today that I’ve heard a lot recently. His teams are buried, AI is clearly on the table, and he asked whether we could just send in a forward deployed engineer to sit with his people and fix their processes. Take it off their plate. He'd rather buy the outcome than go find the hours.
It might be the most reasonable question I get. After all, making AI adoption easier is what I do full-time, and it sounds like doing this is an obvious way to help. And yet the more I sat with it, the more I thought the answer was more interesting than a yes or a no.
If you haven't heard the term, "forward deployed engineer" is the hottest job in AI right now. In one six-week stretch this spring, Anthropic, OpenAI, AWS, and Databricks each stood up a forward-deployed operation, with AWS alone putting a billion dollars behind a unit it plans to seed with thousands of engineers. Postings jumped around 800 percent in a year, at comp running from $300,000 to well past $600,000.
Beyond the fact most companies can’t afford that investment, what I find strange is what these engineers are actually for. You’ve heard stats about projects not returning an ROI. What it shows now is that the problem isn’t that the models didn’t work, it’s that the deployments didn't. And four of the biggest names in technology seem to have looked at that same gap and concluded the bottleneck isn't the AI anymore. It's the last mile, the distance between a model that can do the work and a business that actually runs differently because of it.
So the CEO had put his finger on something real. I just think he had the direction slightly wrong.
Once you strip away the salary and the mystique, most of what an FDE does on a given day looks pretty unglamorous. They sit inside a real workflow, watch where the time and money leak out of it, and change how the work moves so that afterward the same job runs differently than it did before.
That's most of it. What they're really trading on is workflow judgment, understanding a process well enough to rebuild it around what AI can now do. It's the line between a consultant and an FDE, and I think it's a sharp one. A consultant writes you a report and leaves. A forward deployed engineer builds the actual thing and stays until it runs.
Here's what caught me, though. Re-engineering a workflow isn't really a computer science skill. It's operational judgment. It's knowing which step in your intake process is the one that clogs, and imagining what it would look like if AI drafted the first version instead of a person staring at a blank page. A leader who understands the tools can often see that as well as anyone flown in from outside. Which makes me think the scarce ingredient was never the engineer at all. It was someone paying deliberate attention to how the work actually flows.
Almost every leader I talk to tells me some version of the same thing. They don't have time to figure out AI.
I believe them. But then I sit with their week, and something odd shows up. It's full of exactly the tasks AI is now pretty good at absorbing. The email triage. The first draft of everything. The meeting prep. Pulling those numbers again.
So I've started to think the real sentence underneath "I don't have time to figure out AI" is closer to this: “I'm too busy doing the things AI could do to build the capacity to make AI do them.”
And hiring it out doesn't get you out of the loop. Bring in an outside engineer and you still have to choose which processes matter, still have to get your people to change how they work, still have to own the result once the engagement ends. The expert can raise your ceiling. They can't remove the floor, and the floor turns out to be your attention. I suspect that's the quiet reason so many of those projects stalled. Someone hoped they could skip the investment by writing a check.
The thing you fund is the thing you're serious about. Equipment, hiring, marketing, you don't scrape that together from whatever's left at the end of the month. You allocate it, on purpose, because it matters.
Almost nobody does that for improving how the business itself runs, and the numbers bear it out more starkly than I expected. The most rigorous read I've found is the Census Bureau's Annual Business Survey, which covers millions of US companies. In its most recent data, only about one in five made a significant change to a business process in a full year. Four out of five changed nothing about how the work gets done. And when they were asked what got in the way, the top answers were cost and a shortage of skilled people. Not "it can't be done." Resourcing.
So maybe the reframe is this. You don't have a time problem, you have a resource you haven't allocated yet. The leaders I see pulling ahead with AI didn't stumble onto a spare afternoon. They decided it was worth real capacity and then protected it, the same way they'd protect a hiring budget.
In practice, allocation looks like a person. You name someone inside your business whose actual job, with real protected hours, is to own making AI change how the work runs. Most people call this an AI champion, and I'd like to rescue the term a little, because it's drifted into LinkedIn wallpaper. A champion isn't just the person on your team who's excited about ChatGPT.
The version I'd argue for is three things. First, protected time, genuinely defended, not squeezed into the corners of a full-time role. Second, the real work of re-engineering workflows, except done by someone who already knows your business and where the bodies are buried. And third, a roadmap, so they fix the workflows that matter most in the right order, rather than whatever happened to annoy someone this week.
This, I think, is the forward deployed engineer the CEO was asking for. Not an outsider imported at $600,000, but someone inside, backed with time and authority.
There's a number I keep circling back to here. I've started calling it AI Velocity: how many of your workflows actually run differently than they did 30 days ago. Not how many tools you bought. For most teams it's zero, and I think it's zero for a simple reason. Nobody's job is to move it.
The AI industry just spent billions telling you where the hard part is. It isn't the model. It's the deployment, and deployment is a human investment, not a purchase.
1. Name your champion this quarter. Pick someone who knows your operation and give them real, defended hours to own AI deployment. Not evenings and weekends. Capacity you've decided to fund. I'd do it now because the tools finally reward it; the wall used to be the technology, and today it's almost entirely attention.
2. Re-engineer one workflow, don't just assist it. Have them pick a process that eats hours every week and ask what it would look like rebuilt around AI, not merely sped up with it. A faster bad process is still a bad process. Rebuilding it is where the time actually comes back.
3. Sequence with a roadmap. The failure mode I see isn't weak effort, it's effort spent on the wrong workflows in the wrong order. Decide deliberately which to fix first, based on where the time and money actually leak. This is the piece most businesses are missing, because sequencing takes seeing across a lot of workflows to know which one comes first.
The leaders who pull ahead this year won't be the ones who found more time. There is no more time. They'll be the ones who decided that changing how their business runs was worth funding on purpose, and named someone to do it.
Two bits of news I'm glad to share. Stellis AI is now the Strategic AI Partner of AMCI, the Association Management Company Institute, the body representing the firms that run associations behind the scenes. It's a group I have a lot of respect for, and the partnership means we'll be helping AMCs and the associations they manage figure out exactly the kind of workflow re-engineering this issue is about.
And in August I'll be at ASAE’s Annual Meeting, leading two sessions: on Saturday, the CEO Summit, where I'll help association executives build industry-specific AI strategy for their organizations; and on Monday, a session on how to take the most advantage of AI in their own roles. If you're going to be there, sign up, or reach out so we can meet!
Trent Gillespie is CEO of Stellis AI and a keynote speaker helping business leaders understand and operationalize AI in their companies. He spent almost nine years leading global innovation efforts at Amazon before leaving to help other companies build the capabilities they need to compete. Book Trent to speak to your group or book a call to discuss using AI within your business.
Tell us what you thought of today's email. |
Did someone forward this newsletter to you? If you're not already signed up, you can subscribe to AI SPRINT™ for free here.