Why 78% of AE Firms Use AI But Only 38% See Results

I've been watching this play out inside AE firms all year. Everybody's adopting AI, but hardly anyone can point to what it's actually done for their business. This year's Deltek Clarity data put a number on it: 78% of AE firms are now using generative AI, but only 38% report a measurable business impact. That gap didn't surprise me. I've seen it firsthand, spending months running my own daily workflow through AI, not because someone told me to, but because I wanted to know what it actually takes to make it work. And it's consistent with what's showing up across business in general, not just architecture and engineering.

That 40-point gap is the real story this year. Nearly every firm has adopted AI at this point. What separates them is what happens after you turn the tool on.

What's actually causing the 40-point AI adoption gap?

I don't think this is a technology problem. I think it's an operating model decision.

Firms that land in that 38% aren't using a fundamentally different AI tool than everyone else. They're operating with intent, and everybody else is operating on vibes. That distinction matters more than which product you bought.

What are AE firms who are seeing AI results doing differently?

Three things, and none of them are exotic.

First, they're picking workflows, not features. The question isn't “where can I use ChatGPT?” That's backwards. The question is: “what real business problem or workflow am I trying to improve, and how does AI actually enhance it?” Start from the problem. Let the tool follow.

Second, they're investing in the data underneath it before they invest in the AI on top of it. AI on top of clean, connected data is leverage, full stop. AI on top of disconnected spreadsheets is chaos at best. A meaningful share of that 38%'s AI budget went to work that isn't glamorous: cleaning up project taxonomy, connecting business development to delivery and resourcing, getting the foundation actually stitched together. That's the unsexy answer, but it's the one that moves the number.

Third, they're measuring what they're doing. A lot of AI adoption right now runs on gut feel. The firms seeing impact know their AI is saving four hours per proposal or cutting project setup time by 60%. The rest have adoption with no measurable impact because they're not measuring anything. That means they can't tell if a tool is working or tell when to cut it loose and move to something else. You need to fail fast here, and you can't fail fast if you're not tracking anything in the first place.

What does AI as a capacity multiplier look like in business development?

Business development is a good place to watch this play out, because it's where AI adoption showed up first and hardest.

The advantage isn't that a firm “uses AI in BD.” Most firms claim that now. The advantage is what AI lets a small BD team actually do: pre-qualify five times more opportunities, produce proposals that actually reflect the specific context of the client instead of last quarter's boilerplate with a new logo on top, dig deeper into market intelligence instead of just skimming it because there was never enough time to go deeper.

The firms gaining ground here are treating AI as a capacity multiplier, not a productivity tool. That's an important distinction. Productivity means I get my existing workload done faster. Capacity multiplier means a four-person BD team can now operate like a ten-person one. That's not incremental. That's a competitive moat.

How can AE firms build AI fluency in 30 days?

AI fluency isn't a curriculum, it's a habit, and you don't read your way into building one. You have to actually build something.

Here's an exercise that any firm can run in 30 days to test this out; no training program required: pick one workflow per role. Proposal drafting for business development. Project status for project managers. Client communication for principals. Give people explicit permission to use AI on that one workflow during normal work hours, not as a side experiment squeezed in after hours, and put a 30-minute, “what worked, what didn't” session on the calendar once a month. That's it. Practice beats programs every time here.

And one more thing, because I think it's the piece firms skip most: make AI someone's job. Stop making it everyone's hobby. If nobody owns moving this forward, it doesn't move forward — it just sits there as a subscription line item.

Which brings me to the thing I'd actually tell firms to stop doing.

Stop buying more AI tools. I talk to firms that have more AI subscriptions than they have outcomes to show for them. The 38% seeing measurable impact aren't running more tools than everyone else. They're running less and using them better. Before you add another one, go measure what the ones you already have are actually doing. Then double down on those.