Owner-led agencies think they’re ahead of their peers on AI. More than half of respondents in our recent SAGA Agency AI Survey of 62 agency owners and senior leaders said they were somewhat or far ahead of agencies like theirs. Only 13% said they were behind.
That math doesn’t work.
Either we’re seeing self-flattery or the bar for “ahead” has been set so low that opening ChatGPT a few times a week qualifies. I suspect both may be true.
The rest of the survey shows that the confidence may be misplaced, and that matters more than where any individual agency lands on a measuring stick. The full briefing is available for download.
Adoption is broad. Depth is not.
Nearly nine in ten respondents say they’re using AI regularly or widely. Three out of four use it personally every day. None reported using no AI at all. Whatever debate the industry was having two years ago about whether to adopt AI is over.
But scratch the surface and the picture changes. Only 29% report widespread use across the agency. The other 60% describe their use as regular in several areas, which is another way of saying in pockets.
That shows a real gap between using AI and integrating it across the agency.
The use cases tell a similar story. Research, brainstorming, summarizing meetings, drafting, and editing dominate. Each of these is one person at a desk going faster on existing work. Almost no one named creative and design or new business as the area where AI has delivered the most value. Agencies are accelerating what they already do, not reimagining what they offer or how they deliver it.
The impact data confirms this. Nearly nine in ten report a positive effect on productivity. Only 27% report a significant positive effect on quality, and 15% report no real effect on quality at all.
AI is making agencies faster more than it’s making them noticeably better.
There’s nothing wrong with being more efficient. But going faster alone is a short-term play. The agencies that will be in the best shape three years from now are the ones using this period to do things they couldn’t do before, not the ones using it to shave 20% off the time it took to do what they were already doing.
Your team is probably further behind than you are
Owners rate themselves highly on AI knowledge. More than four in five call themselves moderately or very knowledgeable. They rate their teams less generously. Among respondents with employees, more than a quarter say their teams are slightly or not at all knowledgeable.
If you’ve integrated AI into your own work and assumed your team has done the same, the survey suggests you should check that assumption.
The conductor metaphor that gets used a lot in AI conversations only works if there’s actually an orchestra. If you’re the only one who knows the music, you don’t have an agency that uses AI. You have an owner who uses AI in an agency that mostly doesn’t.
This is fixable, but it requires real effort. Training, shared workflows, time set aside for experimentation, and clear examples of what good use looks like.
None of which happens on its own.
Time once again an obstacle to progress
When we asked what’s holding back respondents from expanding their AI use, 42% said lack of time to learn and implement. That was the runaway answer. Cost, ethical concerns, client resistance, and skill gaps all came in below 5%.
Time often serves as a proxy for priority. When something hasn’t actually become a priority, we don’t find time for it.
If meaningful AI deployment were treated as a core business requirement rather than a nice-to-have side project, time wouldn’t be the bottleneck. The fact that 42% of owner-led agencies are blocked by available time tells you that AI is still getting squeezed around the edges of the business at most agencies, even at the ones that use it daily.
The pricing risk most agencies aren’t watching
Interestingly, the oft-cited worry that clients will demand lower fees doesn’t seem to be materializing, at least so far.
One question asked whether clients have requested lower fees because of AI. Almost nine in ten said no. That’s worth noting. The client-driven AI discount conversation is mostly not happening yet at owner-led agencies.
But that doesn’t mean all is well in the world of agency profitability.
Today’s AI tool prices reflect a market still being heavily subsidized by startups eager to claim market share at any price. As subsidies unravel and providers price toward what their services actually cost, AI expenses for agencies will start heading up.
That doesn’t mean you need to charge clients separately for AI. Most don’t, and most have no plans to. But it does mean you need to factor in realistic AI tool costs into the pricing for renewals, upsells, and new business right now. The agencies that build their current cost base on this year’s prices (and usage levels) and don’t adjust will feel the squeeze on their P&L well before any client asks for a discount.
Experiment now, while it’s cheap
The same subsidized pricing cuts the other direction too. Right now, AI is unusually cheap to experiment with. That won’t last. The owners who use this window to try things, fail at things, learn, and build genuine capability across the agency will have a real lead, not an imagined one, when costs rise and the competitive bar moves up.
The SAGA Agency AI Survey data shows a community that’s engaged, optimistic about AI’s direction, and using it daily. That’s the right starting outlook. What’s missing for most respondents is the next step: treating AI as something that reshapes the work, not just something that speeds it up.
Believing you’re ahead is comfortable. Acting like you’re behind is more useful.
Turn ideas into action
Audit the gap between your AI use and your team’s. Spend an hour this week comparing how often and how deeply you’re using AI to how often and how deeply your team is. If there’s a meaningful gap, that’s the first thing to address before any new tools or workflows.
Pick one service or workflow to redesign, not just accelerate. Identify one piece of what you do where AI could let you offer something different or more substantial. That might mean a new deliverable, a different service tier, or a reporting cadence the client couldn’t get from you before. Block an hour to sketch what that might look like.
Review upcoming renewals with realistic AI tool costs in mind. Rebuild the pricing assuming tool costs rise 25 to 50% over the contract term. Be sure to account for shifts in labor or other costs that come along with increased AI usage. If the margin won’t hold with your new projected costs, your pricing model needs adjusting now, not later.
