How can your agency scale and thrive in the age of AI? The answer isn’t about the latest tips and tricks; it’s about taking a strategic approach—and connecting with other leaders.
In August 2025, I joined a small group of agency owners and executives for the third annual AI & Your Agency workshop in San Francisco. Unlike a big conference with hundreds or thousands of attendees, this was deliberately small—which meant candid conversations about what’s working, what isn’t, and what keeps owners up at night. I’ve recommended it since attending The Bureau’s first AI immersion event back in 2023.
A few themes stood out across two days of discussion:
- We’ve moved beyond the hype, but the path forward is uneven.
- Leadership in the AI era requires a different approach from that in the pandemic era.
- Agencies need to balance rapid experimentation with building a long-term economic or competitive “moat.”
- The most successful leaders aren’t waiting for a perfect playbook—they’re creating momentum despite uncertainty, one small win at a time.
This article isn’t about predicting the future of AI. It’s about preparing your agency to thrive no matter what the future brings. At over 3,000 words, pour yourself a beverage and let’s dig in!
Learning more via those in the know
The themes here span multiple speakers and audience members. That is, I can’t always attribute each point to the exact person who said it due to The Bureau’s confidentiality rules, which mirror the Chatham House Rule.
So you can follow their latest advice, here are the 2025 speakers and their session titles:
- Getting Found in the Age of Search, Social, and AI (Wil Reynolds, Seer Interactive)
- Maximizing AI While Mitigating Your Legal Risk (Gabe Levine, Matchstick Legal)
- Building AI Agents and Automations (Sebastian Chedal, Fountain City)
- The Messy Middle (Megan Notarte, AI Portland)
- No Code, All Flow: Automating Agency Deliverables Without Engineers (Mike King, iPullRank)
- Transforming Agency Operations Using AI (Brent Weaver & Manali Sharma, E2M)
I’d also recommend following original speakers Britney Muller and JP Holecka, who had last-minute conflicts this year.
Now, let’s dig into AI themes and practical tips for your agency
The messy middle: Why AI adoption feels stuck
One of the strongest through-lines at the event was a recognition that we’ve entered what speaker Megan Notarte dubbed the messy middle of AI adoption.
The hype cycle—via Gartner and otherwise—has shifted. We’ve moved past the “wow” demos of early tools, when it seemed like AI might instantly transform every part of agency life. But we’re not yet at the plateau of smooth workflows, universal adoption, and clear ROI. Instead, most agencies are wrestling with uneven progress, cultural friction, and plenty of frustration.
On the surface, usage numbers are impressive. Surveys find that nearly half of U.S. adults use AI at least occasionally. But inside agencies, adoption is patchy. For instance:
- Designers may be testing prompt libraries while project managers resist change.
- Account leaders want efficiency, but they also worry about errors showing up in client work.
- Deterministic QA testing doesn’t work with probabilistic models.
- And in many cases, individual “hacks” don’t scale beyond the enthusiastic person who built them.
That’s the messy middle: some breakthroughs, some wipeouts, and a lot of trial-and-error.
Keep going
The temptation is to wait it out until the technology “settles.” But agency AI experts at the workshop were clear: that’s a dangerous bet. The agencies that will thrive are the ones that experiment through the chaos. They’re willing to try, fail, learn, and move forward.
That means accepting a few realities:
- Hype doesn’t match reality. Tools often over-promise and under-deliver. It’s your job to sort through that, for your agency and your clients. AI is not a human-free solution.
- Solutions don’t scale instantly. What works for one power user won’t roll out agency-wide without training and structure.
- Frustration is normal. Team resistance, broken integrations, and half-baked features come with the territory.
The antidote? A mindset shift. Don’t expect to leapfrog directly to the “plateau of productivity.” Instead, start with specific, painful bottlenecks and test whether AI can help. Accept that your first attempts won’t be perfect and that no AI output is a finished product. That is, you can complete things faster—but you still need human participation. As Megan observed, embrace the first pancake. The messy initial try is the step required before you can make a full stack.
Leadership in the AI era: Bridging the owner vs. employee mindset gap
Another recurring theme was how leadership and people management must evolve in the age of AI.
Many agencies are discovering that the people they hired to stabilize operations during the pandemic may not be the right fit for today’s AI-driven needs. Back then, success meant maintaining consistency, systematizing delivery, and keeping things moving during unpredictable times. Today, success often requires speed, experimentation, and a willingness to rethink old assumptions.
That creates tension. Owners feel urgency to adopt AI because they see the competitive pressure. In contrast, many employees want to keep doing last year’s job a little better. I’ve written before about this owner vs. employee mindset gap—and it shows up again here, in how teams approach the adoption of AI. Owners are usually motivated by long-term growth and survival. Employees tend to be motivated by stability and avoiding mistakes.
Add misaligned incentives and the gap widens. A salaried manager might embrace an AI tool that shaves ten hours off their week. But an hourly employee might be less enthused about a tool that reduces their hours and introduces a myriad of unknowns for which they are responsible. Without clear leadership, even well-meaning team members may drag their feet.
Not everyone will cross the gap
The hardest truth leaders confronted at the event: some people won’t make the leap. You can coach, you can train, but you can’t force someone to embrace change if they fundamentally don’t want to. At some point, leaders must ask: Who do we keep, who do we coach, and who do we let go?
As one participant put it, “We hire and attract people for who they were when they joined. They became the employee for what the agency was, not where we’re going.”
Leaders who cling to pandemic-era expectations risk holding their agency back. Leaders who set clear expectations—and make tough decisions—open the door to growth.
I would also observe that AI is a catalyst; given how fast our industry changes, some change-resistant employees may not be a long-term match regardless. But you also have to give your team space to learn and explore, rather than expecting magical outcomes.
Your responsibility as a leader in the age of AI
Your employees won’t magically “get” AI—especially if you aren’t clear about your vision for the business and the role of AI in reaching your goals. They also won’t make progress if you expect them to become AI experts on top of their existing responsibilities. AI results require mindset, training, time to experiment, and a safe environment to make mistakes.
As an agency leader and manager, it’s your job to give your team the guidance and resources they need to get things done. AI is no exception.
This also includes recognizing that “finishing” something takes time. For instance, I used AI to draft this article—but then I made edits, received feedback from a human team member, and made more edits. You might have a first draft in five minutes—but the final draft might take a couple more hours. And if you’re building something more robust—for instance, that does mission-critical calculations—it’ll be more than a couple more hours.
While in San Francisco, I saw a relevant billboard for an AI tool designed for large codebases: “You just raised your Series B. Vibes won’t cut it.”
Why agencies should focus on innovation, not scale—at least for now
One of the most useful insights came from speaker Wil Reynolds, in reframing where agencies are right now: this is innovation and sales season, not scale season.
Too often, owners get stuck thinking, We can’t adopt new tools until we’ve built the perfect process around them. But the reality is that AI tools are changing too quickly for perfection. Waiting for the “right moment” means you’ll always be behind.
Instead, think of this as a time to experiment boldly—even if it creates some messiness. If your clients are ready to buy AI-related services, consider selling them the services they want and need.
Choose to incur intentional tech debt. That means allowing imperfect systems in the short term, because the learning and client value outweigh the cleanup costs. You can systematize later, once you know which workflows are worth scaling.
The lesson: resist the urge to build a pristine machine before you’ve tested what matters. Agencies that thrive in this season are the ones willing to experiment, iterate, and accept a bit of chaos along the way.
Building an economic moat your competitors can’t cross
Wil raised the idea of creating a proprietary moat. Often attributed to Warren Buffett, a business or economic moat is a durable advantage your competitors can’t easily copy.
AI tools themselves aren’t moats; they’re widely available and constantly changing. What matters is how you use the AI tools—including the proprietary data in your systems—and whether the result is something unique to your agency.
From the conversation, several agency leaders described business moats in three forms:
- Knowledge capture: When you consistently record ideas, processes, and insights, you create a proprietary knowledge base that AI can draw from. Think transcripts from client calls, highlights from research, or internal playbooks. When indexed and accessible, that library becomes an internal advantage—because your team has better raw material than anyone else.
- Memory: New features like ChatGPT’s memory or Model Context Protocols (MCPs) allow tools to “remember” past interactions. Instead of starting from scratch each time, your AI workflows build cumulative intelligence. Imagine a custom assistant that already knows your brand voice, your positioning, your service templates; no generic prompt can replicate that.
- Trust: Perhaps the most overlooked moat is client trust. Agencies that consistently deliver with AI—while maintaining accuracy, brand voice, and human judgment—become trusted advisors. Over time, that trust in the implementation of AI becomes a competitive differentiator beyond the technology itself.
Building a moat requires intention. It’s not about chasing the next flashy tool. It’s about capturing, systematizing, and safeguarding what makes your agency valuable. Or as one participant put it, every book, article, and transcript is a database. The moat is in how you use the data.
The agencies that build moats now can win clients today—and defend their position when everyone else eventually catches up. And that means centralizing operations, rather than (for instance) each team member building custom GPTs individually.
Gauging maturity and measuring impact
One of the clearest frameworks came from looking at AI maturity as a progression. Rather than thinking of adoption as binary—you’re “in” or “out”—it’s more useful to think of it in stages.
Speaker Brent Weaver and his colleague Manali Sharma identified four stages of AI adoption:
- AI Aspiring: Curious, but not using AI yet.
- AI Curious: Limited or experimental use, often driven by individual team members.
- AI Enabled: Selective, intentional adoption across workflows, with leadership awareness.
- AI First: Integrated across the business, where AI is part of the operating system.
Most agencies today sit somewhere between “Curious” and “Enabled.” The leaders are pushing toward “AI First.” Knowing where you are helps you set realistic expectations for what comes next and how to move up.
Monitor productivity
Speakers shared stats about how AI has promise—but it doesn’t always have a measurable impact.
My recommendation? Monitor your Revenue per Full-Time Equivalent (Rev/FTE). This simple KPI reflects productivity, efficiency, and value creation.
- For acquirers, it demonstrates you’ve built a scalable agency, which can help you secure a higher EBITDA multiple.
- For clients (although they won’t see it directly), it signals whether you’re delivering value without bloating headcount. You can show them how they’re getting more for their money.
- For you, it’s a simple, comparable way to tell (in a distilled financial metric) if AI is making your team more productive.
Put bluntly: if AI adoption isn’t improving Rev/FTE, it’s not creating real value. You may have novelty, but you don’t yet have impact. For more, see my deep dive on Rev/FTE and why it matters for agency growth.
This pairing—maturity model plus productivity metric—gives you both a map (where you are now) and a compass (how to measure progress). That combination makes AI adoption less about hype and more about long-term business health.
Practical ways to focus now: 8 tips
In the messy middle, the best move isn’t to overhaul your agency overnight. Instead, make small, targeted changes that prove value and build momentum.
Here are seven practical steps you can take right away—and read on for an eighth:
- Run an “It sucks that…” session. Instead of asking your team, “How should we use AI?”, ask them, “What’s hard about your job?” Gather a list of the most frustrating tasks, then test whether AI can help with those. This approach reframes AI adoption as relief, not extra work.
- Embrace the “first pancake.” Your first workflow will be messy. Ship it anyway. You’ll learn more from a clunky prototype than from months of debate. At this point, momentum matters more than mastery.
- Build lightweight guardrails. Don’t wait for a 40-page policy manual. A one-page checklist is enough to start, to reassure both your team and your clients:
- What’s safe (and unsafe) to feed into tools?
- What level of human review is mandatory?
- What’s your commitment to accuracy, inclusivity, and brand voice?
- Default to RAG for content. Generic AI output isn’t good enough. Use retrieval-augmented generation (RAG) by feeding the AI your proprietary assets—brand guidelines, case studies, positioning docs. That way, the output is grounded, on-brand, and less likely to hallucinate. For more, see the latest articles from speaker Mike King.
- Choose technical debt on purpose. Move fast and accept some messiness. Don’t over-engineer systems that may change in a few months. Systematize once you know what works.
- Keep humans in the loop. Plan from the start where you want human oversight: accuracy checks, brand nuance, and high-stakes client work. AI can draft, but people should decide.
- Track Revenue per FTE. If AI is boosting productivity, your Rev/FTE will rise. It’s the cleanest metric for productivity—and it helps whether you sell or keep your agency.
Try this exercise with your team
As an eighth tip, I liked these exercise prompts from Brent Weaver:
- AI Quick Wins (free up your time): List 3 tasks that you repeatedly spend 4-8 hours a week on and could be handed off to an AI agent. To free up your time, these should be low-effort, high-impact tasks.
- AI Big Wins (boost your revenue): Write 3 use cases where AI could help you significantly increase revenue or client satisfaction. These tasks should focus on revenue-generating activities.
Individually, these tips aren’t silver bullets—they’re starting points. But the agencies that apply them will learn earlier, move faster, and build confidence across their teams.
Turning AI into client value, not just efficiency
Much of the conversation about AI focuses on internal efficiency—saving time, automating tasks, and reducing overhead. That’s important, but the bigger opportunity for agencies lies in how AI can enhance client-facing value.
Here are some AI success stories the speakers and audience members shared:
- Competitive research at scale: What used to take a week of manual digging can now be completed in a half-day with AI’s help. Instead of analyzing five or six competitors, one team shared they could now evaluate 25 in the same time—delivering deeper insights and stronger strategy for clients.
- Site audits across hundreds of properties: A project that seemed unmanageable suddenly became doable when AI handled the bulk of the review. Human team members still verified accuracy, but they automated the heavy lifting. The result: faster turnaround, broader coverage, and a clearer value proposition.
- Instant client Q&A: Agencies often lose momentum when clients ask data-driven questions, but the right team member isn’t available. By connecting client datasets to AI-powered queries, one agency ensured clients got answers immediately—no delays, no bottlenecks. That speed built trust and reduced friction.
These aren’t just efficiency plays. These examples change how clients perceive your agency—as long as you highlight the benefits. When you deliver quality insights faster, anticipate their questions, and remove obstacles from the relationship, it pays off. You reinforce your position as a strategic partner, not just a vendor.
The real competitive edge comes when AI helps you serve clients better than they thought possible. And that’s better than being known as “cheaper via AI.” It’s up to your agency to make the case to clients: AI is there to make things better, not cheaper. And if you’re still doing hourly pricing, it’s time to change.
Risks and guardrails: Protecting your agency as you scale
Of course, speed and experimentation come with risk. A clear takeaway from AI & Your Agency was that agencies can’t outsource responsibility for AI governance.
The technology is moving faster than regulators, insurers, or industry standards. That means agency leaders must create their own frameworks to protect both their team and their clients.
The good news: this doesn’t require bureaucracy. Even lightweight guardrails can go a long way:
- Acceptable-use guidelines: Be explicit about what data your team can (and can’t) feed into AI tools.
- Human-in-the-loop reviews: Require human sign-off on all client-facing outputs.
- Bias and voice checks: Build in quick reviews for tone, inclusivity, and brand alignment.
- Testing protocols: Non-deterministic tools need ongoing QA, not a one-time test.
- Broader impact: Set standards for quality (so that you aren’t putting “AI slop” out into the world) and help the team work efficiently (via things like clean data sources and prompt pre-vetted prompt libraries) to reduce the environmental impacts of your AI work.
Beyond process, leaders need to consider risk transfer. Several agency owners noted that insurance carriers are beginning to require AI coverage riders. One attendee said that ChatGPT costs him $20 a month per user… and an extra $1,000 a month in insurance premiums. But it’s a fraction relative to the coverage value. Talk to your insurance agent to see what they recommend for your unique situation.
Governance and compliance aren’t just “corporate” issues. As speaker Gabe Levine made clear, there are practical questions about liability, client confidence, and long-term reputation.
The bottom line is that adopting AI without guardrails is like speeding without seatbelts. Agencies that move fast and protect themselves, with fewer wrong turns along the way.
Finding your own path forward
The conversations at the event reinforced what I see every week in my advisory work: there’s no one “right” way to adopt AI in your agency. Best practices are useful, but the best path is the one that works for you. It should fit your leadership style, your team’s readiness, and your clients’ expectations.
The messy middle is frustrating, but it’s also fertile ground. The agencies that thrive won’t be the ones with the fanciest tools; they’ll be the ones where leaders make clear decisions, test relentlessly, and build durable advantages their competitors can’t easily copy.
Thanks to Carl Smith and Meredith Durham—and The Bureau sponsors—for making the event possible. It’s time to align your leadership team, free yourself from bottlenecks, and prepare your agency for the next chapter.
Need a step-by-step path forward? Consider an Agency Growth Diagnostic (AGD). The project is a structured assessment I use with agency leaders to identify what’s working, where you’re stuck, and how to set your business up for long-term success. Rather than an AI evaluation, I focus on your business as a whole. You don’t need a playbook for AI—but you do need clarity about your agency’s future.
Question: What’s next as you enlist AI to help scale your agency?
