The Agentic Agency: How We're Rebuilding a 28-Year Business for the AI Era
By Adrian Arroyos
The Moment Everything Changed
Almost 3 decades. That's how long Catalyst Studio has been in business. We've survived recessions, platform shifts, the social media gold rush, and the "death" of print. But nothing prepared us for what happened when we started using AI. That's when I realized our old processes were unsustainable.
I was spending 6 hours a week on client proposals that had a 20% close rate. My co-founder and creative director, Maggie Kattan, was carrying the creative load while I played catch-up coordinating our dev team through the project delivery lifecycle. We were grinding harder than ever, but our business model was strained. What had once thrived, up until the COVID-19 reset — had become too burdened by a fractured ecosystem of tools, manual processes, and bottlenecks that were entirely human.
So we started using early tools like GitHub's Copilot and ChatGPT. Not as a novelty. As a learning process and a bet on the future of AI assisted workflows.
What we discovered wasn't just a better way to write copy, accelerate coding, or evaluate strategy. It was a fundamental rethinking of how work gets done — and it forced us to have the hardest recalibration we've had in nearly three decades of partnership: What if we rebuilt everything?
This is the story of how we're doing exactly that.
Part I: The Problem with "AI-Powered" Agencies
In 2026, everyone's slapping "AI-powered" on their website. It usually means they're using ChatGPT to write first drafts, running MidJourney for mood boards, or automating some email sequences.
That's adoption. It's not transformation.
The real opportunity isn't using AI as a tool. It's redesigning your entire workflow around agentic AI — systems that can monitor inputs like email and Slack, make decisions based on context, execute tasks autonomously, and loop you in when they need human judgment and decision.
Think less "AI intern" and more "AI operations manager."
Part II: What We're Actually Rebuilding
Here's what the shift looks like in practice at Catalyst.
Proposals: From Time Sink to System
Before: A client emails asking for a quote on a new e-commerce site. We'd spend several hours (sometimes days) researching, writing, pricing, and formatting. The prospect goes silent, or they respond that they went with someone cheaper. Processes that consumed 15+ hours a month could still be profitable with a few wins, but the true cost was our time invested in every sub-task.
After: An AI agent monitors our inquiry channels, runs initial qualification on budget, timeline, and whether we're talking to a decision-maker, then generates a customized proposal with a pricing matrix. It schedules follow-ups or auto-archives low-fit leads. My involvement is about 60=90 minutes of review and adding the human touch. No matter how good the AI is, we treat it as a junior-level employee whose work requires approval, and customization that can only come from deep expertise across different industry segments. We are not out of the loop.
The time savings have been dramatic — I'm estimating around 75-80% faster delivery — and we're qualifying leads far better than before. We're still measuring close rate impact (currently improved by 12%), but the early signals are improving with each quarter.
Content: From Afterthought to Engine
Before: Blog posts got written once a month when we "had time." We had an SEO strategy, but client work was always prioritized and that strategy slipped. The output was inconsistent and the impact was negligible.
After: An AI agent researches industry trends, client pain points, and competitive gaps. It drafts blog posts, optimized for SEO based on keyword research. Everything queues for our review and approval. We're producing roughly three times the volume at higher quality, and for the first time, engagement is actually measurable.
Client Onboarding: From Chaos to Choreography
Before: Manual, real-time kickoff emails. Scheduling back-and-forth. Repeated questions about brand guidelines. Lost files, missed details. Clients were too polite to say it, but the perception was disorganized. Each onboarding cost us about eight hours, or longer with enterprise clients.
After: Automated welcome sequence with personality — Maggie was insistent on this, and she was right. Dynamic intake forms that adapt to the project type. Brand asset collection with AI-powered organization. Calendar scheduling with context awareness.
The feedback we've gotten: "Most professional onboarding we've ever experienced." That's Maggie's creative sensibility running through automated systems, which is the whole point.
Part III: The Agentic Mindset
Here's the shift that matters most:
Old thinking: "How can AI help me do this task faster?"
New thinking: "What if this task never required me in the first place?"
The framework we're using has three steps.
First, map every workflow. List literally everything you do repeatedly. Client onboarding. Proposals. Status updates. Invoicing. Content. All of it.
Second, classify by human necessity. Some things need no human at all: data entry, scheduling, status updates, first-draft content. Some need a human in the loop: client-facing decisions, creative direction, business impact decision trees, final approval. And some are human-essential: strategy, relationship building, taste.
Third, automate the first category, augment the second, and own the third.
Most agencies stop at category two. That's where the time leak lives. Category one is where you get your life back.
Part IV: The Honest Truths
Let me be straight about what this actually takes.
You Will Break Things
Our first agentic workflow sent a client an email with {{COMPANY_NAME}} in the subject line and inserted a timely tip for the incorrect industry. Embarrassing? Absolutely. But it was a one-time debugging lesson, not a recurring salary line item. You fix it, you move on, and the system never makes that mistake again.
Your Team Will Need to Evolve
"AI is going to take my job" is a real fear, and you owe people a real answer. Ours is: "AI will take the parts of your job that didn't light you up. We're investing in your growth. Your value will only increase when you focus on your unique talents and skills."
Not everyone will make the leap at the same pace, and that's a conversation you'll need to have honestly. But in our experience, the people who lean in become significantly more valuable — not less.
You'll Question Everything
Once you start seeing what's possible, you'll realize how much of your current process is legacy habit masquerading as best practice. That realization is uncomfortable. Lean into it anyway.
The ROI Is Exponential, But Delayed
Month 1: Chaos. Debugging. Frustration. "How is this ever going to work?"
Month 3: Small wins. Time savings. Cautious excitement.
Month 6: Systems humming. More output, less stress.
Month 12: You're operating in a different weight class.
Part V: The Real Competitive Advantage
Here's what I keep telling other business owners:
The window is closing, but it hasn't closed yet.
Right now, most businesses are paralyzed by hype. They're waiting for "best practices" to emerge or for someone to tell them it's the right time. Meanwhile, every month you spend building agentic systems is a month your competitors spend debating which AI tool to buy, wondering if they should hire a consultant, or doing things manually because "it works fine."
The advantage compounds. We're not ahead because we're smarter. We're ahead because we started.
Part VI: Where We're Going Next
Our roadmap for the rest of 2026 is ambitious, and I want to be transparent about what's in progress versus what's planned.
Agentic client delivery is actively in development. We're building AI-driven project management so clients get real-time updates without us manually generating status reports. Early prototype is running internally now.
Predictive pipeline management is next in the queue. The goal is AI that predicts which leads will close, which clients are at risk of churn, and which projects are trending over budget. We're scoping this but haven't built it yet.
Autonomous service expansion is the strategic bet I'm most excited about. We were hit or miss with offering video editing or dedicated social management profitably before. With agentic workflows, we can scale services without scaling headcount. We've started testing this with content services for existing clients.
Full attribution stack is our long-term product play. Every dollar a client spends on marketing tied directly to revenue. We're calling it AttributionPro internally, and it's in early R&D. It's the difference between being sold on vanity metrics and show me the money. More on this as it develops.
How to Start
If you're a business owner or agency leader, here's your Monday morning plan.
Week 1: Audit. Map your top 10 recurring workflows. Time how long each takes per week. Calculate the hourly cost — your rate multiplied by hours. You'll be surprised, and probably a little angry, at the total.
Week 2: Pick one workflow. Start with something low-risk but time-intensive. Proposals, client onboarding, or content drafting are strong first targets.
Week 3: Build or buy. On the DIY side, platforms like Make.com or n8n paired with the Claude or GPT API and your email or Slack integration can get you surprisingly far. On the productized side, look at the emerging agentic orchestration tools — the space is evolving fast, so evaluate what fits your stack.
Week 4: Test, break, fix, scale. Run it in parallel with your manual process. Fix the bugs. Cut over once you trust it.
Month 2 and beyond: Repeat. Add another workflow every two to four weeks. Compound the time savings.
The Bottom Line
Catalyst Studio isn't just surviving the AI era, where agency services are losing ground to vibe coding and vibe marketing. Because of our Agentic AI pivot, we're on offense – helping clients see real ROI through integration, automation, personalization, and speed.
But I want to end with honesty: we're still figuring it out. Every week we break something, learn something, build something better. Technology is only as good as the human judgment steering it. That's the differentiator in every one of our client engagements. If you use ChatGPT like most people, you're getting the predictive output from a Large Language Model (LLM) that is optimizing for average outcomes. That's not what you want. With the right AI Consultant/Partner, you're optimizing for your own business objectives.
The difference between us and most agencies isn't that we have it all figured out. It's that we decided to rebuild while the building was still standing, and we're documenting the whole messy process along the way.
If that resonates, I'd love to hear what you're building.
Adrian Arroyos
Co-Founder & Digital Solutions Architect, Catalyst Studio
catalysts.net