What Happens When Your Agency's SOPs Finally Have Teeth with Andy Janaitis | Ep #910

Most marketing agency SOPs fail for one reason: they are static documents in a dynamic business. In this episode, Andy Janaitis breaks down how his digital agency built an AI-powered context engine that transforms SOPs from ignored documentation into living operational infrastructure. By combining shared agency context with personal workflow context — and introducing approval-based updates instead of open editing chaos — his team created a system that continuously improves how the agency operates.

The deeper lesson is not about AI tools. It is about operational leverage. Agencies do not scale because they document processes. They scale because systems reinforce behavior without founder intervention.

What You'll Learn

• Why most SOP systems fail after initial implementation
• How to structure shared context versus personal context inside an AI-driven workflow
• The operational difference between static documentation and living systems
• Why founder bottlenecks often come from unclear outcomes, not bad team members
• How to start implementing AI operational systems without overengineering everything
• The importance of approval-based knowledge management inside growing agencies
• Why giving AI outcomes instead of tasks dramatically improves output quality
• How context-rich systems reduce dependency on founder oversight

Key Takeaways

• Documentation alone does not create operational consistency. Reinforcement systems do.
• Shared AI context needs governance. Without ownership and approval workflows, systems become unreliable.
• Most agencies fail with AI implementation because they try to build an entire operating system at once instead of solving one friction point first.
• AI amplifies briefing quality. Weak context produces weak output.
• The founder’s role shifts from task management to system architecture as the agency matures.
• Outcome-based communication improves both AI performance and team performance.
• Operational leverage comes from structured context, not more meetings or more SOP folders.

Have you ever written a process that nobody followed? Or built a folder of SOPs that your team politely ignored and you quietly stopped updating?

That was a big struggle for today’s featured guest, but six weeks before this conversation, he and his team built something that solved a problem most agency owners have tried and failed to fix for years: an AI context engine that makes their operating procedures actually stick. In this episode, he walks through exactly how it works, how they structured shared and personal context layers, how to get your team started without overwhelming them, and why giving AI an outcome rather than a task is the thing most founders are still getting wrong.

Andy Janaitis is the founder of PPC Pitbulls, a boutique digital marketing agency focused on Google Ads and Meta Ads for small to medium businesses. His background is in industrial engineering, data science, software engineering, and product management. Throughout these different stages of his career, he always worked at agencies. So naturally, when it came to starting his own business that seemed like the obvious choice.

He launched the agency in 2020 alongside a former colleague, the same week his first child was born and COVID hit. PPC Pitbulls' differentiator is measurement: every ad dollar is tracked, client behavior on-site is understood, and optimization follows the data rather than intuition.

In this episode, we’ll discuss:

• Andy’s solution to the common owner SOP problem

• Shared context vs. personal context

• Get next-level results by providing outcomes, not tasks

Subscribe

Apple | Spotify | iHeart Radio

Sponsors and Resources

This episode is brought to you by Wix Studio: If you’re leveling up your team and your client experience, your site builder should keep up too. That’s why successful agencies use Wix Studio — built to adapt the way your agency does: AI-powered site mapping, responsive design, flexible workflows, and scalable CMS tools so you spend less on plugins and more on growth. Ready to design faster and smarter? Go to wix.com/studio to get started.

The SOP Problem Most Agencies Have Given Up On

Every agency owner knows the rhythm. You write the process. You put it in ClickUp or Notion or a shared drive. You announce it to the team. Three months later nobody is using it, and you are back to making every decision yourself because it is faster than watching the system fail in real time.

Andy has run this loop and now, just six weeks before the recording, managed to use AI to create a tool that changed everything.

It was an AI context engine that pulled from every client touchpoint, including meeting recordings, email, and Slack, and converted that information into living context files the team can query in real time. The key detail is what happens when someone wants to update a shared file. Every central skills file has an owner. Changes get queued for approval rather than overwriting existing rules. What used to be a static document that slowly went stale is now a system that learns, updates, and actually enforces how the agency operates.

Shared Context vs. Personal Context: Why the Distinction Matters

The context gathered in this way is structured across the team in two tiers:

• First tier: The central bank holds client context, agency-wide skills files, and general operating rules. That lives in a shared Google Drive folder that auto-syncs to every team member's desktop.

• Second tier: Personal context, meaning individual rules that only apply to a specific person's workflow, like filtering certain emails that have nothing to do with the agency.

The reason this distinction matters is that most teams building shared AI context run into one of two problems: the files are so locked down nobody updates them, or they are so open that updates overwrite each other and nothing is reliable. The queue-and-approve structure Andy built threads that needle. Team members can flag a better way to do something. The file owner reviews it. If it makes sense, it gets merged into the main store. The agency gets smarter without the chaos of everyone editing the same file in real time.

Start With One Specific Thing, Not the Whole System

Most founders decide to build an AI operating system and then make the mistake of trying to build everything at once, load too much context into a single document, and end up with a system so heavy it cannot function efficiently. Jason describes his own early version as trying to get every person in the company to approve a single letter change. The architecture was right but the structure was wrong.

Andy's starting point recommendation is specific enough to actually follow:

• Pick one workflow. The one that creates the most friction or the most inconsistency.

• Open Claude desktop, describe what you want, identify the tool or source you want to pull from, and ask it to build a file structure that keeps client context organized and retrievable.

The plan it generates is not perfect. That is fine. You approve, adjust, and run it. From that first working piece, everything else becomes an iteration. The common mistake is waiting for a complete vision before starting. The agencies making real progress right now started with something small six weeks ago and have been adding ever since.

Give It an Outcome, Not a Task

The tactical shift that runs through this entire conversation is the difference between assigning AI a task and giving it an outcome. A task is "write me a sales proposal." An outcome is "we need to win this client, here is everything we know about them, here is our agency's positioning, here is what a strong proposal from us looks like, produce a first draft." The output from the second prompt is not in the same category as the output from the first.

This is the same principle that makes or breaks the first few hires at a growing agency. Most founders who have struggled with underperforming team members can trace it back to the same root: they handed someone a task without ever communicating the outcome they were trying to reach. AI amplifies both good and bad briefing habits instantly. Give it strong context and a clear destination, and it operates well above expectations. Give it a vague instruction and ignore the output quality, and the tool looks broken when the real problem is the brief. Building the context engine is how you make that outcome-focused briefing the default rather than the exception.

Do You Want to Transform Your Agency from a Liability to an Asset?

Looking to dig deeper into your agency's potential? Check out our Agency Blueprint. Designed for agency owners like you, our Agency Blueprint helps you uncover growth opportunities, tackle obstacles, and craft a customized blueprint for your agency's success.