Process is the New Code

I’ve been working a lot with AI Agents over the last two weeks. It’s been quite the circuitous route. A few years ago when I started consulting again I created a lot of the tools to track clients, progress, tasks and onboarding. I built it all in the Google ecosystem using Google Sheets, Docs and Forms. I did that on purpose, because I wanted to customize my process according to how I worked. All of my information was already there in Gmail and Google Drive. I wanted to have a lot more granular control, so it seemed obvious to build it around the tools I was already using. I built a lot of tools. Next thing you know I had a dozen documents I was referencing for process, templates and communication scripts. On top of that I had a half dozen spreadsheets that I used daily, if not weekly to manage my business. I started connecting them with Google Apps Script and automating processes as much as I could. Cursor became the hot programming tool and I had the bright idea of combining a lot of my tools into one to make it easier to run my business and communicate with my clients. That one tool turned into three tools and OPSTŪL.AI was born. 

We are wrapping up development on OPSTŪL Core and have already opened signups for beta users. If interested, please go to the site and sign up. OPSTŪL Relay and OPSTŪL Signal are working their way through development and will be available soon after. It’s been a fantastic experience to build the tools I’ve always wanted, but could never find. I also realized I didn’t want one app to do everything. In fact, I only want each app to do a few things very well. In particular, I am not trying to reinvent the wheel here or replace any of the fantastic baseline tools we already use such as Gmail, Calendar, Spreadsheets and Docs. I want the OPSTŪL suite to work with those tools. This evolved direction led me back to how I was using those tools now, how my clients will be using those tools and what type of effortless integration we can create.

Introducing Spark

A few months ago, Google debuted Google Spark, a more enhanced Google Gemini agent. Similar to Gemini in a lot of ways, Spark has deeper functionality with your connected apps and can be empowered with custom apps, skills and schedules to work closely with you day to day. There are plenty of articles that explain what agents are and how to use them, so I’m not going to go into detail here about how that works, but essentially, in my opinion, what makes an AI Agent an Agent, is a Powerful LLM + Tools + Knowledge. I can’t emphasize enough how important all three of those components are. Six months ago Google didn’t have any products like this. Just six months ago AI Agents were talked about like you would talk about a person. You would have an AI Agent that is focused on a specific type of job and that’s what they did. From there we evolved to having multiple agents that could have multiple skillsets that would all work in parallel on the same project. Today, we are moving to a “single orchestrator” model with dynamic “subagent creation.” These “agent instances” are temporary, modular in context and load specific skills into memory only when needed, preserving token efficiency. That is a substantial evolutionary change in just a short period of time. This is what Spark is today.

Extensibility

First came prompts. Large detailed documents that provided background, tool capabilities and instructions. When skills became available, these tools and instructions could live within the agent like memory. Everyone then loaded up their agent with skills. Context windows immediately were clogged with skills and prompts using thousands of tokens, well before any solutions were provided. This led to slow responses, increased hallucinations and less accurate results. Optimization of the context window was needed. It’s important to note that there hasn’t necessarily been huge innovations to the core concept of the context window, just reworking the technology that was already there. What you put in as a prompt, is not what’s presented to the model, that’s the system prompt. Agent Skills are restructured to provide a reduced footprint while the model is now instructed to find which skill handles the request. Skills manage the tools to use and Just-In-Time context loading allows for information to be loaded from outside documents as needed. This piece is important because information obtained from these external documents is dynamic and if not structured correctly, they will be chunked and once chunked, can be reassembled inaccurately unless proper formatting is followed. What we now have today is a dynamic optimized workflow that expands and contracts on demand that is more precise and financially efficient.

Where do we go from here? In past traditional operations, automating a business meant writing custom software, embedding business rules into applications, creating tailored databases and implementing expensive integrations. The arrival of Agentic AI has created a dramatic shift in how we scale business operations. Natural language is the primary communication method and Standard Operating Procedures (SOPs) is the new executable code. 

How Process is the New Code

Historically, business documentation was passive. Standard Operating Procedures (SOPs) were written in PDFs, binders, or wikis to onboard employees, satisfy audits, or store tribal knowledge. In practice, they often suffered from becoming outdated quickly, forgotten, or ignored in favor of bad habits.

When operational workflows are coupled with AI agents, the role of the SOP fundamentally changes. It evolves from a passive reference guide into an active, real-time execution tool.

Natural Language as the Programming Language

The Past: A human read an SOP, interpreted the instructions, logged into three different software platforms, and manually performed the actions.

The Present: An AI agent reads the SOP document directly from a folder and translates those instructions into API tool calls (creating folders, updating spreadsheets, pre-drafting emails, and organizing task cascades) in seconds.

The Takeaway: The SOP is no longer a guide about the work; it is the executable program of the work itself.

Real-Time Operational Updates

The Past: Changing a company-wide process required re-training staff, updating software configurations, or commissioning custom code changes from developers.

The Present: To change how your business operates, such as making changes to client onboarding or updating invoice standards, you edit the text in the Online SOP. The next time an Agent Skill runs, the updated logic is executed immediately across your tools.

The Takeaway: Editing text in a document is equivalent to updating code in a production application.

Alignment for Humans and AI Agents

The Past: Automated software systems and human teams operated on completely different planes. Systems ran on code logic; humans ran on manuals and memory.

The Takeaway: A well-structured SOP governs both human contractors and AI assistants simultaneously. If a human contractor takes over onboarding, they read the SOP. If an AI agent executes onboarding, it reads the exact same file. This creates 100% operational alignment across your human and AI workforce.

Enforcement of Quality and Guardrails

The Past: Human oversight was required to ensure steps weren't skipped, naming conventions were respected, and communication standards were met.

The Present: The SOP embeds strict operational guardrails directly into the execution flow, enforcing rules like "Never delete a task," "Pre-draft client outreach emails in Drafts," or "Follow this folder naming structure."

Impact for Current and future AI-Forward Companies

For current and future AI-forward organizations, the transition of SOPs into a real-time execution tool represents a fundamental restructuring of company architecture, human capital, and operational economics.

The Leadership Shift: From Supervisor to System Architect

In traditional companies, founders and operational leaders spend significant time acting as manual supervisors, chasing task updates, enforcing quality control, reviewing file structures, and coordinating handoffs between departments. In an AI-forward organization, the operator’s primary job becomes System Architecture.

As the system architect, your job boils down to three main things. First, you’re designing the logic, writing clear, simple rules and standards in plain-English SOPs so the AI knows exactly what to do. Next, you’re the one connecting the dots by mapping out which Agent Skills talk to which folders or tools in your workspace. Finally, you’re managing the edge cases by double-checking things like drafted emails or tricky expense categories, then tweaking the instructions whenever something new pops up.

Advantages for AI-Forward Organizations

Lean Unit Economics (Non-Linear Scaling)

In the past, growing your business meant you had to keep hiring more people, more account managers, more coordinators, more admins, just to keep up with the extra work. But AI-forward companies work differently. They can grow their revenue without needing a large payroll. Now, a small team or even a solopreneur can handle a bunch of specialized projects all at once. This means the business can scale way up while keeping costs low and profit margins high.

Reduced Technical Debt

When a company spends money on custom software or tries to patch together a bunch of fragile SaaS tools, they end up with a mess of maintenance and debt. But when they go AI-forward, they skip all that. They just use their standard workspace tools and turn their SOPs into the actual code that runs everything. If they need to change how things work, they aren't paying a dev to push an update, the just open a Doc, tweak the text, and they're done.

Eliminate Tribal Knowledge and Outdated Documentation

In most companies, documentation is rarely used. People write things down, then immediately ignore them and start doing their own thing, taking that "tribal knowledge" with them when they leave. When you move to an AI-forward system, that lack of use just doesn't happen. Since the AI agents are actually reading and running your SOPs every day, those docs stay alive and accurate. Your SOP becomes the one real source of truth that keeps everyone, humans and AI, on the exact same page.

Built-In Operational Safety & Risk Mitigation

Quality control of the past usually depended on someone remembering everything or manually double-checking work, which is how things like missed steps, lost files, or messy client emails happened. AI-forward systems fix this by building rules right into the workflow. Simple things like saving client emails as drafts first or sticking to a specific folder structure make sure the work stays consistent and predictable without you having to worry about it.

Practical Playbook for Building an AI-Forward System

Getting your AI-forward system up and running really starts with writing down your processes in plain, clear steps, making sure to cover exactly what goes in, what happens next, where files should live, and what to do if things go wrong. Then, you'll break those SOPs into manageable chunks by linking them to specific Agent Skills, like giving them direct access to the right folders so they never have to guess. For the big stuff, like emailing clients, sending invoices, or paying bills, it’s smart to keep a human in the loop by having the agent just draft the work for you to approve. And whenever you run into a mistake or an edge case? Just think of it as a bug in your SOP. Tweak the doc, and your AI agent will pick up the new instructions right away.

Creating Operational Systems

Looking back at how working with Google Sheets and scripts, to building the OPSTŪL suite, and now orchestrating everything with Spark, it’s all been about achieving the same goal, finding a smoother, and faster way to get stuff done. Moving from hand-coding everything to designing AI-driven SOPs is such a monumental workflow change compared to how I’ve worked in the past.

By working with Spark, focused Agent Skills and clear SOPs, my written processes have effectively become the "brains" of my whole operation. For those of us running businesses, our role has totally shifted. We aren't just managers or tech maintainers anymore, we're System Architects. Every process you document or skill you connect helps build an automated, scalable business.

Truly, process is the new code. The future of business success will be determined by how well we document our businesses. What an incredible time.