Before AI Can Help, It Needs to Understand Your Workflow

“61% of organizations manage half or more of their content outside official systems, leaving critical processes undocumented and informal.”

(AIIM white paper PDF)

Ask five people how they complete a routine task, and you’ll hear five different stories. These aren’t tales of inefficiency; they’re snapshots of reality. People find their way. They rely on memory, sticky notes, screenshots, and instincts honed over time. The real magic of getting work done often lives between the lines of your company’s SOP.

This is where AI meets its first obstacle. No matter how sophisticated the AI tool, it can’t improve what it doesn’t understand. If your workflows are scattered, improvised, or undocumented, your AI investment may only end up perpetuating confusion.

Even tried-and-true methodologies like Lean or Six Sigma depend on having reliable, observable processes to analyze. Without visibility into how people actually work, neither humans nor AI can make meaningful improvements.

Why SOPs Aren’t Enough Anymore

Standard Operating Procedures (SOPs) outline how things are supposed to work. But in most organizations, what’s documented is only half the story. What’s not written down, the side conversations, the tab-switching rituals, the workaround scripts, is where most of the real workflow lives.

If you try to layer AI on top of an SOP without understanding how people actually do the work, you’ll get unpredictable results. Optimizations will be misaligned, reports will be formatted incorrectly, and automation bots will click the wrong buttons. Tools like NotebookLM can’t surface what hasn’t been captured.

Think of SOPs as stage directions. What you need is the actual performance.

What Is an Individual Workflow?

An individual workflow is the path someone takes from “task received” to “task complete.” It includes every decision, every shortcut, and every nuance, even the ones they don’t realize they’re making.

For example, submitting a travel expense involves more than uploading receipts. It might include digging through emails, renaming PDFs to a specific format, cross-checking with a personal spreadsheet, and waiting for an unofficial nod from finance. That’s the workflow. Unless AI can see the whole process, it can’t offer meaningful help.

This is the work layer where AI can become a brilliant assistant or an annoying interloper.

Why AI Needs Workflow Context to Be Useful

AI needs specificity. It needs a map. Drop it into a process with missing pieces, and it will start guessing. And guesswork is expensive in business.

We’ve all seen automation fail because it lacked context. Maybe a chatbot gave a customer the wrong answer, or a document generator filled in the wrong template. These aren’t technology failures; they’re clarity failures.

To make AI work for your business, you must teach it the rules and rituals. And that starts with documenting what your people are actually doing. NotebookLM can give AI that context, but only if the information is clear and grounded in real practice because it’s documented.

How to Capture Real-World Workflows

Here’s a five-step approach to documenting workflows in a way that actually prepares your business for AI:

• Pick a high-impact task. Start with something repeated often, like onboarding, responding to inquiries, or submitting reports.
• Shadow the people who do the task best. Observe their screens, watch their clicks, ask questions, and capture even the “this is just something I do” moments.
• Map the steps in sequence. Use simple language. Include software tools, bookmarks, email templates, folders, and physical aids.
• Ask: What’s manual? What’s repetitive? What requires judgment? This helps you distinguish between what AI can handle and should remain human-led.
• Compare this to your current SOP. You’ll likely find gaps, areas where your documentation needs to catch up with reality.

This is also a good time to pull from established frameworks like Lean or Six Sigma. These methodologies emphasize reducing variation, identifying bottlenecks, and eliminating non-value-adding steps, all critical when looking for opportunities to improve or optimize how work gets done. With AI analysis tools like Google NotebookLM, these frameworks can be tailored to fit your company’s actual workflows and SOPs. By grounding your documentation in real behavior and layering it with structured analysis, you create a powerful system for continuous improvement that delivers real-world results.

Once your workflow is mapped, consider loading it into NotebookLM to make it easier for AI to understand, search, and reuse.

Example: One employee uses a macro to reformat spreadsheets before sending them to leadership. This isn’t in the SOP, but it saves hours weekly. Once it’s documented, that task is ripe for AI assistance.

Why I Use Google NotebookLM

Once you’ve mapped your real-world workflows, the next step is making that information accessible to your AI tools. That’s where Google NotebookLM comes in.

NotebookLM is an AI-powered research and documentation tool that lets you organize your knowledge into source-based notebooks. You can upload SOPs, meeting notes, workflow observations, PDFs, and even entire knowledge bases and query that content using natural language.

Unlike traditional AI chat tools, NotebookLM is grounded in your sources. That means every answer it gives is traceable, contextual, and specific to your content. It’s not just guessing; it’s pulling from the material you’ve taught it.

NotebookLM gives AI the context to act like a smart assistant, not a clueless intern.

This makes it an ideal companion for workflow discovery. As you document how work is really done, you can feed that knowledge directly into NotebookLM and use it to:

• Generate accurate SOPs and quick-reference guides
• Ask follow-up questions across projects or tasks
• Collaborate with team members on shared knowledge
• Train other AI tools (like ChatGPT or Copilot) with better prompts

Integrating NotebookLM into your process makes your workflows visible, searchable, explainable, and scalable. Now, add documents to NotebookLM’s “Sources” containing well-documented methodologies like Six Sigma and tell the AI to look for ways to optimize your SOPs and/or workflows. You now have an extremely powerful consultant at your disposal. Optimization becomes achievable and sustainable.

Modern SOPs: Blending Structure and Flexibility

Once you’ve captured individual workflows, it’s time to upgrade your SOPs. This is not to rigidly lock everyone in, but to provide clear guidance on where AI belongs, and human flexibility remains essential.

A next-generation SOP might include:

• Clear, documented steps for both humans and AI
• Built-in flexibility for edge cases and exceptions
• Specific AI commands like “Use Copilot to summarize notes before submission.”
• Checkpoints for human validation

These dynamic SOPs become more than documents—they become playbooks for collaboration between people and their digital assistants, especially when paired with tools like NotebookLM, which turn those documents into interactive knowledge bases.

Real Results: What You Unlock by Doing This

“74% of companies prioritize AI, but few align it to real workflows.”
(PwC Global AI Study white paper PDF)

The companies that invest time in workflow discovery and integration see real benefits:

• More useful AI tools: Optimizations that actually save time and reduce stress
• Improved training: SOPs reflect reality, so new hires get up to speed faster
• Fewer errors: Tasks happen the same way each time, with AI or humans following reliable steps
• Innovation: Employees can offload grunt work and focus on improving systems

Start With a Conversation

AI can’t guess how your business runs. It needs you to show it. And that starts by talking to your team.

“How do you really do this task?”

That one question can uncover the hidden knowledge you need to unlock optimization. It’s not about documenting every click; it’s about surfacing the logic behind the work and using that insight to make AI smarter, faster, and more helpful.

AI isn’t here to replace your people. It is here to learn from them; the sooner it does, the sooner it can contribute.

Check out my next post HERE for a tool I use personally to interview SMEs and document individual workflows.