Good AI results needs good workflow design…

AI Can’t Save You From a Broken Workflow

Why Understanding "How You Work" Matters More Than You Think

Recently, I was chatting with two friends about books when we stumbled onto an interesting question:

At first, our answers seemed simple.

I am a Kindle lover (I get to carry 1000s of books in my pocket), so I go straight to Amazon, check reviews, browse recommendations and even use AI-generated summaries (this is awesome!).

Tom picks books based on the cover—if the design and title grab him, he's in.

And Jane? She relies on recommendations from trusted book clubs and friends.

Same goal—choosing a book—but completely different decision-making processes.

But as we kept talking, I realised something important: Jane’s process wasn’t as simple as she thought.

The Hidden Complexity of Simple Decisions

Jane described her approach as “Oh, I just go with recommendations from trusted sources.”

But when I asked her to walk me through it, her actual process turned out to be much more complicated and intricate:

She starts with recommendations, but then she also...

• Checks Goodreads ratings,

• Reads book critic reviews, and

• Scans blog posts before making a decision.

There’s a gap between what Jane thinks she does and what she actually does. We all have some sort of gap.

This gap doesn’t just exist when we choose books.

Imagine I had built an AI bot based on Jane's initial description. It would have simply purchased any book recommended by her trusted sources. But that's not what Jane actually wants or how she makes decisions.

This same gap exists in professional workflows—especially when businesses try to implement AI.

Most people don’t recognise the full complexity of their own decision-making, and that’s a problem.

This small example mirrors a much larger challenge in AI adoption—when we don’t fully understand our own decision-making, we can’t effectively teach AI to assist us.

Why Understanding Your Workflow Matters

For AI to be truly effective, it must reflect how you think, not just execute broad instructions.

People often oversimplify their workflows. They say things like:

• "I open the document."

• "I do a web search."

• "I write the report."

But they don’t mention the micro-decisions:

• Where do they look first when opening the document?

• What information do they prioritise or ignore?

• How do they verify what’s accurate?

This lack of awareness creates a major issue when designing AI solutions. If we don’t capture the real decision-making process, AI will never truly mirror human thinking.

Imagine venture capitalists (VCs) screening startups. They often say they evaluate "team, market, revenue." But their real process includes:

• How they verify founders’ backgrounds?

• Which market signals matter most?

• How they weigh numbers against team dynamics

• What triggers deeper investigation?

If an AI system was built only on their surface-level description, it would completely miss these critical steps.

The “Magic Button” Myth

Many people assume AI is a one-click solution:

• Give it data, and a perfect report comes out.

• Ask it to analyse trends, and the insights are flawless.

But here’s the reality: AI doesn’t “just know” what you want. It follows patterns and instructions—just like humans do.

Read our article on The AI “Magic Button” Myth.

This brings us to a critical issue in AI adoption: People rarely articulate their full decision-making process, yet expect AI to 'just work.' But AI doesn’t 'just know'—it mirrors the patterns it's taught. If those patterns are incomplete, so are the results.

So how do we bridge this gap?

Bridging the Gap: The “Think Aloud” Method

Getting to the real workflow, and understanding how my client works is often the biggest part of the work.

And my biggest pain points when implementing AI is:

• Undefined workflows

• Unstructured documents and bad data. Check out: AI’s Kryptonite: Bad Data

• Badly designed/non-existent processes

• Lack of templates

• The workflow

You would think that building the AI is the hard part, but it is actually the workflow design.

To uncover these hidden workflows and micro-decisions, I use a method called the Think Aloud Study.

A "Think Aloud Study" is asking someone to narrate their inner monologue while they work. They explain what they're doing and thinking as they do it.

If we want AI to work for us, we first need to understand how we actually work.

When working with clients, I ask them to narrate their actions in real time while performing a task. This reveals:

• Micro-decisions they don’t realise they make

• Hidden rules and patterns in their work

• Unspoken expertise that AI needs to capture

Once we map their real workflow—not just their perceived one—we can design AI solutions that work the way they actually do.

Building Modular AI Workflows

Many assume AI should run entirely on its own, but AI lacks true understanding—it processes patterns but doesn’t reason or verify its conclusions.

Let’s explore…

The “Black Box” Workflow

When people aren’t clear on their process, and can’t communicate it to AI, I call it a “black box” workflow.

A "black box workflow" is a process where the inner workings are hidden or unclear.

 You know what goes in and what comes out, but the steps in between are obscured. 

An unsophisticated AI workflow looks like this:

1. AI receives raw data.

2. AI processes everything at once, with no clear guidance or structure.

3. AI generates a final report.

Problem?
If AI makes a mistake early in the process, that error propagates through the entire workflow—and users only realise it at the end.

AND… you can end up with generic and flat, obviously AI generated output… bleh!!

The Modular Workflow

A modular AI workflow breaks the process into structured steps, allowing AI to follow a clear process, and give output at every stage.

It also allows human oversight along the way:

1. AI gathers data → User checks for missing or unreliable sources.

2. AI analyses data → User reviews and identifies what’s important.

3. AI refocuses and refines analysis → User questions AI’s assumptions and digs deeper.

4. AI creates a draft report → User verifies accuracy before finalising.

Every time a human gives good feedback, the quality of the AI’s output improves… but only if you are an expert and understand AI.

This is a great example of Augmented Intelligence: AI makes you smarter.

Human Feedback Trick

This also allows the expert to consistently guide, correct and refocus the AI.

People underestimate the impact of their expert opinion and guidance throughout the AI process.

Breaking up complex tasks is good practice. It forces you to think about how you do work. And you get better quality outputs.

Instead of one overcomplicated AI, modular workflows keep things simple, structured, and correctable.

Key Human Checkpoints in AI Workflows

Initially everyone thought that AI would completely replace e.g. radiologists. Not true. AI can assist radiologists with image recognition tasks, but it can't yet handle the entire diagnostic process.

Get used to building in checkpoints in your AI engagements:

1. Directing AI: Asking the Right Questions

• AI is only as good as the questions you ask. A vague request leads to generic or misleading results.

• Why it matters: AI doesn’t set priorities—you do. Defining scope, relevance, and focus improves accuracy.

2. Evaluating AI Output: Never Accept at Face Value

• AI’s responses can be wrong, biased, or lack depth. Blindly accepting results increases the risk of misinformation.

• Why it matters: AI doesn’t question its own output—humans must.

3. Challenging Assumptions: Spotting What AI Misses

• AI finds patterns, not meaning. It reports correlations but doesn’t ask 'why' or 'what if.'

• Why it matters: AI makes surface-level connections—you add critical thinking.

4. Refining and Iterating: Improving AI Over Time

• AI isn’t a one-and-done tool. The best results come from continuous feedback and adjustment.

• Why it matters: AI learns from structured feedback—but only if you provide it.

The Future: Working With AI, Not Being Replaced By It

Right now, AI workflows need human orchestration—breaking tasks into steps, reviewing outputs, and refining the results.

Over time, AI will:

• Get better at predicting what you need.

• Learn to anticipate missing information.

• Adapt to your unique workflow over time.

But even in the future, AI won’t replace the human role of strategic oversight, judgment, and experience.

Think about Excel—it didn’t replace financial analysts; it made them more powerful. AI is similar. It’s not about replacing human judgment; it’s about amplifying it.

The future isn’t one mega-bot that does everything. It’s a system of specialised AI tools working together under human orchestration—each focused on a specific task, but guided by human expertise.

Your value isn’t in the mechanical steps AI can automate—it’s in setting direction, understanding what matters, spotting patterns, and making decisions. 

The better you understand your own process, the better you can leverage AI to enhance it.

Ready to AI-ify Your Workflows? Let’s map it out in a free consultation.

If you're looking to streamline your workflows, reduce manual effort, and make AI work for you, let’s set up a time to talk.

• Identify where AI can save you time, money, and mental effort

• Break down complex tasks into modular AI-driven workflows

• Ensure AI works with human oversight for better accuracy and efficiency

AI isn’t just a tool—it’s a way to work smarter. Let’s explore how it can transform your processes.

👉 Schedule a free consultation today and take the first step toward an AI-powered workflow designed for real results.

If you forget everything else, remember this…