Replaced by AI? Workflows, Service Design, and What AI Doesn’t Know

TL;DR? We don’t share everything with AI— and probably never will. In that gap is the boundary you get to set between your service’s value proposition and what folks can reasonably do with AI (instead of you).

Beyond just “How can I take advantage of AI for myself?”, I keep seeing the same two questions pop up:

• “How has AI impacted your business?”
• “Why not use AI for this instead of me?”

In some circles, it’s followed by a third along the lines of:

• “If they’re going to use AI, should I make AI a part of my service offering?”

Depending on the moment … These questions could be reflecting any mix of crisis (“Am I about to be made irrelevant?”) and pragmatism (“The landscape has changed. What should I do about it?”).

And here’s the truth that validates both:

With good instincts, a solid knowledge base, a framework for making judgement calls, and a little experience, AI can probably get you about 70% of the way to where you want to go.

(Yes, that’s a made-up number, but …. It’s there to help make the point that that’s probably sufficient for most things. Perfectionism and decision paralysis aside, most of us will — justifiably — stop right there.)

When it’s not sufficient … It’s not a question of a better prompt or an updated model. It’s about what’s missing from the data used to train the models — and finding the boundary for when that matters and when it doesn’t.

How I Use AI for My Own Business

Tying back to that 70% I mentioned earlier, I’ve found that AI can help me with a lot of real work in my business.

Draft Feedback That Maintains POV

As a personal MO, I almost never use AI to generate ideas (although a conversation might spark one); I find it easy to get derailed by AI’s confident pushiness while I’m still germinating. So I typically pull together a working draft — or, at the very least, a robust outline –and then run it by the AI. That helps to preserve my POV and my own judgement calls, which I think are essential to connecting with my audience and staying true to my vision and brand.

Quick-and-Dirty Research

These will be prompts along the lines of: “Research reddit and similar forums and tell me what kinds of questions folks are asking about dealing with whitespace in their Excel data” or “Act like a funnel strategist. What kinds of lead magnets or audience builders or other marketing mechanisms would you recommend for [service URL]? Make people want the paid version. Use data from places like Google Trends, X, Reddit, and anywhere else you find useful.”

This can help me understand how my efforts fit into the larger landscape. Sometimes this helps me validate or refine something like a video script before I start scripting. Sometimes this helps me find new ways to market a service. And sometimes it helps me assess whether what I’m thinking about is more of a blue ocean or a dry well based on factors like how others are productizing similar services or finding workarounds that mean that no one is really paying for this.

Different Perspectives

Different models are trained in different ways. Additionally, once you’ve been in a single chat long enough, you and the AI have created an echo chamber. That’s why I don’t rely on just one AI when what I’m working on really matters. I’ll take the idea that I’m in love with, that seems to hold together really well, and pop it into another AI to see what I get back on a cold read. (Data policies matter here — so this isn’t always an option.)

And, when it really, really matters, I’ll still run it by a human at some point.

Where AI Hits a Wall

In each of the cases listed above, AI is latching on to what’s popular in a “what’s normal” or “majority rules” kind of way. It’s picking up on patterns. Outliers are easily overlooked. These issues are model issues, and they might get solved at some point. (In the meantime, it means AI might have a harder time helping you decide if your next business idea is a dry well or a blue ocean.)

But what’s really missing is … all the stuff we don’t type in. All the stuff we haven’t made public.

Think about that post that turns a failure into a lesson learned. In packaging it into a story, that experience has been glossed over. Sanitized.

What’s missing from the data, then, are things like:

• interpersonal struggles and negotiations (the retired spouse or the mismatched work schedules)
• energy levels and “baby brain” — and how they affect performance
• the feeling of the ugly cry
• intuition that you don’t articulate — but that’s built from lived experience, and helps you make the split second decisions that pan out
• the closed door business meetings and negotiations — the things we agree will never be shared outside this room

And this is where a human really has an advantage.

Things that stress test well for AI, or that sound confidently great on paper … don’t always hold up for humans who are dealing with exactly this kind of stuff on their own behind-the-scenes. We’ve failed and ugly cried and struggled and been through it in a way we’ll never put online for the public, even when we gloss up our failure stories for public consumption.

This is as true in technical, data, and SOP spaces as it is in coaching spaces.

Reimagining the Service Boundary

For whatever your service category, the question is: Where can you be reasonably and justifiably replaced by AI, and when does it matter that the “ugly cry” data is missing from the AI’s data?

In my own business, the thing most under “threat” is probably my 1-to-1 Advisory service. Which has meant getting real about where the boundary needs to realistically be for this — so I can keep delivering value. Here’s where I’ve landed for now:

• AI can be a “companion” to our work together. (1) I can help my clients stress-test AI plans that sound logical, but might be overlooking some very human elements. (2) I can use AI to jumpstart a work session. This can look like using AI with my own judgement, experience, and expertise to mock up a positive proposal, so that my client has something “concrete” to bounce off (e.g., marketing plans, 1-pagers, etc.).
• Beyond that, there’s all the stuff we’re not sharing with AI models (intentionally and not) — but that we do tend to share during conversations with humans we trust. In a confidential space, with someone who is invested in your business, a lot of that will be drawn out and discussed with candor and realism. Like what it’s like to negotiate contract specs with a team of lawyers from a large firm … when you’re just one person.

Points to Ponder

Where is your business situated with respect to how AI is impacting the business landscape?

In particular:

• How are you using AI in your business? Where does it fit in? What’s working well? What tools are you using?
• Have you redefined any of your services in response to how your clients use AI? What’s not included in the AI training data that you bring to the table? What have you let go of that used to be part of your service’s value?

I’d love to hear where you’re sitting with this.

— –

This content was featured in the March 2026 Blou Designs newsletter. If you’d like more content like this delivered to your inbox, so you can better align your efforts (one manageable piece at a time), subscribe.