The Question Businesses Are Afraid to Ask About AI

Everyone's Chasing AI. Most of It Is Theater.

You've seen the meme. The guy walking with his girlfriend, head turned to stare at someone else — captioned something like "CEOs" on the left, "their actual business problems" as the girlfriend, and "AI" as the one they can't stop looking at.

It's funny because it's accurate.

Since 2023, the pressure to "do something with AI" has been relentless. Vendors slap "AI-powered" on everything. Board members ask about it in every meeting. LinkedIn is a wall of AI hot takes written by AI. And somewhere in the middle of all of it, the real question gets lost:

Is AI actually solving your problem — or are you just chasing it because everyone else is?

This post is for the businesses that want to think clearly about AI before they spend money on it. We're going to be honest about what we've seen work, what we've seen fail, and what we're doing ourselves.

The Most Important Sentence About AI You'll Read This Year is:

You can automate stupid at scale.

If your quoting process is inconsistent, AI will generate inconsistent quotes faster. If your customer onboarding is confusing, an AI chatbot will confuse customers more efficiently. If your data is a mess, AI will surface messy insights at a speed your team can't keep up with.

AI is a force multiplier. Like any multiplier, it amplifies whatever it's applied to — good processes and bad ones equally. The businesses that are winning with AI right now didn't start with AI. They started with clear, documented, functional processes — and then applied AI to make those processes faster or more scalable.

The ones struggling are the ones that treated AI as the solution before they identified the problem.

What the Theory of Constraints Tells Us About AI Adoption

Eliyahu Goldratt's The Goal introduced a framework that's as relevant today as it was in 1984: every system has a single binding constraint. Improving anything that isn't that constraint doesn't improve throughput — it just creates a different kind of waste.

We've written about how this applies in the age of Industry 5.0 in depth. The short version for AI: the question is never "how can we use AI?" The question is "what is our actual constraint, and does AI address it?"

If your constraint is that your best estimator is backlogged three weeks, AI-assisted estimating tools might be transformative. If your constraint is that customers don't trust your quotes because they're inconsistent, AI will generate bad quotes faster — and your estimator's backlog wasn't the real problem.

If your constraint is that your team can't find information when they need it, an internal AI assistant with access to your documentation can be a genuine force multiplier. If your constraint is that your documentation doesn't exist or isn't accurate, AI will make it easier for people to retrieve wrong information faster.

Identify the constraint first. Then ask if AI addresses it.

Where AI Actually Works (From Experience)

We're not writing this as observers. TotalCare IT is in the middle of migrating our own website to an AI-native platform specifically because we want our content to be queryable, composable, and available to AI agents in ways that a traditional CMS doesn't support.

We're using AI agents in our own workflows. Not chatbots. Agents — AI that can take multi-step actions, reason over context, and hand off to the right person or system at the right time.

Here's what we've seen work, in our own operation and with clients:

Drafting and editing, not replacing. AI is excellent at producing a first draft that a human then edits, fact-checks, and makes sound like a person. The value is in collapsing the time from "blank page" to "something worth editing" — not in replacing the human judgment about what should be said.

AI agents for repetitive multi-step workflows. The real productivity gains aren't from AI generating content. They're from AI handling the routing, summarizing, triggering, and coordinating that used to require a human to track. An agent that monitors incoming requests, categorizes them, pulls relevant context, and routes to the right person — without needing someone to babysit it — that's where the time savings are meaningful.

Predictive maintenance where the data foundation exists. For manufacturers with instrumented equipment and clean sensor data, AI-powered maintenance forecasting is legitimate and the ROI is clear. The catch: "where the data foundation exists" is doing a lot of work in that sentence. Most operations need to build that foundation before AI can use it.

Documentation and knowledge management. The average manufacturing company has process knowledge locked in the heads of their most experienced employees. AI-assisted tools that can capture, organize, and surface that knowledge make it available to new employees faster and reduce the single-point-of-failure risk when someone leaves.

Where AI Is Still Theater

Having AI write your company policies.
Quick, comprehensive, legally-worded — what's not to like? The problem is that policies written by AI reflect generic best practices, not your actual workflows, your team's reality, or your company's values. When employees violate a policy that was never really designed around how work actually happens here, you have a document problem dressed up as an HR problem. Policies need human judgment baked in from the start, not just a signature at the bottom.

Using AI to send connection requests and outreach on LinkedIn at scale.
The logic sounds reasonable: more outreach, more pipeline. The reality is that everyone can feel it. The "personalized" note that references your job title and a recent post you made — nobody's fooled. You're not building relationships at scale. You're burning your reputation at scale with people who didn't know you before and now actively don't want to. Trust is built slowly and destroyed quickly. AI-powered spam just speeds up the destruction.

Creating content at scale — and losing your brand voice in the process.
Volume without editorial judgment doesn't build authority, it erodes it. If your customers have been reading your content for years because it sounds like you — because it has a point of view, because it's honest and a little direct — and suddenly it reads like everyone else's, they'll notice before you do. We've been working through our own blog audit for exactly this reason. Several posts we're consolidating or deleting were AI-generated filler that had no TotalCare voice in them. More content is not the same as better content.

Using AI to rate or rank your employees.
Productivity scores, performance metrics, output tracking — AI can measure a lot of things. What it can't measure is the person who quietly unsticks every project before it becomes a crisis, the one who keeps morale up during a hard stretch, or the one who's been carrying institutional knowledge for twelve years that would take years to replace. Optimizing for what AI can count will cost you what it can't.

Putting all your business data into AI and expecting it to find the answers.
Connect your ERP, your spreadsheets, your CRM — let AI surface what you're missing. Except: AI finds patterns in whatever data you give it. If your records are inconsistent, if three people track the same metric three different ways, if there are gaps nobody documented, AI will produce confident-looking insights built on a shaky foundation.

"Garbage in, garbage out — just faster, and with better charts."

The Three Questions to Ask Before Any AI Investment

Before your business pursues any AI initiative, ask these three questions:

1. What is our actual constraint right now?
Not what AI could theoretically improve. What is the specific thing limiting your throughput, revenue, quality, or capacity? If you can't name it clearly, you're not ready to deploy AI against it.

2. Does this AI tool address that constraint directly?
If it addresses something else — something that isn't the bottleneck — you'll generate activity without improving outcomes. That's a waste of money and attention.

3. Do we have the process clarity and data quality that AI requires?
AI amplifies what's there. If the underlying process isn't clean, the data isn't organized, or the expected behavior isn't documented, AI has nothing to work with. Fix those first.

Treat AI like a new employee, not a magic button.

New employees need clear instructions, well-documented processes, and feedback to improve. AI tools do too. The businesses that get the most from AI are the ones that treat deployment as an ongoing investment, not a one-time setup.

AI is genuinely useful. It's also genuinely overhyped, and a lot of what's being sold as AI strategy is theater designed to capture budget before the dust settles.

The businesses that will look back on this era and say "we made the right calls" are the ones that asked hard questions before moving fast — that identified their real constraints before selecting their tools.

At TotalCare IT, the conversations we have with manufacturers, engineering firms, and operationally-critical businesses in Boise and Idaho Falls start with your actual situation — not with a product pitch. If you want to think through where AI fits in your operation, we're happy to have that conversation.