Every week, another vendor pitches your business on another AI tool or system. Automate your marketing. Automate your accounting. Automate your customer service. The promises are seductive, especially when you’re running a $20M to $200M business and feeling the pressure to modernize. But here’s what nobody selling you AI wants to admit: most companies that adopt AI without a clear operational framework end up spending real money on marginal improvements to processes that were never the problem in the first place.

There’s a better way, and it starts with a concept that’s been around since the 1980s. It’s one that most middle market operators already understand intuitively, even if they’ve never put a name to it.

The Theory of Constraints in 60 Seconds

Eliyahu Goldratt introduced the Theory of Constraints (TOC) in his book “The Goal“, and the core idea is disarmingly simple: every system has a bottleneck. The output of the entire system is limited by that single constraint. If you want to improve the system, you have to find the bottleneck and fix it. Investing in anything other than the bottleneck is, at best, a waste of resources and, at worst, a driver of excess inventory, confusion, or misaligned capacity upstream and downstream.

For a manufacturer, the constraint might be a single machine on the production line. For a services firm, it might be the proposal-writing process that sits between qualified leads and signed contracts. For a distributor, it might be the warehouse pick-and-pack operation that limits how many orders can ship per day.

The discipline of TOC says: find the constraint, maximize its efficiency, align your other processes to support it, only then do you invest additional resources in increasing capacity. That sequence matters. And it matters enormously when you’re deciding where to deploy AI.

The Problem with “AI Everywhere”

Walk into any industry conference right now and you’ll hear a version of the same pitch: AI is transformational, it can be applied to virtually any business process, and the companies that move fastest will win. All three of those statements contain some truth, which is exactly what makes them dangerous.

Here’s what happens when a middle market company takes the “AI everywhere” approach. The CFO’s office automates invoice processing and is able to shave two days off an accounts payable cycle that was never causing a real business problem. Marketing deploys an AI content generator and now produces three times the blog posts for a website that wasn’t generating meaningful leads in the first place. The IT team implements an AI-powered helpdesk chatbot and resolves internal tickets 40% faster in a department that had adequate capacity all along.

Each of these projects can show an ROI on paper. Each one can be presented in a board meeting with impressive metrics. And none of them moves the needle on the thing that’s actually preventing the company from scaling.

Meanwhile, the real constraint — maybe it’s the quoting process that takes your sales engineers three weeks to turn around a custom proposal, or the quality-control bottleneck that limits your production throughput, or the onboarding process that means new hires don’t become productive for six months — sits there, untouched, getting worse as the rest of the organization speeds up around it.

What Happens When You Use AI to Speed Up a Non-Constraint

This is the part that the AI vendors won’t tell you, because it requires actually understanding how your business operates as a system rather than a collection of independent departments.

When you speed up a process that isn’t your constraint, you don’t get more output. You get more work-in-process sitting in front of the actual bottleneck. You get more pressure on the constrained resource. You get frustration from teams that are now producing faster but seeing no improvement in overall results. And you get a growing pile of AI subscription costs with no corresponding growth in revenue or margin.

Find your bottleneck. Aim there. Everything else can wait.

Think about it concretely. If your constraint is that you only have three project managers who can oversee complex implementations, and you use AI to generate twice as many qualified proposals, you haven’t scaled the business. You’ve created a backlog. Your close rates might actually drop because prospects get tired of waiting. Your PMs burn out. Your reputation suffers.

You spent time and money making things worse.

A Better Approach: AI on the Constraint

Now imagine the opposite. You identify that the constraint in your business is the quoting process – the step between a qualified opportunity and a signed contract. Your best estimators are maxed out. Every custom quote takes specialized knowledge, historical pricing data, and engineering judgment. You’re leaving deals on the table because you simply can’t produce quotes fast enough.

This is where AI becomes genuinely transformative. You build (or buy) a system that pulls from historical pricing, applies engineering rules, drafts an 80%-complete proposal, and puts it in front of your estimator for review and refinement. What used to take three weeks now takes three days. Your estimators aren’t replaced; they’re leveraged. The bottleneck opens up. More quotes go out. More deals close. Revenue grows.

That’s real ROI. Not theoretical. Not projected. Not “soft”.  Real, measurable throughput improvement that flows straight to the top and bottom line.

How to Apply This in Your Middle Market Business

If you’re a middle market business owner or operator considering AI investments, here’s a practical framework.

First, identify your constraint. This requires honest assessment, not wishful thinking. Where does work pile up? Where are your best people overwhelmed? What’s the one process that, if you could double its capacity overnight, would directly result in more revenue or higher margins? Talk to your front-line managers. Look at your lead times. Follow a customer order from intake to delivery and find where it stalls.

Second, understand the constraint deeply before automating it. Don’t just throw AI at the bottleneck. Map the process. Understand why it’s constrained.

> Is it a knowledge problem?  People don’t have the right information fast enough
> Is it a decision problem? Too many approvals, too much back-and-forth
> Is it a capacity problem? Not enough skilled people

AI is exceptionally good at certain types of problems (pattern recognition, data retrieval, draft generation, anomaly detection) and poorly suited to others (logical reasoning, common sense, novelty, self-correction, genuine understanding). You need to match the tool to the actual problem.

Third, measure throughput, not activity. The metric that matters is system output, not local efficiency. You don’t care how many invoices AP can process per hour. You care about how many customer orders ship per day, or how many deals close per quarter, or how many projects your firm can deliver per year. If your AI investment doesn’t improve that system-level metric, it’s not working, no matter how impressive the department-level dashboard looks.

Fourth, resist the temptation to automate everything else. Once your constraint is addressed, a new one will emerge. That’s the nature of systems. Follow the same discipline: find it, understand it, apply AI (or other solutions) specifically to it. This iterative approach means every dollar you spend on technology is directly connected to a measurable improvement in your ability to scale.

The Competitive Advantage of Implementing AI With Discipline

In a market where every company is rushing to adopt AI, the ones that win won’t be the ones that adopt it the fastest or most broadly. They’ll be the ones that adopt it the most strategically.

Middle market businesses have a structural advantage here. You’re close enough to the operation to see where the real constraints are. You don’t have seventeen layers of management between the CEO and the shop floor. You can move quickly once you know where to aim. But that advantage only matters if you resist the pressure to spray AI across your entire organization and instead apply it with surgical precision to the one thing that’s actually holding you back.

The Theory of Constraints isn’t new, but AI is. The marriage of the two — using the most powerful automation tools ever created, directed by the most time-tested operational framework we have — is the approach that will separate the middle market companies that scale from the ones that just spend.

Before investing in another AI platform, make sure you’re solving the right problem. Oak Hill works with middle market companies to identify operational constraints, improve scalability, and align strategic investments with measurable business outcomes. If you’re trying to determine where AI fits into your organization,  let’s talk.

 

_ _ _ _

Erik Owen is the President of Oak Hill Business Partners and has over 20 years of professional experience in Finance and Accounting, Administration, and General Management.  You can call Erik at 262.299.5526 or email him at erik.owen@OakHillBP.com.

Oak Hill Business Partners is a boutique business advisory firm serving middle market, closely held companies. Based in Milwaukee, WI, our partners focus on scaling companies using functional excellence in finance & administration, sales, marketing, and operations.  Oak Hill also helps company owners plan and execute transition/exit planning holistically. Oak Hill partners work with a team of advisors including wealth and legal advisors to help owners understand their options for transition in the business and execute the plan that meets their specific needs..