Here's the Missing Link Between AI Adoption and ROI

By John Marshall

Most businesses do not have an AI problem; They have a work-design problem. After buying licenses and experimenting with various AI programs, someone has written a better email, summarized a meeting, analyzed a spreadsheet or built a report faster. That’s useful, but it’s not transformative. To get real business results, you need to redesign the workflow.

By workflow, I mean the path a piece of work follows from start to finish—who touches it, what information is needed, what decisions are made and what result comes out the other end. Take quoting in an LBM business: Someone has to understand the job request, fill in missing information, select products, check availability, apply pricing rules, consider freight, handle exceptions, prepare the quote, and follow up.

AI may help with several of those steps, but if all you do is make one step faster, you have not necessarily improved the result. This is how businesses end up with employees saving ten minutes while the customer still waits two days. 

The question is not “Where can we use AI?” The better question is “What result needs to improve?” Then work backward through the process required to produce it.

Do Not Automate a Bad Process

There is a temptation to look at an existing process and ask which steps AI can automate. Sometimes that is exactly the wrong place to start. Automating bad work just lets you do bad work faster. A process may contain steps that exist only because information was hard to find, analysis took too long or someone had to move data from one system to another. If AI removes those limitations, the answer may not be to automate the old process. It may be to eliminate part of it.

For example, consider a weekly management report that takes four hours to assemble from three systems. AI may reduce that chore to 30 minutes—Useful, but perhaps management no longer needs the same report every Friday. The better answer may be to make the information available when a manager needs it and allow follow-up questions immediately. Now you have not just made the old work faster. You have changed the work.

That is where I think the next phase of Generative AI will create its biggest value. For the first few years, most attention was on the individual user: prompting, choosing tools, and making one person more productive. That still matters, but individual productivity is not the finish line. The larger opportunity is moving from individual productivity to company performance by redesigning the workflows. That is where AI begins to affect capacity, speed, service, quality and margin, and where its value and ROI become easier to see.

But beware: Redesigning the workflow does not mean handing unrestricted authority to AI.

The goal is not maximum automation. It is the right division of labor between the person and the machine. The operating guideline of “Trust but verify” is not a sign of resistance to AI. It is how responsible companies use any powerful tool.

Start With One Workflow

Most small and midsize businesses do not need an AI transformation program. In fact, I would be suspicious of anyone trying to sell them one. They need one important piece of work that is worth making better.

Choose something that happens frequently, creates real friction and matters to the business: quoting, collections, purchasing, customer follow-up, inventory exceptions, warranty claims or month-end reporting. Then follow the work from beginning to end, taking the time to investigate what actually happens, not what the procedure manual says is supposed to happen. 

Find the waiting, searching, re-entry, unnecessary handoffs, and decisions that keep climbing the organization because nobody knows who owns them.Then decide where AI belongs and where it does not. Give the AI program the company context it needs, establish where human approval is required, and put someone in charge of the workflow, not just the technology. 

Measure the result. Did the quote get back faster? Did collections improve? Did the error rate fall? Did the salesperson spend more time selling? If not, keep working on the workflow or stop.

The point is not to prove that AI works—We already know it can do useful work. The harder question is whether your company can redesign the work well enough to capture the value. The next AI advantage will not belong to the company with the best chatbot. It will belong to the company with the best-designed work.

John Marshall is owner of AI Growth Partners. Previously, he was CEO at Wilson Lumber and President at AllTemp Windows. Along with his work at AI Growth Partners, he is interim CEO and Board Chairman at Tara Manufacturing.