Part of our Practical AI series: tactical, tool-specific looks at what’s actually usable in AI right now. New to AI for business generally? Start with our beginner’s guide first.
Almost every company now uses AI somewhere. Almost none of them can point to real AI ROI from it. According to McKinsey’s State of AI research, nearly 9 in 10 organizations report regular AI use in at least one business function, yet only a small single-digit percentage can attribute a measurable profit impact to it.
That gap isn’t a technology problem. It’s a process problem, and it’s the single biggest factor separating the companies getting real AI ROI from the ones still waiting for one.
AI ROI Is a Redesign Problem, Not an Adoption Problem
Here’s the number that matters more than the adoption headline: McKinsey’s research found that fundamentally redesigning a workflow around AI has the strongest measurable link to profit impact of any factor studied, yet only about 1 in 5 companies that adopted AI had actually redesigned a workflow to take advantage of it.
The companies seeing real AI ROI generally aren’t using better models than everyone else. Frontier and mid-tier models are widely accessible at this point, with high performers and laggards often building on the exact same underlying technology. What separates them, and what ultimately determines their AI ROI, is what they built the AI around.
Bolting On vs. Rebuilding: What the Difference Looks Like
Bolting on takes an existing task and makes it a little faster:
- Someone drafts an email, and AI helps them draft it quicker.
- A rep looks up account history, and AI summarizes it faster.
- The task still exists. The workflow still exists. AI just rides along.
Rebuilding removes the task, or restructures the workflow so the task no longer needs a human step at all:
- Instead of a person summarizing account history before every call, the system delivers it automatically the moment a call is scheduled.
- Instead of a document routing through four approval emails, an agent checks the criteria and routes only the exceptions to a person.
Same underlying AI. Completely different AI ROI.
Where Legacy Systems Quietly Cap Your AI ROI
This is also where legacy system modernization keeps showing up in the same conversation as AI transformation. A lot of AI initiatives stall out not because the AI is weak, but because the workflow they’re bolted onto runs on a 20-year-old FoxPro, VB6, Delphi, or Access application that was never built to expose its data or logic to anything else.
You can’t meaningfully redesign a workflow around AI if the system underneath it can’t talk to modern tools in the first place. In our experience, the highest-value AI projects tend to follow, not precede, a legacy system that’s already been brought onto a modern platform.
Practical Starting Point for Better AI ROI
Rebuilding a process is harder than installing a tool, which is exactly why most companies haven’t done it yet. The good news: you don’t need to rebuild everything at once to see real AI ROI. Three moves tend to matter most.
Pick one process, not everything. Choose the workflow that eats the most time or loses you the most deals, and redesign that one first. Trying to transform every department at once is how AI roadmap initiatives stall out in the pilot stage.
Put an AI usage policy in writing. Most companies don’t have one, and employees are already using AI tools whether leadership has decided on a policy or not. A clear, written policy protects the business and gives the team confidence to use these tools well instead of quietly guessing at what’s allowed.
Train the people, not just the process. AI tools are only as good as the judgment of the person directing them. A few focused hours of training changes how a team actually uses what you’ve given them.
If you’re not sure which process to start with, diagnostic conversation is usually the highest-leverage next step.
Running on a Legacy System? Start There First
The three steps above work well once your core systems can actually support a redesigned workflow. For a lot of the small and mid-sized businesses we work with, that’s the real blocker to a strong return on AI investment: the AI strategy is sound, but it’s aimed at a system that can’t yet participate in it.
That’s exactly the intersection Ticomix works in every day:
- If your core system is the aging application everything else has to work around, our Ticomix REV methodology is built to modernize it faster and at a lower cost than a traditional rebuild.
- If the system itself is modern enough but the workflow around it isn’t, our AI automation team designs agents and integrations around your actual process, not a generic template.
- Not sure which applies to you? Run your numbers through our legacy modernization ROI calculator to see the cost of standing still versus the investment to move forward.
This fits into the bigger shift we’re seeing across the board. See our full look at how AI integration in business works for the wider picture.
Frequently Asked Questions About AI ROI
Most AI spending goes toward making an existing task faster rather than removing the task or redesigning the workflow around it. Per McKinsey's research, companies that fundamentally redesign a workflow to use AI see a meaningfully stronger link to profit impact than companies that simply add AI tools to their current process.
It means changing the process itself, not just adding an AI-assisted step to the process you already have. Instead of using AI to summarize a report faster, for example, you redesign the process so the report is generated and routed automatically, and a person only steps in for the exceptions.
Start with whichever process costs you the most time or the most lost deals today. That's usually the one where a redesign pays for itself fastest. If it isn't obvious, a short diagnostic conversation with a team that's done this before is faster than guessing.
Most likely yes. Employees are generally already using AI tools with or without a formal policy. A written AI usage policy protects the business and gives your team clear guardrails, regardless of how far along your official AI roadmap is.
Often, yes. If your core business system is an aging application that can't easily share data with other tools, there's a ceiling on how much any AI initiative built around it can accomplish. Modernizing that system first tends to unlock the AI automation work that comes after it.
Want more practical AI insights like this? Check out our other Practical AI posts: Claude for Small Business: What It Actually Does and how to build an AI chatbot in Microsoft Teams. Mark Perez, Ticomix’s Chief AI Strategist, also hosts a monthly AI Office Hour for clients and prospects — a working session on what’s actually useful in AI right now, not the hype. Email Mark for details on the next session.