Everyone's chasing AI transformation. The companies actually seeing returns are targeting something much smaller — the dead time between steps. Here's what that looks like in practice.
Every mid-market operator I talk to right now is somewhere on the same spectrum. Either they've tried an AI pilot that quietly died, or they're about to launch one and quietly nervous it will.
The question they're really asking: *Where does AI actually pay off?*
The honest answer isn't glamorous. It's not your large language model trained on proprietary data. It's not a chatbot on your website. It's the dead time between steps — the handoff lag, the approval queue, the admin that piles up after every meeting or customer interaction.
That's where real ROI lives in 2026. And most companies are still missing it.
There's a pattern in the AI workflows that consistently deliver measurable returns. They share one trait: they *remove* work rather than *assist* with it.
Assisting with work looks like this: a rep gets an AI writing tool to help draft follow-up emails. They still have to open the tool, write a prompt, review the draft, edit it, and send it. You've added a step while trying to remove one.
Removing work looks like this: after a sales call ends, the CRM is automatically updated, a follow-up email is queued for review, and the next action is surfaced in the rep's task list — without the rep doing anything.
Same outcome. Completely different impact on their day.
The workflows that consistently deliver in regulated industries follow this same logic. Think about what actually burns time in your operation:
None of these are headline-grabbing. All of them compound.
Here's the problem specific to the companies we work with: compliance overhead makes it easy to default to caution in a way that kills useful AI before it starts.
The legal team flags liability. The compliance officer wants a full audit trail. IT wants a security review. Change control needs documentation. Six months later, you've approved a pilot for a low-impact use case that nobody's excited about, and the business unit that asked for AI has moved on.
Meanwhile, the workflows that would actually pay off — the ones touching real volume, real repetition, real handoff friction — never made it to the table because they sounded too risky.
This is backwards. The highest-value AI workflows in regulated industries are often *lower* regulatory risk than people assume, precisely because they're operating in the space *around* core processes, not inside them. You're automating the admin layer, not the clinical decision or the underwriting judgment or the legal opinion.
That's a meaningful distinction that most organizations haven't made clearly yet.
If you're trying to identify where AI can actually move the needle for your organization, start with this filter:
Step 1: Map the handoffs. Where does work sit waiting — waiting for input, waiting for approval, waiting for someone to move it to the next stage? These are your candidates.
Step 2: Count the frequency. A handoff that happens 500 times a month is worth solving. One that happens 10 times isn't. Volume is what makes automation meaningful.
Step 3: Assess the decision type. Is the action at that handoff predictable and rule-based, or does it require judgment? AI handles the former reliably. It shouldn't own the latter in a regulated context.
Step 4: Check the compliance surface. What does the regulatory framework actually require here? In most cases, AI handling admin, routing, and drafting — with a human reviewing before anything consequential happens — clears the bar. That's a workable model.
Step 5: Pilot on one workflow end-to-end. Not a proof of concept. Not a demo. A real workflow, with real users, tracked against a real baseline. That's what tells you whether the ROI is there.
The companies getting real returns from AI right now aren't doing anything exotic. They found a workflow where work was piling up, automated the predictable parts of it, and measured the difference.
That's reproducible. It doesn't require a transformation initiative. It requires a clear-eyed look at where time is actually going — and the discipline to start with one workflow instead of ten.
If you're in a regulated industry, you have real constraints. But you also have real volume, real admin overhead, and real handoff problems that AI is well-suited to solve. The opportunity is there. Most companies just haven't looked in the right places yet.
Dealing with a similar challenge?
We work with mid-market companies in regulated industries to build AI workflows that actually hold up.
Let's TalkSean Cummings
Founder of Laminar Flow Analytics. Specializes in AI workflow automation for regulated industries — medical device, financial services, and complex logistics operations.