Menu
The talking heads keep shoveling hype and making everyone feel behind. You're not.
Most of the companies we talked to recently are still getting the middle of the org to use what they already bought.
That's normal. It's just not what you'd guess from scrolling LinkedIn, where everybody seems to have a fleet of agents running their week. I get why it happens. Urgency gets attention, and "you're already late" is an easy thing to say when you don't have to sit in the room with the team that has to do the work. But it doesn't match what we see when we sit down with actual organizations, and it pushes people to skip steps they can't really skip.
One leader told us recently that her team feels slow, but everyone they talk to keeps telling them they're right on time. That's closer to the truth than most of what's trending.
Here's how most of them got here. The tools got rolled out with very little guidance or training, and people were expected to learn it on their own time. Some did. The self-starters dug in and filled the gap themselves. Most people didn't, and I don't blame them. Turning the tools on isn't the same as teaching people how to use them in their actual work, and in a lot of organizations nobody really owned that second part.
When we sit down with a team, they're usually in one of three modes. Training first, then workflow redesign, then custom builds. Here's what each one looks like in practice, how you know you're in it, and what to do next.
The tools went live 18 months or more ago, and usage is all over the place. A few people are building agents. Some never opened it. The middle is still asking what it's for. 75% of the companies we talk to are in this stage of adoption & enablement.
In practice it sounds like basic questions, and that's fine. Where's the internal hub? Is my email connected, or do I have to add it? Can I hand a project to someone else, or do I just keep passing a link back and forth? We also hear about people dropping a file in and expecting it to magically produce whatever they wanted. When it guesses, they decide it doesn't work.
How you know you're in it:
What to do next: train on real work, with the tools people actually have. Show it first, then teach it. The goal is for the whole team to have the same working knowledge, like what context to give before asking for anything and why you don't accept the first answer that comes back.
There's a big difference between one person getting good at AI and a team or a whole org getting good at it. One person can try the newest model the day it ships. A team gets whatever the AI usage and governance policies allow, and the latest models often aren't on the approved list.
That's not a failure. It's how organizations with customer data, regulated work and a security team are supposed to operate. But it means a lot of the advice you see online doesn't apply to you. The demo you saw probably used a model your org hasn't approved, connected to systems your security team won't open up.
We see this all the time. In one organization, a group ended up on a different approved tool than the rest of the company, and they felt like they'd been handed the weaker version. When someone showed their leader what it could actually do, the reaction was "I had no idea you could do this." The gap wasn't the tool. Nobody had shown them. Another team can only reach its files through one approved connector, with limits on how much it can hold. That's the real environment, and enablement has to start there, with the tools people actually have.
Once people have the basics, we sit with the team and look at the work they already do. Where does it slow down? Where does AI actually help? Where does the handoff always break?
One line from a client keeps coming back to me: "to expedite a process, you gotta know the process." A lot of the time, nobody has written the process down. It lives in one person's head. Someone told us, pretty candidly, that there are basically no processes there. An operations leader with business units that "work pretty differently" figured they might need a full workflow redesign first, with technology added on top after that.
How you know you're in it:
What to do next: pick one everyday workflow. Write down how it actually works today, including the judgment calls that only one person knows. Decide together where AI fits and where it doesn't, then run it on real work until it holds. Put an agent on a process nobody wrote down and you get the same mess, faster.
Then we build internal tools around the workflow. Sometimes that's agents in Copilot. Sometimes it's custom software with agents in the background. Either way, it's measured against KPIs you've already set, with custom reporting so you can see whether it's working.
That part matters to us. We're not inventing a new scorecard to make the build look good. If your organization already tracks turnaround time, review rounds or hours spent on a recurring report, that's the bar.
The leaders who are ready for this mode tend to describe it plainly. One told us she sees these projects as pilots: working engagements that end in a deliverable that's "usable, it's working," and that the team can repeat. Another, with a tool already launched, wanted to know who owns the feedback and the fixes once it ships. Those are the right questions to be asking.
How you know you're in it:
What to do next: scope one build against one workflow. Use the KPIs you already report on, and set up the reporting before launch so you're not reconstructing a baseline later. Decide who owns it once it ships.
So if you're still on step one, you're not behind. You're doing it in the right order.
Most of the organizations we sit with aren't late. They're early in a sequence a lot of people online are acting like they've already finished.
If you're not sure which mode you're in, the free AI Operations Reality Check is a short read on one everyday workflow.
If you already know which workflow is stuck and want a written read on what to do next, that's the AI Operations Review. Either way, it's not an open-ended contract and not another rollout. Scot and Susan Westwater run every engagement themselves.
Which mode are you in?
Scot Westwater is the CSO and Co-Founder of Pragmatic Digital. He is an architect of practical AI operating systems that help operations and marketing teams move from robotic output to governed, brand-safe workflows. With over 25 years of building digital platforms for Fortune 500 brands, Scot focuses on turning AI experimentation into repeatable, measurable processes that drive real business impact. He is a co-author of Voice Strategy and Voice Marketing.
Weekly insights on AI adoption, workflow, and what's actually working in mid-market organizations.