I hope this helps.
When you’re coordinating a server infrastructure deployment across 50 sites in six states, “ready” means different things to different people. The technical team needs to know the infrastructure is ready for the change. The people at each site need to know when it will happen, what it means for their work, and whom to call if something goes wrong. Both answers matter when you’re deciding whether to go live.
I’m a National Account Manager at Mid-South Telecom, and my work spans technical consultation, solution design, project coordination, and deployment. That has taken me from server infrastructure to a 440-camera surveillance deployment across 32 locations, as well as work with healthcare and education organizations. Seeing projects through from the first conversation to implementation has shaped how I think about adopting AI.
Start with the work people are trying to do
A request for new technology often comes with a solution already in mind. My first job is to understand what prompted the request. Where does the current process break down? Who has to work around it? What would a better day look like for those people? The answers tell you what the technology needs to accomplish and give you a way to judge whether it did.
Say a small business wants to use AI to respond faster to routine customer inquiries. Start with one type of inquiry and give the tool an approved source to work from. Have the people who already handle those requests review the drafts. Measure the time it takes to get a response ready to send, including review and corrections. A draft that takes seconds to generate can still take longer to verify than it would have taken someone to write.
Listen to the people who handle the exceptions
On a multisite project, a plan can look consistent on paper while the work varies from one location to another. The people at each site know which tasks can wait, which interruptions would cause trouble, and what they already do when the usual process fails. Bringing them in early gives you a chance to account for those differences before rollout.
For an AI pilot, I’d ask the team to bring a few examples that already require judgment. What happens when a customer asks about an exception to policy? What if two approved sources give different answers? Those examples help you decide when the tool can produce a useful draft, when it should ask for more information, and when a person needs to take over.
Make sure someone can own the decision
This is what I mean by People > AI. If an AI tool drafts a customer response, someone owns the promise that response makes. That person needs to know where the answer came from and have the time and authority to correct it.
It’s also part of why I started TheNomadicIT. In The Infinite Machine, I explore the ideas and discoveries that shaped computing, robotics, and AI. In AI Cyber Brief, I look at what developments in AI, cybersecurity, and privacy mean for decisions people make at work and at home. I want people to understand enough about the technology to ask better questions and feel confident making their own calls.
Before expanding an AI workflow, I’d ask the people using it three questions:
- What part of your work has actually gotten easier?
- Which answers still take the most time to check, and why?
- When the tool gets something wrong, who can correct it or pause its use?
Ask those questions while the pilot is small enough to change. The people closest to the work can tell you what the results look like in practice and what needs to happen before you expand it.
As always, I hope this helps.