Most AI webinars only show you the wins. This isn't that.
On May 21, Kyle Paalman and Pete Terryn spent an hour with a room full of business leaders walking through nearly four years of real AI implementation: what worked, what didn't, and what they'd do differently.
Between them, they've watched AI become part of daily operations, built autonomous workflows that save teams hours every week, seen promising ideas stopped by software limitations, and experienced their own I.T. department pushing back on internal AI deployments for security reasons. Those experiences shaped the discussion.
Here's what they shared.
AI has changed. What we do with it has, too.
Pete's enthusiasm for AI started early. He found his first internal email about ChatGPT dated December 2022, barely a month after OpenAI made it public.
"I'm a sci-fi kid," he said. "I lived long enough to see all these movies that I watched as a kid… they're becoming a reality now."
He started evangelizing language models across the NuWave team. Technicians began using AI to troubleshoot issues. Leadership used it as a thought partner for planning. Sales teams refined outreach with it.
What's changed over the past year isn't the excitement around AI. It's the shift from language models to agentic AI.
You used to ask AI a question and wait for a response. Increasingly, AI can complete work on your behalf.
Every morning, one of NuWave's internal AI agents scans competitors' websites and social channels, identifies promotions, new service offerings, and office openings, then delivers a briefing directly to the sales team's inbox. The process runs autonomously.
"It's a great way to leverage AI to work on your behalf," Pete said.
Kyle shared a very different example from earlier in the journey.
The vision was to build a Microsoft Copilot agent that could coach a sales intern through cold-call practice and provide feedback. The project never reached that goal.
"Not all of that vision came true," Kyle admitted.
Microsoft's terms of service prevented Copilot from acting in a manager-like role by evaluating employee performance.
After several years of implementation, they've learned an important lesson: AI projects don't fail only because the technology isn't capable. Sometimes governance, security, or platform limitations become the obstacle instead.
One Client’s Workflow Story
One food manufacturer was spending more than ten hours every week on a repetitive document process.
Every week, employees received certificates of authenticity from vendors, manually transferred the information into the company's own documents and software, and then forwarded those completed certificates to customers.
Their I.T. lead was spending hours simply keeping the workflow functioning. The company had already invested significant time and money trying to automate the process using OCR, but it didn't work.
The problem wasn't scanning documents. The problem was that every supplier formatted certificates differently.
"Because that would shift when they get it from different vendors, they were unable to make those adjustments," Pete explained.
Small layout changes repeatedly broke the OCR workflow.
During Prescott's AI Assessment, interviews with stakeholders uncovered more than a dozen opportunities where AI could improve operations. This workflow quickly rose to the top because it consumed valuable employee time and had resisted previous automation attempts.
Instead of relying on rigid templates, Prescott built an end-to-end AI workflow capable of handling the document variations that had defeated the OCR approach.
The manual process was eliminated, returning more than ten hours each week to the team for higher-value work.
The hours saved matter, bbut Pete pointed to a larger lesson.
Until AI matured to where it is today, that workflow simply didn't have a practical solution. The technology finally caught up to the problem.
Two companies, one offer.
When you're solving an AI problem, you don't want your network engineer moonlighting as an AI strategist. And when you're securing your infrastructure, you don't want your AI consultant guessing at your firewall.
NuWave's wheelhouse is the I.T. foundation:
-
Network infrastructure
-
Cybersecurity
-
End-user support
-
The systems your business runs on
Operational efficiency, custom AI solutions, workflow automation — that's Prescott's lane. Two dedicated practices, each built around a specific skill set.
Kyle explained the decision to structure it that way. "It was important to us that we really spun that component out of NuWave," Kyle said. "It's a different skill set, different people, different talent."
You get both. Neither is a side project.
AI data security isn't a marketing claim. It's a fight.
"You would think, with me running operations, that I could make things happen in the organization," Pete said. "But when I started asking for AI access to more and more tools within our ecosystem, I got a lot of pushback."
That's exactly how it should work, and Pete was thankful for it.
What data is the tool accessing? Read or write? What do the permissions look like? Who gets access to what gets aggregated? These are the questions every AI deployment should survive before it goes live.
That internal accountability shaped how NuWave and Prescott now approach every client engagement.
The conversations every business is having right now showed up throughout the security discussion:
-
Leadership wants to deploy AI on company devices
-
Employees are asking AI to access their email and calendar
-
Shadow AI, as people install tools without approval
NuWave addresses these proactively and applies the same scrutiny it survived internally.
The Q&A: what attendees wanted to know
The hour could have gone longer. Here's what was on people's minds.
What's the biggest AI security mistake companies make?
Buying a stack of Microsoft Copilot licenses and turning everyone loose. By default, Copilot starts indexing everything each user has access to: SharePoint, OneDrive, and shared folders. Kyle added a real example: a shared folder where one employee dropped personal information, and now anyone with Copilot access could surface it with the right keyword search.
Pete called it the single most common mistake he sees. "Lack of a plan. If you don't sit down and be intentional about your AI implementation, you're already exposing yourself to some risk."
Which AI tools are worth paying for?
Pete's rule of thumb: use the free version until it stops doing what you need, then jump to the $20/month tier. That's where most users will plateau for a long time. The $100–$200 tiers buy more tokens and earlier access to frontier models, but the average user won't notice the difference.
Is AI going to replace our jobs?
Both Pete and Kyle were emphatic. "We are not going to replace people with AI," Pete said. The goal is to grow the company faster than the team, keeping headcount stable or expanding it more slowly than revenue. That makes the business more profitable while protecting roles. They repeat that message at every quarterly meeting.
What's next.
One theme surfaced repeatedly throughout the hour: businesses rarely struggle because AI isn't capable enough.
More often, projects stall because of unclear governance, unrealistic expectations, or security questions that weren't addressed early.
The technology will continue to evolve, but successful AI adoption still depends on understanding your workflows, protecting your data, and solving problems that are genuinely worth solving.
Your network is the foundation everything else runs on.
NuWave helps businesses keep that foundation secure, stable, and ready for what's next.
If you're ready to identify where AI can create meaningful value inside your organization, Prescott's AI Assessment begins with stakeholder interviews, identifies your highest-impact opportunities, and delivers an AI strategy document you keep as your roadmap for implementation. From there, Prescott can help bring that strategy to life.
AI may be the destination.
A secure network is still the road.
Kyle Paalman is President of NuWave Technology Partners. Pete Terryn is Chief AI Officer at Prescott and co-founder of the West Michigan AI Lab. Together, they've tested AI in their own business, built custom workflows for clients, encountered real-world implementation failures, and developed governance practices shaped by those experiences.
