In the whirlwind of AI hype, where grand visions of autonomous agents often dominate the conversation, some of the most impactful stories are being written by leaders who focus on the practical. We sat down with a chief information officer who has successfully guided his midsize company through the early stages of AI adoption, not by chasing futuristic trends, but by relentlessly pursuing small, tangible wins. This conversation explores the real-world journey of integrating AI—from training non-technical staff and automating tedious workflows to empowering departmental “ninjas” and navigating the crucial challenges of data governance. It’s a masterclass in turning AI potential into measurable business value, one targeted use case at a time.
Many companies see a significant jump in employee AI tool adoption after targeted training. Beyond usage metrics, how do you measure the real-world impact on productivity, and what does an effective AI workshop for non-technical staff look like? Please share some step-by-step details.
You’re right, the numbers can be dramatic, but they have to translate into something real. After we brought in a partner to run hands-on Gemini workshops, we saw our employee usage leap from a mere 20% to over 71%. But the number I’m most proud of is that nearly three-quarters of our people reported that the tool saved them up to an hour every single day. That’s the real-world impact. An effective workshop isn’t a lecture; it’s about empowerment. We started by demystifying the technology, showing them it’s just a tool to help with their existing jobs—writing better emails, summarizing reports, or brainstorming ideas. Then, we moved into guided, role-specific exercises. For example, our marketing team learned to generate creative copy, while our sales team practiced crafting personalized outreach. The key is making it directly relevant to their daily tasks, so they walk away not just knowing what the tool is, but how it will make their Monday morning easier.
A common early AI project involves automating a high-volume task like invoice processing. When you automated the nightly categorization of hundreds of invoices, what was the biggest technical challenge, and how did you manage the human side of transitioning an employee’s role as AI took over?
That invoice project was a huge milestone for us. We handle about 120,000 invoices a year, and we had a part-time employee manually categorizing customer artwork orders to feed data into our Oracle NetSuite ERP. The goal was to understand our end-users better for marketing purposes, but it was a mountain of busy work. Technically, the biggest challenge was the integration. We built the pilot on Google’s Vertex AI platform, using Gemini to do the analysis, but we had to create a custom SuiteScript within NetSuite to make the whole process seamless. Now, every single night, we automatically push 600 to 700 invoices to Gemini, and it handles all the artwork analysis and categorization flawlessly. On the human side, this was never about replacing someone. The employee doing that work was fantastic, and we saw an opportunity to elevate their role. By automating the administrative burden, we freed them up to become a full-time company representative, focusing on higher-value interactions that a machine could never do. It was a win for efficiency and a win for her career.
For many midsize businesses, the advice is to focus on small, targeted use cases. How do you identify and prioritize these “small win” projects? Could you walk us through the criteria you use to decide which workflow to automate next and how you measure its specific return on investment?
My philosophy is simple: hunt for the friction. I tell my team to look for the places where we’re spending a disproportionate amount of time or money on a repetitive, low-value task. That’s your starting point. The criteria are straightforward. First, is the process well-defined and rule-based? AI thrives on that. Second, what is the resource cost? Are we paying a person, or multiple people, significant hours to do this? And third, what is the downstream impact of automating it? For the invoice project, the ROI was clear: we eliminated the cost of the manual work and simultaneously got faster, more consistent data for our marketing efforts. The measurement isn’t always a hard dollar figure, though. Sometimes the return is improved employee satisfaction because they’re no longer stuck on tedious tasks, or it’s faster customer service. You have to look at ROI holistically—efficiency, employee morale, and the journey toward becoming a more mature AI-driven organization are all part of the equation.
Some organizations find success by cultivating “technology ninjas” within business units like marketing or finance. How do you empower these champions, and what role do they play in identifying use cases that IT might overlook? Please provide a specific anecdote about how this collaboration works.
I absolutely love my “technology ninjas.” They’re my eyes and ears on the ground. These are curious, motivated people in accounting, marketing, manufacturing—everywhere—who are genuinely excited about what technology can do. Empowerment starts with giving them access and permission. We provide them with the tools, like Gemini, and the training to use them effectively. Their role is crucial because they understand the nuances of their own workflows in a way IT never could. A perfect example is our marketing team. While IT was focused on core infrastructure projects, the marketing ninjas came to me with an idea for social media. They are now actively working on a plan to use an AI agent that can post content at optimal times, adopt different personas for different platforms, and engage with our audience. That idea would have never come from IT, because we don’t live and breathe social media strategy. It’s that partnership that uncovers the most creative and valuable use cases.
Implementing an AI voice agent for customer support is a popular goal. What are the key performance indicators for its success, and how do you design the system to both learn from interactions and know precisely when to escalate a complex issue to a human representative?
The AI voice agent is our next big win, and we’re being very deliberate about it. With about 500 orders moving through our building every day, the volume of simple customer inquiries is high. The primary KPI for success, initially, will be containment rate—how many calls for order status, information, or bill payment can it handle without human intervention, especially after hours and on weekends when we’re closed. The design is two-fold. First, it has to learn. Every interaction, especially those escalated to a human, will be used to refine its responses. Second, the escalation logic has to be crystal clear. We’ll set firm rules: if a customer uses words indicating frustration, asks the same question twice, or has a problem that falls outside a predefined scope like a complex billing dispute, the system must immediately and seamlessly transfer them to a live employee. The goal isn’t to create a wall; it’s to create a highly efficient front door that intelligently routes people to the right place.
As business units begin building their own AI solutions, CIOs must balance user autonomy with the need for data security and governance. What is the first practical step you recommend for cleaning up foundational data to support these new tools and ensure compliance?
This is the challenge that keeps CIOs up at night. You want to empower your people, but you can’t have a free-for-all. The very first, most practical step is to conduct a data-systems audit with an “AI readiness” lens. You have to map out where your most critical data lives. You’ll inevitably find that one department, like marketing, is operating on a modern, clean system, while another, perhaps finance, is running on a legacy platform that can’t support these new tools. That mismatch is a huge risk. Cleaning up your foundation means identifying those legacy systems and creating a concrete plan to either modernize them or create a clean, accessible data layer on top of them. You can’t build a strong AI house on a crumbling foundation of technical debt. Reframing the conversation back to these foundational investments is the only way to be successful and secure in the long run.
What is your forecast for AI adoption in small to midsize businesses over the next two years?
Over the next two years, I believe we’ll see SMBs move from broad experimentation to deep integration. Right now, many of us are still figuring out what agentic AI can really do; it’s a bit of a gray area. The immediate future is about clarifying that gray area. The first phase was giving everyone a tool like Copilot or Gemini and seeing what happened. The next phase, which we are in now, is about surgically applying AI to solve specific business problems—automating invoices, generating product designs, and handling customer calls. Looking forward, the next two years will be about connecting these isolated wins. We’ll see business units, empowered with their own AI literacy, start building agents that automate entire cross-departmental workflows. The focus will shift from “How can AI help this task?” to “How can AI transform this entire business process?” It will become less of an IT-led initiative and more of a deeply embedded, operational reality.
