After my internal controls session, I was thrilled to hand the virtual stage to our friends at Synergy Technical for the portion of the symposium many of you came for: AI for the nonprofit sector. Cathy Redford introduced the Synergy team, Don Lewis shared what he is seeing from his work in licensing, Jeff Lau laid out a practical roadmap for getting started, and Thyler Courter of Evansville Goodwill Industries walked us through a real-world use case. Here is a recap of the session, including highlights from the Q&A.
How This Session Came Together
About four or five months ago, I ran into Cathy Redford at a networking event. As we talked about our work, we realized we both serve nonprofits. When I asked whether she was seeing AI used much in the sector, she said they're starting to, and that they really want to do more. I asked her to come speak to our nonprofit audience, and she brought her team with her.
About the Company
Cathy, who serves as vice president of business development, shared some background. Synergy Technical was founded in Richmond, Virginia, in 2011 by a global CIO who couldn't find cloud resources and bet her career on it. Cathy joined in 2013 to build out the sales organization, bringing experience from IBM, Arthur Andersen, and a long career in technology leasing.
The company started in a kitchen in Midlothian, Virginia, and has scaled significantly. Synergy has deployed more than 20 million seats of Microsoft cloud solutions, has clients in all 50 states, and has done project work in more than 70 countries. Its industry verticals include nonprofit, energy, manufacturing, retail, pharma and healthcare, banking, and legal, with clients ranging from small and mid-sized businesses to enterprises in both public and private markets. The team likes highly complex, highly regulated environments. Services include cloud implementations, technology strategy, AI enablement, IT managed services, licensing and advisory services, and technology carve-outs and consolidations, along with a strong private equity practice supporting PE firms' portfolio companies.
From Counting PCs to Counting Users to AI
Don Lewis has been with Synergy since 2017 and manages its licensing practice, working with nonprofit and commercial customers alike. He has made a career of licensing, including a tour at Microsoft. He joked that he still knows where a few skeletons are buried, and admitted that Microsoft licensing is sometimes compared to the tax code. His goal is to make it make sense for customers, and one of his main tools is an annual license assessment.
Don noted that licensing historically meant counting PCs, and then shifted to counting users. In his view, AI is a fundamental shift in what you're actually getting out of what you buy, and he believes nonprofits are very well positioned to take advantage of it. As he put it, you can never have enough volunteers in the nonprofit space, but AI is tireless.
What He's Seeing in Nonprofit License Reviews
Within about a month, Don completed licensing reviews for five nonprofit customers. The consistent theme: all of them were leaning into AI. None were refusing to engage with it. Everyone, Synergy included, is concerned about cost. Don pointed out that the industry is moving toward consumption-based pricing, with some tools charging by the task, and that Microsoft is heading more in that direction, which makes costs hard to predict. Synergy's advice is to focus on control rather than prediction: figure out what your budget needs to be, and then put controls in place to maintain it.
Why Smaller Organizations Should Lean In
Don said he was a little surprised by the results of my earlier poll, where most attendees reported not using AI in a significant way. I had commented that smaller organizations may not be embracing it as much, and Don argued it should be the opposite. He sees AI as a way for small organizations to act big and deliver services at a higher level, particularly in fundraising, grant writing, and marketing. You don't have to be an expert in those areas anymore.
On a personal note, Don shared that he has never liked creating PowerPoints, and now he can write a couple of sentences and get a deck back. He also admitted to some skepticism as a member of what he called the Skynet generation. Still, his encouragement to nonprofits large and small was to embrace AI as a way to provide a higher level of service without adding staff.
He also noted that it's sometimes easier for an outsider to come in and ask why you're doing something a certain way, which helps break what he calls the "way we've always done things" syndrome. The Synergy team offered to talk with organizations about these areas and share their learnings.
Jeff Lau of Synergy Technical took us from the "why" of AI to the "how." Jeff joined Synergy in 2021, and his main role is helping clients adopt AI: assessing where they are today and building an implementation strategy, whether they're starting from ground zero or already experimenting. Here's his practical roadmap.
What Attendees Told Us
Jeff polled attendees throughout his session. Asked where they would start if AI could eliminate one major source of work, most chose administrative and reporting tasks. Asked which tasks they'd be most comfortable handing to an AI agent, most wanted to start with simple tasks first. And when asked what's holding their organization back from using AI more, most chose data privacy and security. Jeff said that makes complete sense, and that it's something you want tight before pushing AI out to the whole organization.
Three Moves in 60 to 90 Days
Jeff acknowledged that getting started can feel overwhelming given the number of strategies and tools available. He breaks it down into three practical moves an organization can make in a quarter:
The general path he follows with clients: start with an introduction to AI and training on quick wins, move to hands-on sessions applied to real work, identify use cases for agents, and surround it all with governance. The key is that you don't have to solve every problem at once. Start small, demonstrate value, and expand slowly.
Use the Tools You Already Have
You don't need to buy a bunch of new tools to begin. Many organizations already have secure AI available. For example, organizations with Microsoft 365 have access to Copilot Chat, a free AI chat tool similar to ChatGPT or Claude, but secure, so what you enter stays within your tenant boundary. CRMs, meeting platforms, marketing tools, and other applications you use may also have AI built in.
Jeff drew a clear line between enterprise-approved tools and consumer tools. Tools you give employees should be researched, understood, and officially approved. Otherwise you get what he called tool sprawl, where everyone uses a different tool, it's hard to understand the value you're getting, and it's hard to govern everything at once.
Get Off the Blank Page
One of the most valuable immediate uses of AI is what Jeff calls the blank page. Most people aren't looking for AI to make the final decision or produce a perfect document. They want help getting started. Examples include donor thank-you notes, email campaigns, program outlines, meeting agendas, and transcribing a weekly team meeting into a summary. Think about your daily tasks and where you could delegate that first draft.
Jeff emphasized that you never remove the human from the loop. Someone should always review, approve, personalize, and check accuracy before the work is used.
Understanding Agentic AI
Once you're comfortable with chat tools, agentic AI is the next step. The biggest difference is moving from asking AI for a response to delegating a goal or outcome. With a chat tool, you might ask for a document summary. With an agent, you give it a goal, connect it to information sources (such as your intranet or external websites), define the steps or actions it should take, and let it work through the process. For example, an agent might review an incoming request, identify what the person is asking for, locate relevant information across sources, prepare a response, and route it to the right employee for review.
From the many agents Jeff has built alongside organizations, his conclusion is that building the agent isn't the hardest part. Understanding the underlying process is. Map it end to end: how it starts, what information is needed, what decisions are made, and what exceptions exist. Without a good process, you typically won't build a useful agent, and you may find you don't need an agent at all. A chat tool might do the job.
Examples he has seen work well at nonprofits include:
An AI Usage Policy People Can Actually Use
Jeff stressed that an AI usage policy is key at the beginning of your journey, because there are safe and unsafe ways to use AI. It doesn't need to be long or complex, but it should answer: What tools can I use? What information can I enter? When do I need approval? What should always be reviewed by a person?
One approach he likes is a green, yellow, and red method:
The goal isn't to stop people from using AI. It's to give them the clarity to use it confidently every day.
Stand Up an AI Council
Jeff has had a lot of success helping organizations create an AI council made up of frontline users from different departments, each bringing a different perspective. The council might meet once or twice a month depending on your progress, and it can create training sessions, host office hours, share success stories, and carry enthusiasm back to their teams. You also want IT and security working alongside the council. I'm a big fan of this idea. When we implemented AI at our firm, we had a committee too, and I don't think we would have seen as much use or passion without it.
A Question of Conscience
One attendee raised a question I want to acknowledge here. Their small nonprofit colleagues are asking whether they should use AI at all, given that they're trying to better their communities while seeing concerns about the technology's environmental impact and limited regulation. They described it as a heavy cost to weigh. I told the group honestly that I feel the same way, and I understand where this person is coming from. Jeff said he understands that perspective too, and reiterated his advice to start small and find some quick wins so you can build comfort and confidence over time rather than changing how your organization works all at once. Thyler Courter also spoke to this theme in his session, which I cover below.
One of the highlights of the session was hearing a real-world use case straight from a nonprofit. Thyler Courter, system administrator for Evansville Goodwill Industries, walked us through how his organization is using AI in ways that go well beyond chatbots.
Meet Thyler
Thyler explained that each Goodwill operates as its own regional organization. Evansville Goodwill covers southern Indiana and parts of western Kentucky and southeastern Illinois. Thyler has six years of IT experience, all in the nonprofit sector at Evansville Goodwill. Like many nonprofit staff, he wears many hats, specializing in Microsoft 365 administration, automation development, single sign-on, and networking, and he also manages the team's data analyst.
He came to IT from a logistics background. When he started, he automated processes in the logistics department, especially forms and paperwork, to cut down on repetitive work, and that's what got him noticed and moved into IT. He looks for ways to improve efficiency, automate repetitive tasks, and help the organization get the most value from its technology investments. In the nonprofit sector, he said, that's huge. He wants a real return on investment rather than frivolous spending on shiny objects.
A Practical Agent: The Handbook Helper
Building on Jeff Lau's discussion of agents, Thyler shared one his team built. Employees frequently asked about standard operating procedures and policies, so Thyler uploaded the SOPs, employee handbook, and similar materials into an agent and deployed it in Microsoft Teams. It sits in the left sidebar as a "handbook helper" where anyone can ask about policies and procedures. He acknowledged it isn't the coolest thing in the world, but it saves a lot of time and improves accuracy, because people aren't guessing or misremembering policy.
The Problem: Tally Marks on a Clipboard
Thyler noted that in networking with other nonprofits and Goodwill organizations, a recurring theme is the desire to find uses for AI beyond chatbots. Here's the one his team found.
Goodwill receives donations when people drive up to a donation door and hand items to donor greeters. For tracking, the team relied on a piece of paper on a clipboard: a greeter made a tally mark for each car, and the marks were added up at the end of the day. That introduced a lot of human error. Greeters would forget, or when five or six cars were lined up, they'd guess rather than make donors wait while they went back to mark the sheet. Evansville Goodwill wanted better accuracy and to eliminate a repetitive manual task.
The Solution: AI Car Counting with Existing Cameras
The team worked with a company called i3, which offers a cloud solution that connects to existing security cameras (or sells and installs cameras). i3 offers other capabilities, such as people counting and license plate reading, but Evansville Goodwill was primarily interested in counting cars.
The system uses a zone drawn around the area where donors typically pull up. It detects vehicles and ignores cars outside that zone. What Thyler particularly liked was how lightweight the solution was. Because security cameras were already in every location, the only purchase was a small media server, essentially a networked video recorder. Cameras connect to it, recording data goes to i3's cloud, the AI does the work there, and the team receives the analytics.
Testing and Results
They started with a pilot location. To measure accuracy, the team audited camera footage, which Thyler admitted was very boring but important, manually counting cars and comparing that audit count to both the employees' manual count and the AI count. They liked what they saw. The project began about a year ago, and the last location went live about a month before the symposium.
The team still audits each location at least quarterly over a three-day spread, with staff continuing to count manually as accuracy is tracked over time. Results vary by location, since people differ, but some locations' manual counts were around 75% accurate, while the AI count averaged about 92% accuracy. There will always be some mistakes, but the improvement has been significant.
Insights Beyond the Count
The system also measures dwell time, meaning how long donors wait. That data has helped managers schedule more help at the door during peak hours, something they didn't have the data to do before. It also shows where bottlenecks are, which locations may need more marketing support, and how weather affects traffic, such as lighter traffic on hotter days and the need to schedule more staff on nicer days. Because the organization relies on donors to keep things moving, a better donor experience matters.
What's Next: AI-Assisted E-Commerce Listings
Thyler also shared an idea that came to him during the session. Anyone who has sold something on eBay or Facebook Marketplace knows you have to take photos and write up a description and condition. Evansville Goodwill is looking at a system where an employee sets an item down and takes a picture, and AI identifies the item, brand, and condition and creates the full listing. The employee reviews it for accuracy and hits publish.
When an attendee asked which tool this is, Thyler explained that ShopGoodwill.com, an eBay-style marketplace for Goodwill organizations, is developing the tool specifically for other Goodwills. For other organizations, he suggested researching mobile apps that can help create listings for whatever marketplace you use.
Technology as a Tool, Not a Replacement
An attendee asked whether the car-counting technology eliminated jobs or reduced hours for donation lane employees. Thyler's answer was an emphatic no. He advocates for technology as a tool rather than a replacement and does not want to automate people out of their jobs. A human still needs to greet donors, accept donations, offer receipts, and invite them back. The AI simply takes the tally marks, and the end-of-day paperwork of scanning and sending the sheet, off their plates so they can focus on the donor experience and on keeping the lane swept and safe.
Thyler also spoke to sustainability, a concern raised earlier in the Q&A. He noted that everyone working in the nonprofit sector cares about making the world better, and that it's easy to get caught up in frivolous uses of AI. He'd rather focus on real-world uses that make a greater impact on the mission. Evansville Goodwill's mission is to help people thrive through the power of work, empowering people to help themselves, and he said they've seen a lot of success using AI to help teach people and expand their reach.
Thank you, Thyler, for such a great use case. It was a perfect example of what AI can do beyond the chatbot.
We closed the session with an open Q&A with the Synergy Technical team and Thyler. Here are the highlights.
Who Owns Enforcement?
I brought forward a question an attendee asked at the end of my internal controls session: You can create a committee and write policies, but who ultimately makes sure people follow them?
Thyler shared how Evansville Goodwill handles it. As head of IT, he writes the policies, which are approved first by his supervisor, then by the CEO, and sometimes by the board of directors. Enforcement, in general, falls to direct supervisors. For example, a regional director is responsible for making sure a store manager follows policy, and can escalate questions or confusion up the ladder. IT has a lot of visibility into the environment but can't watch everything. Thyler can't stand over people's shoulders to stop them from sharing passwords, so IT writes the policy and handles violations through discipline when they're discovered. Controls also help: multi-factor authentication, for instance, makes password sharing much harder. His bottom line is that direct supervisors are ultimately responsible for enforcing policy for the people they manage.
Cathy Redford added that Synergy spends a lot of time with key stakeholders and leadership teams developing codes of conduct. Synergy has a very clear AI code of conduct internally, and every client signs an acceptable use agreement that spells out which subscriptions can and can't be used, which data can and can't be entered, and why. Leadership and key stakeholders play a role, and if you formalize an AI committee, which Synergy recommends, you should have governance and security guardrails in place. Because AI changes quickly, Cathy stressed that this must be an ongoing cadence as you review potential use cases and run proofs of concept.
Jeff Lau agreed, recommending that the council meet at least twice a month to discuss whether the policy is working and what needs to change. He described the policy as a living document.
Transcription Tools for In-Person Meetings
An attendee asked for a recommended tool for transcribing in-person meetings, not just Teams or Zoom. Cathy noted there are so many tools that it's hard to be specific without understanding your environment and what you've already invested in. Synergy always tells people to start with the low-hanging fruit, which is what you already own, since many organizations aren't fully aware of the capabilities they have.
One panelist mentioned using Notability for a unique need, documenting conversations with a builder during a cabin remodel, but was clear that it wasn't an endorsement. Thyler shared that he's seen people start a Teams meeting on their device and leave the microphone on to transcribe an in-person conversation, using a tool they're already comfortable with. An attendee also suggested that Google offers a nonprofit platform and that Google Meet works easily for remote board meetings.
Cathy's overarching advice: make sure you understand the tool, what it does with your information, and that it's locked down with proper security and data governance. Synergy always starts foundationally with security and builds from there.
A Note on NotebookLM
During Jeff's session, an attendee asked whether a shared Google NotebookLM is more secure than broader AI tools. Cathy, who recently completed a Google certification, explained that security depends on the underlying settings of the entire environment, including governance and how identities are set up, rather than just the specific tool.
As the session wrapped up, the Synergy team noted that hearing from a customer about real use cases is important, and that the one thing everyone can agree on is that AI will evolve, and quickly. It's up to all of us to keep learning, approach it with caution, and recognize that it's here to stay.
Thank you again to our friends at Synergy Technical and to Thyler for a great use case, and thank you to everyone who attended. I hope you took away at least one or two nuggets to implement at your organization.
We hold these symposiums three times a year, and our next one will be in January. After each event, we send a survey asking for feedback and topic ideas. Please be candid and share any technical topics or other subjects you think would benefit nonprofits. We're always happy to bring in a guest speaker.
Finally, in late October we're launching our first-ever nonprofit-specific podcast, BEyond the Misson. I'll be leading it, and episodes will run monthly at about 20 minutes each, covering technical topics as well as mission-driven conversations with our nonprofit clients about what they're doing at their organizations. Watch for it on social media. It's another way to stay connected with us.