At our 19th annual Innovate Conference in Washington, D.C., the air was thick with a particular two-letter acronym. You couldn't walk ten feet near the registration desk or the exhibitor hall without hearing it: AI.

I am acutely aware that I am not the first person to stand on a stage or sit at a keyboard to discuss Artificial Intelligence. In fact, if you open your LinkedIn feed or turn on the news today, the "noise" surrounding AI is deafening. We are currently living in a cycle dominated by FUD: Fear, Uncertainty, and Doubt. We hear sensationalized warnings that the machines are coming for our jobs, alongside utopian promises that we will never have to perform a manual task again.

My intention here isn't to add to that noise or to claim a monopoly on the "correct" vision of the future. Rather, I want to offer my perspective, honed over 35 years of leadership experience, on how we, as leaders, can navigate this transition without losing our professional souls. While the world evolves at a breakneck pace, our role as the "human anchor" has never been more critical. We are tasked with a complex challenge: harnessing the immense power and promise of AI while ensuring that the human remains firmly at the center of the equation.

What can AI do well in nonprofit finance, and where does it fail?

There is a fascinating dichotomy at play with AI. On one hand, the promise is undeniable. Many of us have experimented with tools like ChatGPT, finding ourselves amazed by the speed, effectiveness, and quality of results when tasked with routine work. In the realm of nonprofit finance, AI excels at handling reliable, routine tasks, from automated transaction coding to rapid anomaly detection. It is an undeniable force multiplier for operational speed and scale.

However, this great power comes with calculated risks that no executive can ignore. We must acknowledge that AI, in its current generative state, is designed to provide an answer, even if that answer is factually incorrect. Hallucinations and algorithmic bias are not just technical concepts; they are real-world risks that can lead to poor governance and flawed strategic decisions.

The greatest danger is not that AI will replace the CFO; the danger is that in leaning too heavily on the "easy button," our own critical thinking, analytical depth, and strategic insight will atrophy. If we stop questioning the data because the machine presented it in a beautiful chart, we have failed in our duty as stewards of our missions.

Where AI stops in restricted fund accounting

This is easiest to see in the work that defines nonprofit finance. A model can code a transaction, flag a duplicate, and summarize a variance. It cannot decide whether a cost is allowable under a specific grant agreement, whether a release from restriction is properly documented, or whether an allocation across three entities will hold up in a single audit. Those judgments rest on donor intent, grant language, and board policy, which are specific to your organization and rarely written down anywhere a model can read them.

A useful test before handing any task to AI: if the answer would have to be defended to an auditor, a program officer, or your board, a person owns it. If it is volume work that a person then reviews, AI is fair game.

This is the line our client advisory and accounting team works on every day, and it is why we describe the work as co-sourced rather than outsourced. The routine work can be automated. The accountability cannot.

Why the human has to stay the pilot

In my recent talks, I have used the metaphor of the Human Anchor. I think of the human as the "pilot" of this technology. While the machine handles the mechanical execution of the flight, the human pilot provides the vision, the destination, and the moral compass. AI is a tool for speed, not a substitute for thought.

When contemplating the integration of AI into your organization, I invite you to filter your strategy through these three foundational principles:

1. Quality in, quality out: the precision principle

AI is a mirror; it reflects the quality of what it is given. Higher-level results require precise input, clear explanation, and a deep understanding of the desired outcome. If we provide mediocre prompts or shallow, unorganized data, we will receive mediocre, and potentially misleading, results. As leaders, our job is to define the "prompt engineering" of our entire department: ensuring the data we feed into our systems is clean, compliant, and contextualized.

2. The iteration gap

There is a dangerous myth that AI makes work "instant." While the first draft might take seconds, real, meaningful results still require significant time and deep thought. I call this the Iteration Gap. It is the space between the first draft an AI produces and the final, polished product that has been refined through multiple iterations of human judgment. If you are not iterating on what the AI gives you, you aren't leading; you're just spectating.

3. The thought leader mindset

We should use AI to help us structure and word our ideas, but we must never allow it to replace our professional judgment. At JMT, we advocate for an "expert-first" approach. Our value as consultants and finance leaders lies in our ability to navigate the thornier aspects of nonprofit financial management: compliance, restricted fund accounting, and multi-entity stewardship. These are areas where human experience, empathy, and strategic alignment cannot be automated. Use AI to draft the report, but use your brain to determine if the report's conclusions actually serve the mission.

How should a nonprofit finance leader start?

Contemplating the use of AI is not a conceptual exercise; it must be a practical one. You don't need to be an AI expert to be an AI-capable leader. To move forward safely, I encourage my colleagues to adopt a deliberate roadmap:

• Adopt a growth mindset. Do not let the "noise" or the "fear" paralyze you. Learn how to use these tools in real, practical ways. Understand the specific improvements being integrated into the technology you use today, such as Sage Intacct's AI features or Microsoft Business Central's Copilot.
• Plan for human growth. This is perhaps the most important leadership task of 2026. As AI reduces the structural workload of routine tasks, you must plan for how daily work changes for your team. If your staff is no longer spending 20 hours a week on manual data entry, what will they do instead? We must plan to keep our team members involved, growing, and analyzing, rather than just "monitoring the machine."
• Stabilize, optimize, strategize, and innovate. Use the efficiency gained from AI to move your team up the value chain. If the "manual grind" is reduced, redirect that human energy toward operational decision-making and long-term financial transformation.

If you want a place to start that is not a tool at all, start with your data. Most of the disappointment we see with AI in finance traces back to the input rather than the model: a chart of accounts that grew without a plan, dimensions used differently by different people, and years of history nobody has ever cleaned up. AI is a mirror, and that is what it will reflect back. Fixing it is unglamorous data work, and it makes every tool you adopt afterward work better.

The pilot and the machine

The "new frontier" of AI is already here, and despite the FUD, it represents the single greatest opportunity for nonprofit finance in a generation. It allows us to move away from being "history bookkeepers" and toward being "future strategists."

But remember: the value of a force multiplier is determined entirely by what is being multiplied. If you multiply zero strategic insight by AI, you still have zero. If you multiply deep, human-led expertise by AI, you change the world.

As you head back to your organizations, I encourage you to bring your own growth mindset to this world. Learn the tools, understand the risks, but never let go of the steering wheel. Ensure that it is your quality, your reliability, and your critical thinking that lead the way, not the AI telling you what to think.

Let's lead our organizations into this future by being intentional, being skeptical when necessary, and always being "mission-driven."

Questions we hear from nonprofit finance teams

Will AI replace nonprofit accountants and CFOs?

No. The tasks most exposed to automation are the repetitive ones: coding, matching, reconciling, and first-draft reporting. The judgment work is not, because compliance, restricted funds, and multi-entity stewardship depend on context a model does not have. The role shifts toward analysis, it does not disappear.

What finance tasks can AI safely handle today?

Routine, rules-based, high-volume work where a person reviews the output: transaction coding, anomaly and duplicate detection, document summarization, and first drafts of reporting narrative. Treat every output as a draft, not an answer.

What are the biggest risks of using AI in nonprofit accounting?

Three. Confidently wrong output, because generative tools are built to produce an answer whether or not it is correct. Bias carried in from the underlying data. And the slow one, which is a team that stops questioning what it is handed. The third is the hardest to notice and the most expensive to unwind.

Is it safe to put donor or grant data into an AI tool?

It depends entirely on the tool and the agreement behind it. Features built into your accounting system, such as Sage Intacct's AI features, operate inside that system's existing controls and permissions. A general purpose chatbot does not. Before any data leaves your systems, confirm where it goes, whether it is used for training, and who at your organization approved it. Write that down as a policy before your team needs it, not after.

Can AI help with nonprofit budgeting and forecasting?

Yes, in a narrow way. It is good at drafting scenarios, spotting inconsistencies between versions, and explaining a variance in plain language. It is not good at deciding which scenario your board should plan around. Dedicated budgeting and planning tools are where that work belongs, with the assumptions owned by a person who can defend them.

How should a small finance team get started?

Pick one repetitive task that takes real hours and has a clear right answer. Run AI against it for a month alongside your current process, and compare the two. One honest pilot teaches you more than a year of reading about it.