Summary: Most nonprofit finance teams are still early in their AI journey. In a recent Aprio webinar, our specialists showed where AI is best positioned to help, identified the lowest-risk starting points, and explained how to score your organization across six readiness pillars. We found that starting with AI you already know is a strong starting point, but human review is still essential.
If your nonprofit organization is at the beginning of this AI journey, whether you’re still learning, investigating, or just starting to implement, you are not behind. In this article, we cover what AI is and what it is not, the terms you should know, low-risk places to start, how to gauge your readiness, and how your data shapes the outcome.
What Is AI, and What Is It Not?
AI is a broad term. Think back to your first use of a generative AI chatbot, then consider the various ways you use it in your current personal and professional life. AI can be a capable ally across your accounting processes, but it cannot take over the aspects that require interpretation, context, and judgment. Because AI is not conscious or sentient, it is important to be clear about the limits of how you can use it in your processes.
- AI is not an all-knowing oracle, and it can produce confident but incorrect answers.
- It has no original thought or lived experience.
- It is not automatically objective or unbiased.
- It is not a replacement for the human judgment of a well-trained technical accountant.
AI is not correct 100% of the time. We should treat AI as “an accelerator or an advisor,” and never as a factual, neutral, or accountable decision-maker.
What Do the AI Terms You Keep Hearing Actually Mean?
A shared vocabulary will help your team weigh opportunities clearly and help reduce confusion. Here are the five terms that are often used when speaking about AI:
Operational reporting:
Think of this as the dashboards and reports that your accounting system already produces. Not AI in itself; however, AI makes it faster and more efficient.
Process automation:
This automates routine, rule-based tasks by following a set of predefined steps to produce output.
Generative AI:
These are large language models (LLMs) that write and generate content. They are used to summarize meetings, draft narratives for operational reporting, etc.
Machine learning:
This process uses historical data and statistical modeling to make predictions or decisions automatically.
Agentic AI:
These are solutions trained on your data that help complete tasks and are typically used with other tools to build end-to-end solutions.
Understandably, people can tend to blur the last three together. An easier way to remember and differentiate: traditional AI answers a question, machine learning learns patterns from data, and agentic AI takes action to accomplish a goal.
In addition to the terms above, several other concepts in the AI world are similar enough in nature that they are often conflated:
1. Generative AI versus Predictive AI:
generative AI writes, summarizes, and drafts, while predictive AI forecasts, scores, and flags risks or trends.
2. Chatbots versus agents:
a chatbot answers a question or handles one task, while an agent can pull data, build a report, draft a narrative, and route it for approval.
3. The AI you own versus the AI you build:
the former are the tools already embedded in your Microsoft 365 or accounting system, compared with the latter, which are custom workflows you commission.
Where Should Your Nonprofit Start With AI?
For most, the best starting point is to look for work where the workflow repeats, the data is well understood, and someone can review the output before acting. Lower-risk examples include meeting and document summaries, accounts payable and accounts receivable workflow support, close and reconciliation preparation, reporting narrative drafts, knowledge-based question-and-answer, and data cleanup and classification. Let’s explore further:
1. Meeting and Document Summaries (Lower Risk)
Most video conferencing platforms now include recording or transcription, and those transcripts can be leveraged to create action items or draft minutes. This is a relatively low-risk entry point, with the main risk being mistranscription or misattributed speakers. However, with access to the recording itself, you can mitigate this risk and address any transcription ambiguities by reviewing the video. Our nonprofit consulting team members use transcription during departmental assessments, which helps them stay present in an interview instead of writing notes.
2. Excel Formula and Macro Support (Lower Risk)
AI can help with Excel-related work by assisting with complex formulas, formatting, and debugging. The output still depends on the quality of the source data you feed it, and human review is still vital.
3. Technical Accounting Research (Medium Risk)
AI can provide a serviceable framework for a technical question. However, you shouldn’t rely on it to interpret complex standards or unusual circumstances without professional review. When an AI prompt is unclear, it can make assumptions, so an appropriately trained professional should review it to reduce the risk of creating incorrect deliverables based on those assumptions.
What Can the AI You Already Own Do Today?
Before you buy a new tool or commission a custom build, you should look at the AI already sitting in your enterprise resource planning (ERP) or core accounting system for several reasons:
- It uses data already in your system of record.
- It operates inside your existing security and permission boundaries.
- It is often faster to turn on, sometimes within a day or two.
- It can cost less than a custom-built solution.
Because the data and permissions are already yours, you work within the governance your organization has already agreed to, without any extra effort.
How to Assess your Organization’s AI Readiness?
AI is only as useful as the data and workflows beneath it. As such, as part of this process, honestly assess where your organization stands across several areas, such as information technology (IT) infrastructure and security, your ERP and/or customer relationship management (CRM), data, reporting, and analytics tools. Take time to honestly assess where your organization is now, not where you expect it to be in a few months, to determine your overall readiness for AI.
Typically, we assess an organization’s readiness by reviewing its AI capabilities across six pillars:
- Strategy
- Data
- Process
- Technology
- Governance, and
- People
Use these six pillars to develop a rubric to score your organization’s AI readiness. For example, review each pillar and score your organization on a scale of 1-5, where a “5” means your organization has fully mastered the pillar and a “1” means you have not yet begun work. Your total places you in one of three bands:
- A total score of 6-14: Focus on building your foundations,
- 15-23: Run controlled pilots, or
- 24-30: Scale with governance.
Your read on the technology pillar may differ from what your IT leadership thinks, and your view of donor data may differ from your development teams; that is why this process must be a collaboration across all impacted parties in your organization. Use these five questions to compare your perspectives and focus your discussion:
- What organizational challenges are we trying to tackle?
- How can AI support our mission or strategic plan?
- Which repetitive tasks consume the most staff time?
- Do we have appropriate governance, with every relevant department at the table?
- Is our data accurate, comprehensive, and up to date?
Why Data Integrity Comes First
AI can surface patterns and generate insight, but it cannot overcome flawed data on its own. Poor data can cause reporting errors, compliance issues, reputational risk, and difficult donor conversations. Your team must design review processes for both sides of every exchange with your AI tools, including what goes into the AI and what comes out of it, before delivering any AI-created or -edited output.
What Does Responsible AI Adoption Look Like in Practice?
As part of our work with clients, we have encountered many examples of clients putting AI to practical use within their current environment and processes. Here are some examples of low-risk, low-barrier applications that the Aprio team has worked on both directly and indirectly:
A Foundation With No Prioritized Roadmap
Our nonprofit systems team worked with a client whose ERP environment needed review, with fragmented analytics and reporting and inconsistent data entry. It needed a clear picture of its current state and architecture, stronger operational and financial visibility, and an ROI-driven roadmap. Aprio delivered a fit-gap assessment and an AI and data readiness assessment to help leadership and IT build their own AI roadmap.
A Nonprofit Insurance Provider With Limited Internal Staff
Another client was taking on a digital and AI transformation with manual workflows, legacy data spread across several systems, a CRM rollout ahead of it, and no AI roadmap. They needed help aligning their tech stack and defining a practical path forward.
Aprio acted as an outsourced digital and AI transformation office, supported the new CRM implementation to align with the existing tech stack, and developed an AI strategy and roadmap.
A Foundation Using AI for Grant Agreement Analysis
Lastly, our nonprofit consulting group worked with a client who experienced a full accounting team turnover from one audit cycle to the next. They lacked strong internal processes for revenue recognition review, and with revenue not finalized for the year, they felt immense pressure as an audit deadline approached.
They decided to use AI to take a first pass at reviewing grant agreements for the year to determine proper revenue recognition treatment. To start, the team went to IT and legal to help ensure they were operating under current policies. They redacted donor information to meet their AI policies and security infrastructure, then used AI to draft first-pass analyses of the year’s grant agreements. They also validated the approach by feeding in a prior-year agreement whose treatment auditors had already accepted; the analysis matched the earlier conclusion.
While it did save them time overall and helped them meet deadlines, the key point was that they did not shortcut the process by eliminating human review by their accounting team, which ensured their deliverables aligned with generally accepted accounting principles.
Final Thoughts: How AI Readiness Can Transform Your Finance Function
AI adoption is a progression, not a single project. The organizations getting value from it are not the ones with the most advanced tools; they are the ones that understand their systems, clean their data, and bring the right people and tools into the room first.
