Looking forward to it!

As Blackbaud rolls out new AI tools, how are you making sure they’re environmentally responsible? For example, are you looking at things like the energy required to train and run these models, and how does that fit into your broader responsible AI approach? How should charities using Blackbaud respond to our supporters concerns about environmental impact of our tools?

Thanks for the question - I will start to answer these thematically in a few response :)

Responsible AI within Nonprofit Fundraising and Constituent Engagement

Responsible AI means defining, developing, and\or deploying AI that is ethical, transparent, and mission-driven - always keeping the social impact community at the center. That suggests it’s not just a technical standard… it’s a commitment to fairness, reliability, and trust.

For more information, please visit: https://www.blackbaud.com/intelligence-for-good

For nonprofits, that means the use of AI should amplify human judgement, ensure fairness and inclusivity, operate transparently so decision can be understood and trusted, prioritize human agency, honor data and donor relationships, and be mission-driven.

Some great resources for us all to leverage include:

• Responsable AI Institute: https://www.responsible.ai/
• Open Data Institute: https://theodi.org/
• Ethics Guidelines for Trustworthy AI: https://digital-strategy.ec.europa.eu/en/library/ethics-guidelines-trustworthy-ai
• FundraisingAI: https://fundraising.ai/

I would be interested in this community’s recommendation on additional resources for us all to leverage.

You noted that very few nonprofits currently have an AI policy in place. Do you have examples of AI policies for charitable groups?

Sharing this Blackbaud Institute Spotlight as well, as it references AI trends as well: What Will Giving Look Like in 2023: https://institute.blackbaud.com/resources/giving-in-2035 - thanks to Dr. Laurene Currie.

And here is a great article from Carrie Goldberger - AI in Action: Deploying Responsible, Effective, and Trustworthy AI: https://blog.blackbaud.com/ai-in-action-deploying-responsible-effective-and-trustworthy-ai/

Carrie is an amazing Data & AI Product Manager and outlines a thoughtful intentional approach - another great resource for you all.

What would you recommend to someone who has not used AI before as a way to dip their toe in to get started? I saw comments about steps on an organization level, but what about on a person to person level?

For our current Gen-AI offerings, Blackbaud doesn’t build or train generative AI models. We use enterprise-grade tools and techniques like prompt engineering, RAG (retrieval augmented generation), and more to create helpful, secure features for our customers. This lets us improve the outputs over standard LLM's for our customers so that we get the benefit of emerging models with billions of parameters and improvements in things like hallucinations and performance - while applying our industry best practices.

I look at them as layering.

AI governance policies and data/analytics/business intelligence (BI) policies share a lot of common foundations, such as data quality, security, and compliance. But AI governance introduces additional layers of complexity because of the nature of AI systems.

Lets start with the key similarities

• Data Integrity & Privacy: Both require strong controls for data accuracy, security, and regulatory compliance
• Accountability: Clear ownership and auditability are essential in both domains.
• Transparency: BI and analytics policies emphasize clear reporting (note: AI governance extends this to explainability of models.

Now lets talk about the additional layers of AI governance

• Ethical Considerations: AI governance explicitly addresses fairness, bias mitigation, and societal impact.
• Model Lifecycle Management: AI policies govern training, validation, monitoring, and retraining of models.
• Risk Management: AI introduces algorithmic risk (ex: unintended bias, drift), requiring continuous oversight beyond standard data governance.
• Human Oversight: AI governance includes human-in-the-loop for critical decisions.

Overall, BI policies ensure accurate, secure, and useful insights from data. AI governance builds on that foundation but adds ethical, technical, and operational safeguards to ensure AI systems are trustworthy, explainable, and aligned with your organizational values.

I would say – just be curious! We are all learning 😊

I’ll start with my favorite books – Prediction Machines and Power and Predictions ( https://www.predictionmachines.ai/ ) and AI Factor ( https://theaifactor.ai/ai-factor-book/ ).

Winning with AI: The 90 Day Blueprint for Success ( https://charleneli.com/winning-with-ai/ ) looks great as well. It's coming soon – but the authors (Katia Walsh and Charlene Li) are incredible and I’ve learned so much from them throughout my career.

Charlene also has a great newsletter – Leading Disruption: https://charleneli.com/ (scroll down to the bottom to subscribe)

If you are looking for technical training, there are some great, free courses out there as well – check out DataCamp or LinkedIn Learning – but there are others. Some have hands on skills labs to practice.

The FundraisingAI Global Summit has some great recordings shared as well: https://fundraising.ai/summit/ . I of course have to give a shout out to Sam Venable (Director, Data & AI Product Management) for his session on The Context Advantage: The Missing Piece in Nonprofit AI Strategy.

And then stay tuned for our AI Certification for Social impact in 2026: https://www.blackbaud.com/ai

I am also interested in what this community has to share on favorites!

Hi Charles,

A risk of historical bias in training data is largely based on the likelihood of harm to individuals. We see laws and litigation focus on bias in training data primarily in scenarios where AI makes decisions on biased data that has discriminatory and significant impact on people, in the form of affecting the cost or terms of one's insurance, provision of healthcare, education, employment opportunities, access to public services etc. 

If the risk of automated decisionmaking on an individual is low or nonexistent, such as categorizing donors or recommending ask amounts, we don't see orgs correcting for historical bias. 

If your organization is thinking about using AI to make significant decisions about people (grants, financial aid, scholarships, compensating fundraisers, school applications), the answer would be different.