AI in nonprofits: The good, the bad and the ugly
What are the latest nonprofit experiences of AI?
While adoption is still uneven, some nonprofits have had positive experiences, including efficiency savings and encouraging organisational conversations about data. But many also raise serious concerns about AI, including unreliable (but overconfident) outputs, damage to trust, and operational risks like data breaches.
Nonprofit Datafolk Club discussion: AI – the good, the bad and the ugly
Each month our Nonprofit Datafolk Club workshops provide a virtual space for nonprofit people to get together and talk about data-related topics. In October, we gathered to talk about AI in a workshop entitled ‘AI – The good, the bad and the ugly’.
Participants were asked five key questions:
Where has AI already made a positive difference in your role, your organisation, and wider field of work (e.g. health, education, environment)?
Where have you seen AI overpromise and underdeliver?
How is AI changing trust – in information, in institutions, in each other?
Where do you see AI being used to manipulate, exploit, or deceive already?
What's the worst-case scenario you worry about and how likely do you think it is?
We used a closed system AI (Claude) to create a first summary of the anonymised data, which we then reviewed and refined, particularly as we've observed the tendency of large language models to omit negative comments about themselves.
Here is a summary of what people said:
Where has AI already made a positive difference in your role, your organisation, and wider field of work (for example health, education or environment)?
Our Nonprofit Datafolk Club fed back that AI adoption is still uneven. Many organisations lack clear AI policies, approved tools and guidance – particularly around sensitive data, security and free platforms. Cost and reliability are also concerns, especially for qualitative analysis. People also highlight that differing views between enthusiastic and sceptical colleagues can create tension, although growing senior leadership interest offers an opportunity to develop a more strategic and secure approach. It is also important to distinguish AI from established automation and data-analysis tools. Some of the ways people speak about AI making a positive difference include:
Helps to raise conversations about data to encourage better data governance, develop data models and responsible technology use.
Saves time and improves efficiency, particularly for managing emails, administration, prioritising task and workplans, producing notes of (non-sensitive) meetings.
Summarising and organising information, including long papers and multiple documents, extracting actions and key messages.
Supporting writing and communication through drafting, reframing, professional phrasing, audience-specific content, fundraising pitches and idea generation.
Assisting with data and technical work, including dataset matching, dashboards, quantitative analysis, data cleaning, visualisation and generating code for Excel, Power Query and SQL.
Strengthening planning and decision-making by providing templates, restructuring project plans, and organising thoughts.
Where have you seen AI overpromise and underdeliver?
Data folk in nonprofits view AI as most useful as a starting point, sense check or support tool, rather than a dependable replacement for human judgement. Its benefits are often reduced by the time needed to verify outputs, correct errors and supply repeated prompts. This makes appropriate oversight, specialist tools and realistic expectations essential. Some of the ways our Nonprofit Datafolk Club members feel it underdelivers on promises include:
Unreliable and overconfident outputs can include plausible but incorrect information, fabricated references, inaccurate numbers and conclusions that require human checking.
Weak performance on complex tasks means AI may overlook parts of a brief, lose earlier instructions, overcomplicate code or processes, and need large projects broken into guided steps.
Inaccurate transcription and summarisation can misrepresent discussions, assign actions to people who were not present and fail to capture decisions reliably.
Bias and misplaced trust arise when friendly interfaces and confident language create a false sense of trust, including in recruitment and selection.
Limited creativity and quality are evident when images, session plans and suggested lines of enquiry are incomplete, irrelevant or generic, sometimes taking longer to correct than creating the work from scratch.
Human and organisational risks include becoming over-reliant on it, reduced original thought, impersonal services, sustainability concerns, subscription confusion and poorly coordinated internal roll-outs.
How is AI changing trust – in information, in institutions, in each other?
AI appears to be causing a shift, from trust being an assumption, to it being something that must be actively earned through verification and human judgement, honesty and transparency of use. While some of the Nonprofit Datafolk Club have seen little visible change so far, the wider concern is that unchecked or undisclosed use could damage confidence in information, colleagues and institutions. Things people highlight when it comes to how AI is changing trust include:
Trust depends on purpose and expertise: outputs are easier to verify in areas such as coding or within a user’s own field, but harder to assess outside existing knowledge.
AI is increasing suspicion of information, hidden AI use and uncertainty about authenticity make people more cautious about what they read and see.
Trust between people can be weakened when AI-generated responses appear impersonal, errors are not checked, or individual knowledge is devalued.
Education faces particular challenges in balancing legitimate AI use with evidence of students’ own learning, contributing to reduced trust in submitted work.
Institutional transparency and oversight are essential, especially around sensitive data, bias, external communications and reputation, environmental impact and disclosure of how AI is being used.
AI may displace established sources of authority while fast access to information encourages over-reliance, less checking and possible disengagement if people feel they can no longer believe anything.
Where do you see AI being used to manipulate, exploit, or deceive already?
AI is seen as enabling deception at both individual and systemic levels: it can obscure authorship, amplify false content, exploit data and automate persuasion at scale. The responses highlight the need for transparency, informed consent, human accountability and protection for creative professional's expertise and copyright. For data folk in nonprofits, concerns include:
Deceptive confidence and persuasive design can present uncertain or biased information as fact and use personable, frictionless interactions to encourage continued use.
Hidden or unclear organisational use makes it difficult to know when AI has shaped communications, recruitment or services, creating a need for proportionate openness and disclosure.
Scams and fake content are growing through deceptive advertising, social media content, impersonation and cybercrime, with potential to influence behaviour and democratic choices.
Misrepresentation in education and recruitment includes AI-generated assessments, applications and cover letters, as well as tools used to disguise previous AI involvement.
Exploitation of personal data is a concern where people’s information and interactions may be incorporated into AI models without clear knowledge or permission.
Loss of human and creative value can occur when chatbots or digital replicas replace knowledgeable people, users cannot tell whether contact is genuine, and creative work is copied or used without consent.
What’s the worst-case scenario you worry about and how likely do you think it is?
For our Nonprofit Datafolk Club, the concerns range from immediate and plausible operational risks – such as data breaches, unreliable decisions, reputational damage and workforce impacts – to much more speculative systemic or existential scenarios. Things like:
AI-enabled data breaches causing major disruption essential resources and services such as banking and cloud infrastructure.
Compounding misinformation could occur as inaccurate material is passed between AI systems leading to wrong decision making, reputational damage, a loss of funding (particularly detrimental for smaller organisations).
Fundamental errors and weak business reliability remain concerns where inconsistent tools make mistakes that people may not make or create extra work elsewhere rather than improving efficiency.
Job displacement and reduced human contact
Loss of critical thinking and capability may emerge if new employees and wider society become dependent on AI without learning how to work, reason or solve problems independently.
Existential threats range from sentient systems acting against human interests to loss of drinking water and catastrophic harm to humanity.
What’s your funniest idea for an AI-related t-shirt?
To cheer things up a bit we then launched a 'funniest AI-related T-shirt slogan' competition – enter before 30 November 2026!
Finding out more about AI for nonprofits
More recently, we’ve been attending sessions at DigiLeaders AI Week 2026, an online event bringing together expert speakers to deliver more than 200 free talks. The topics covered included AI for good, bias and ethics, AI for charities and public services, AI and data, and more.
All recordings are now available on their website for free. Here are a few we particularly recommend:
Not Faster, More Human: What AI Should Really Do in Frontline Advice from Stuart Pearson at Citizens Advice
Artificial Intelligence- the Good, the Bad and the Ugl”ai” from Kirsty Ingleson at Leeds College of Building
From "We Don't Know Where to Start" to Six Months of Delivery: A Charity's AI Roadmap from Blood Cancer UK
Data Orchard’s position on AI
Our experience in the sessions at AI Week reinforced our position that data maturity is the fundamental basis for innovation.
Many of the speakers emphasised that the basics were critical to success in trialling and implementing AI solutions: ensuring that your data is well structured and managed, your staff are equipped with the skills they need, leadership understand and invest sufficient resources into data, and there is a positive data culture in your organisation.
You can find more of our favourite introductory guides to AI in the nonprofit sector and practical AI resources for nonprofits in our previous blogs on this subject.
And if you’re looking to start your journey to becoming AI-ready, we can help – do get in touch.
Join the Nonprofit Datafolk Club
The Nonprofit Datafolk Club is a friendly group of data-interested people working in or with nonprofits. If you found this resource interesting, or if you have any curiosity in nonprofit data more generally, please come and join us at our next workshop. Each month has a different topic, and you’ll be able to find the details on our events page.
