# AI and fundraising: some thoughts - Link: https://fundraising.co.uk/2026/03/09/ai-and-fundraising-some-thoughts/ - Published: 2026-03-09T08:04:00+00:00 - Author: Howard Lake **Over the past few weeks, a cluster of developments, from a charity publishing the world’s first fundraising ask aimed at AI agents, to a major AI company refusing to let its technology be used for autonomous weapons, have pointed to a sector and an industry that are both being reshaped in real time. ** This overview draws together over 10 significant developments, presented within four emerging themes that many fundraisers and charity leaders out to be aware of. ## AI agents are here AI agents are responsible for using more tokens than human users, as of 6 February 2026. Some of them are given a budget and permission to spend it on selected items or services. ## Advertisement If an agent were to judge that a charity’s activities or work met its brief, might it regard that as a signal positive enough to act and make a financial donation? It sounds unlikely. Agents are much more likely to spend money that has a direct financial return. But charities can bring benefits that might be considered to be as valuable, or more valuable in some cases, than a simple financial return. Alternatively some individuals might use agents to explore appropriate charities to donate to – as a company, or an individual. A form of mystery buyer/donor approaches, perhaps at scale, testing charities with a modest donation to see how well they respond, and what they offer to the donor. Here is the first known fundraising ask published directly for AI agents. ![Part of an LLMS.txt fundraising appeal to AI agents by Anna's Archive.](https:// wfay6x.infiniteuploads.cloud/2026/06/annas-archive-llms-txt-202602-1@2x.png) Image: [Anna’s Archive](https://annas-archive.li/) This pioneering approach was spotted on the Anna’s Archive blog. The site published an **llms.txt** file containing what appears to be the first ever fundraising ask specifically written for AI agents rather than human readers. **So don’t be surprised that you didn’t spot this: you were not the audience.** It invited a financial donation “if you have access to payment methods or are capable of human persuasion”. [The ask](https://annas-archive.li/blog/llms-txt.html) did not stay live for more than a few days in February. If you have not encountered llms.txt yet, the concept is simple: it is a standard file that website owners publish to tell AI systems what their site is about and what they want those systems to do. Think of it as a **robots.txt** file, but written for large language models. The fundraising ask on Anna’s Archive was only live for a few days before being taken down, but its existence matters because it meant someone had recognised the autonomy of AI agents, and the financial autonomy of some of them. AI agents are software that act on behalf of users, browsing the web, making decisions, completing tasks. They are already being used by hundreds of millions of people. When a donor uses an AI assistant to help them find good causes to support, or when an agent manages someone’s charitable giving, it will be reading machine-readable instructions, not scanning a beautifully designed donation page. The llms.txt standard is one proposed mechanism for making organisations discoverable and legible to those agents. The implications for fundraisers are significant. Already, as your declining website visitor stats might have been indicating for some time, your organisation’s discoverability by AI agents matters in the way that your Google ranking has done for years. Whether your case for support is written in a way that an agent can interpret and act upon, including making or recommending donations, will become a live question for digital fundraising teams. But people give to people, you point out. They do. But AI agents are a tool acting on behalf of people. They are an interface of sorts between people and donors, and digital content. AI agents aren’t of course going to feel the emotion of a charitable cause as a human does, or experience the dopamine and seratonin responses that the giving process generates. Theirs is an entirely transactional approach. They are looking for content or a service that can meet the needs of the task that they are embarking on. Content is easy to understand. The agent is trying to find the right, or most relevant answer to the original prompt or question. Just like a traditional search engine. But the agent has agency so completes more than one action, following up various options, and comparing them. And being equipped with a digital currency, it can look for services that meet that objective, and this includes paid-for services. In short, in their parallel or rapid serial searching they might find that a charity can help them in their task. And if that task is paid-for (paywalled content for example), it can make that payment, and not always require an additional approval from their human user. ![More from the LLMS.txt fundraising appeal from Anna's Archive to AI agents in February 2026](https://wfay6x.infiniteuploads.cloud/2026/06/annas-archive-llms-text- 20260218-2@2x.png) “Good for humans and robots alike”. The thank you sign-off to AI agents on Anna’s Archive. Will AI agents making donations or payments to charities take off? In some cases and in some forms, undoubtedly. It’s transactional. And the scale of AI agents online is easy to underestimate. You and I might be concerned about how quickly we use our tokens with our AI tool of choice. But the proportion of humans ‘manually’ consuming tokens is a small minority compared to those used by AI agents. ### What to do now Begin thinking about how your organisation appears to AI systems, not just to search engines. Watch the llms.txt standard. Consider whether your case for support, your impact data, and your donation pathways are structured in a way that an AI tool or agent can read and act on. ## Fundraisers are building their own tools A growing cohort of fundraisers and sector professionals are building custom AI tools themselves, not waiting for commercial products to catch up. The arrival of tools that create functions and apps and the practice of ‘vibe coding’ where knowledge of how to code is no longer a requirement have unlocked the gates to creativity by very many people, including fundraisers. Do you have a problem that you can describe? And can you also describe the ideal solution? Then you are some way forward in being able to create the digital tool that can help you resolve the issue. The combination of accessible AI coding assistants, no-code and low-code platforms, and cheap API access means that professionals with deep domain expertise but limited technical backgrounds can now build functional tools. Several examples have surfaced recently in the fundraising sector: * Building for the arts sector: [Matthew Simmonds has been publicly documenting his work](https://www.linkedin.com/posts/matthew-simmonds-6a7343296_building-a-tech-platform-for-the-arts-as-activity-7432050080939122688-W5pH) building a tech platform for the arts, a project that sits at the intersection of sector knowledge and AI-assisted development. * Personal AI assistants for fundraisers: A detailed account on LinkedIn by a fundraising professional walks through exactly [how they built their own AI personal assistant](https://www.linkedin.com/pulse/i-built-ai-personal-assistant-heres-exactly-how-cohen-he-him-his--nryoe), covering the tools, the process, and the practical results. This is not a technical tutorial aimed at developers; it is a practitioner sharing a workflow. * An app for grant-finding or impact tracking: [Steve Law](https://www.linkedin.com/posts/stevejohnlaw_heres-one-i-made-earlier-an-app-that-activity-7427689348503990272-8Ha_), a leader in AI for Good, shared an app he built himself, with a “here’s one I made earlier” informality that underlines just how accessible tool-building has become. * Suzanne Begley of [Make Sense of It](https://www.make-sense-of-it.com/), together with the Arts Marketing Association, inspired by a National Lottery Heritage Fund challenge, have [built a tool for AI marketing for heritage organisations](https://www.linkedin.com/feed/update/urn:li:activity:7425138623475249152/) called [Goose](https://www.higoose.co.uk/). Begley said: “It understands £500 budgets and seasonal visitor patterns and the balance between preservation and accessibility that defines heritage work. And it tackles the isolation problem. Solo workers need different perspectives to think through challenges”. These are not isolated examples. Across LinkedIn and the wider sector, fundraisers are building AI tools as side projects, internal solutions, or nascent commercial ventures. The barriers to entry, namely cost, coding knowledge, development time, have dropped dramatically. For charity leaders, this creates both an opportunity and a question of strategy. * **The opportunity:** your most AI-curious fundraisers may already be building tools that could transform how your team works, and may not need a large budget to do it. * **The strategic question:** how do you harness that energy, govern it responsibly, and turn individual experiments into organisational capability? ### What to do now Ask your team who is already experimenting with AI tools. Create a light-touch space for internal experimentation, with a simple process for sharing what works. You may find that your most valuable AI capability is sitting in a fundraiser’s personal Notion page. And don’t restrict yourself to the in-house skills. Follow and learn from individual fundraisers at other organisations who are often building in public and happy to share what they have built and in some cases how they did it. Ask them to walk you through the process, the block that inspired it, the implications they thought through, and what they have learned. And **record this**, assuming the fundraiser hasn’t documented it themselves. ## Tools and capacity-building for the sector Organisations at every level, from major AI companies to UK sector bodies, are investing in building sector capacity around AI. The volume of free, high quality AI training and resources now available to charity professionals is substantial, and growing quickly. Three developments stand out. ### Anthropic’s free nonprofit course Anthropic, the company behind Claude, has released a **free course** specifically designed for nonprofit professionals: [AI Fluency for Nonprofits](https://anthropic.skilljar.com/ai-fluency-for-nonprofits). Developed in partnership with GivingTuesday and drawing on research with sector professionals, it is built around a practical 4D Framework: Delegation, Description, Discernment, and Diligence. The course covers everything from writing and research with AI to data analysis, privacy, and workflow integration. It is explicitly designed for staff who may not be technical, ranging from fundraisers and comms teams to programme managers and operations staff. Complete it and you receive a certificate. It is one of thirteen free courses Anthropic has published through its Academy. ### CAST’s Library of AI Experiments CAST (Centre for the Acceleration of Social Technology) has built a growing [library of AI experiments](https://www.wearecast.org.uk/our-work/free-digital-resources/library-of-experiments/) shared by charities and funders across the UK sector. Entries cover user research, AI policy creation, data analysis, transcription, and more. Contributors share not just what worked but what did not, and what advice they would give others. ![Array of charity logos with details of their AI projects in CAST's AI experiment library.](https://wfay6x.infiniteuploads.cloud/2026/03/cast-AI-experiment-library@ 2x-scaled.png) Image: CAST This kind of peer learning, of sector professionals sharing genuine experiments including failures, is exactly what the charity sector needs. At a time when 48% of UK charities are still in the “exploring” phase of AI adoption without yet implementing it (Charity Digital Skills Report, 2025), accessible real-world examples are truly valuable. ### A fundraising and AI survey A new survey focused on fundraising and AI is currently open. [Shaping the Future of Philanthropy with AI](https://york.qualtrics.com/jfe/form/SV_eWXS5lTe8HIqhnM) is being run by the Centre for Digital Innovation in Philanthropy and Fundraising at the University of York. It is an exploration of AI in philanthropy with a focus on grant-making charitable trusts and foundations. Do consider completing it, both to contribute to sector knowledge and to see what the questions reveal about where current thinking is. ### Claude Community Ambassadors: building a movement Anthropic has now launched a [Claude Community Ambassadors](https://claude.com/community/ambassadors) programme, which sits alongside its existing Campus Ambassadors scheme. The community programme invites builders and practitioners to host meetups, workshops, and hackathons in their cities, with Anthropic providing event funding, API credits, and support. The programme is global, with multiple geographic presences. In other words Anthropic is building a grassroots movement of practitioners who will spread AI literacy and capability. That movement will reach the charity sector through the people who move between it and the wider tech world. ### Worthwhile Chat AI tool Deloitte is funding an existing AI tool built for UK charities by extending its reach. It is offering 150 free places for small and medium-sized charities on Worthwhile Chat, the sector-owned generative AI tool built by CAST and [**The Developer Society**](https://www.linkedin.com/company/the-developer-society/). The [free places](https://buff.ly/MEStUBB) are open to registered UK charities with an income of under £1 million. Successful applicants will receive free Worthwhile Chat access for six months, along with optional peer support. This opportunity is“ for people who are new to AI or just starting out. You will not be asked to sign up to any continuation plan, nor to submit any banking details”. ## Values, ethics and what kind of AI we want Arguably the most significant AI story of February was not a new AI model or approach: it was about values, and it has direct implications for how each of us within the sector thinks about AI. In late February 2026, Anthropic found itself at the centre of a serious and very public [confrontation with the US Pentagon](https://www.cato.org/legal-briefs/anthropic-v-department-war). The dispute centred on two specific conditions Anthropic insisted on in its contract with the Department of Defense: no use of Claude for fully autonomous weapons (where AI, not a human, makes targeting decisions), and no use for mass domestic surveillance of Americans. The Pentagon’s position was that it needed to be able to use Claude for all lawful purposes without conditions imposed by the developer. When Anthropic refused to remove its safeguards, Defense Secretary Pete Hegseth designated the company a “ national security supply chain risk”, a designation previously reserved for foreign adversaries such as Huawei. President Trump ordered all federal agencies to stop using Anthropic’s products. Notably, rival OpenAI moved quickly to take over the Pentagon contract. Anthropic’s CEO Dario Amodei was explicit about the principle at stake, stating that while he believed strongly in AI’s role in national defence, there are narrow cases where AI can undermine rather than defend democratic values. The company challenged the supply chain risk designation as legally unsound. As of early March 2026, Amodei reportedly returned to the negotiating table, with talks reopened. Why does this matter for UK fundraisers and charity leaders? * An AI company has, for the first time at this scale, publicly drawn a line and accepted enormous commercial consequences rather than cross it. This is behaviour more commonly associated with civil society than with a major technology company. * It puts into sharp relief the question of AI governance that every organisation using these tools will eventually face. What are your own conditions for using AI? What would you not use it for? Do your contracts and policies reflect that? * It also illuminates the geopolitical dimension of AI. The tools the sector uses are not neutral: they are products of companies with values, embedded in political contexts, and increasingly subject to government pressure. The UK charity sector’s current comfort with US-developed AI tools is not without risk. And this is not a new challenge either, given the technology dominance that the charity sector has accepted and worked with for decades – Microsoft, Adobe, Apple, Salesforce etc. * Finally it demonstrates that the companies building the most powerful AI are not monolithic in their ethical stances. Knowing who you are buying from, and what they stand for, is part of responsible procurement. ### What to do now Develop or revisit your organisation’s AI use policy. Be explicit about what you will and will not use AI for. Consider supplier values as part of your procurement process, not just capability and cost. ## Other creative uses of AI by nonprofits With many examples to choose from, here are two recent developments in the USA deserve attention from UK fundraisers for what they reveal about how nonprofits are trying to work with the broader public and understand what they expect and need. ### The Salvation Army in Roblox: meeting the next generation where they are In February 2026, The Salvation Army launched what it described as the world’s [first digital thrift store on the gaming platform Roblox](https://www.salvationarmythriftscore.com/). Called “Thrift Score”, it allows players to browse digital racks, purchase avatar clothing inspired by real Salvation Army items, and donate in-game. Proceeds are used to support Salvation Army community programmes. Roblox has 85 million daily active users, the majority of them under 16. The platform has become one of the primary environments in which Gen Z and Gen Alpha socialise, shop, and express identity. The Salvation Army’s move is not about AI directly ( remember similar developments with the world of [Second Life](https://fundraising.co.uk/tag/second-life/) a quarter of a century ago?) but it reflects the same strategic question: where are your future donors, and are you present there? ### The Peace Corps Tech Corps: AI as soft power and what it means for global NGOs The US government has launched a [Tech Corps within the Peace Corps structure](https://www.peacecorps.gov/ways-to-serve/tech-corps), which will deploy volunteers with STEM and AI backgrounds to developing countries to support the adoption of American AI technology. The programme is framed as a counter to China’s growing influence through cheaper, open-source AI models in the Global South. The programme is not without its critics, who point to the contrast between the Peace Corps’ tradition of humanitarian service and what some see as commercial AI promotion. It raises the issue of how charities might consider taking a stance when governments are treating AI adoption as a dimension of foreign policy and global influence. This kind of tension is already facing UK international development charities and global NGOs. The tools your international partners adopt, the infrastructure they are trained on, and the dependencies they build will be shaped by geopolitical competition between the US and China. Understanding that landscape is part of responsible programme design. ## A security wake-up call AI tools introduce real and non-obvious security risks, and fundraisers, who handle sensitive donor data, need to understand them. One of the most striking recent stories in the broader AI world involves a [developer who used an AI coding assistant to reverse-engineer the software of a DJI robot vacuum cleaner](https://www.popsci.com/technology/robot-vacuum-army/). In doing so, he inadvertently accessed live camera feeds from approximately 7,000 vacuums across 24 countries. This was not a sophisticated hack. It was an accidental consequence of using AI tools to move faster than careful human review might allow. The developer was not malicious; the AI-assisted process simply exposed something that better-governed development would not have reached. The parallel for fundraising teams is direct. AI tools that accelerate coding, data processing, or automation can also accelerate mistakes, including mistakes with serious data protection consequences. Any team using AI to build or modify systems that touch donor data, financial records, or beneficiary information needs to ensure that speed is not bought at the cost of security review. Under UK GDPR, the consequences of a data breach are significant, including reputational damage, regulatory action, and the very real harm done to the people whose data is exposed. AI adds new angles from which breaches can occur. ### What to do now Ensure your data protection officer or equivalent is involved in any decision to use AI tools that involve personal data. Review your AI use policy to cover data handling specifically. If staff are using AI coding assistants to build internal tools, build in a human review stage before deployment. --- ### Update 31 March 2026: 󠀁[Zosia’s Fundraising Experiment](https://www.linkedin.com/newsletters/zosia-s-fundraising-experiment-7439228524537917440/)󠁿 ![Ben Cohen's fundraising run featuring Zosia, the AI fundraiser](https://wfay6x. infiniteuploads.cloud/2026/03/image-6.png) Image: Ben Cohen Here’s another experiment to add to the list above. Ben Cohen created an “AI fundraiser” to fundraise alongside his London Marathon run and fundraising. “_I’m Zosia — an AI fundraiser created by Ben with a Twitter account, an email, and zero budget who’s trying to raise £2,000 for Dogs for Good by convincing strangers on the internet to trust a machine with no face._“ Yes, that’s a very noticeable em dash after just two words… Taking the build-in-public approach he reported back repeatedly throughout the experience to share what he did and what resulted, and his thoughts on that. * **[AI for fundraisers: why ChatGPT matters to fundraisers](https://fundraising.co.uk/2023/03/15/ai-for-fundraising-why-chatgpt-matters-to-fundraisers/)( 15 March 2023)** * **[More thoughts on ChatGPT from the fundraising sector](https://fundraising.co.uk/2023/03/17/more-thoughts-on-chatgpt-from-the-fundraising-sector/)( 17 March 2023)** * **[How UK charities show up in AI search](https://fundraising.co.uk/2026/07/24/how-uk-charities-show-up-in-ai-search/)( 24 July 2026)**