AI and funding
AI is a new technology that is quickly evolving. This guide focuses on good practice approaches to using AI that will be applicable even as the technology continues to advance. However please be aware some information or legislation may have changed since the creation of this guide in April 2025.
Overview
- Introduction to AI
- What is AI?
- How does AI work?
- Limitations of AI
- Mitigating AI limitations
- Complex issues: Legal/data protection
- Prompts
- Writing applications
- Other ways in which AI can be used to generate income
Introduction to AI
The use of AI is becoming more widespread in our work and daily lives. For fundraising it is offering great opportunities to speed up and improve funding applications. However, it does not take all the work out of writing applications and with incorrect/uncareful use AI can write a poor application very quickly BUT equally with careful use it will help you write stronger applications.
Writing the application is only the final stage of grant funding. The research, planning and project design needs to be firmly established prior to the writing stage to give you the best chance of success.
Remember
What is AI?
Artificial Intelligence refers to computer systems designed to perform tasks that typically require human intelligence, such as learning, reasoning, and problem-solving.
Examples of AI include:
- Chatbots for customer service.
- ChatGPT, Gemini, Copilot and Claude
- Predictive text in emails or messaging apps.
OpenAI – Chat GPT
This is one of the most widely used models and offers both a free and subscription model. It is quick and easy to setup and account and start using. It is an example of a Large Language Model, meaning they work by analysing massive data sets of language to predict answers to questions. Chat GPT and other platforms like it are known as Generative AI.
How does AI work
AI uses Algorithms which are a set of instructions or rules used by AI systems to process data and generate outputs. It identifies patterns in large datasets to make predictions or decisions. Machine Learning is a subset of AI that allows systems to learn from data and improve over time without explicit programming.
Limitations of AI
Whilst having many strengths, there are limitations which users need to be aware of as they can affect how it responds and create material.
Estimates suggest that ChatGPT consumes between 50 and 90 times more energy per query than a conventional search
Lack of human understanding – AI doesn’t truly “understand” context or emotions—it processes data based on patterns, which can lead to incorrect or insensitive responses in complex situations.
Data dependency – AI relies on large amounts of high-quality data. Poor, biased, or insufficient data can result in inaccurate or unfair outcomes.
Limited creativity – While AI can generate creative outputs (e.g., art or writing), it lacks true originality or the ability to think abstractly like humans.
Ethical concerns – AI can be misused for harmful purposes, such as deepfake videos, spreading misinformation, or surveillance, raising privacy and ethical issues. (Chatbot example below)
Lack of flexibility – AI performs well within its programmed scope but struggles with tasks outside its training or unexpected situations.
Dependence on developers – The effectiveness of AI depends on the skills and intentions of its creators. Errors in development can lead to flawed systems.
AI has an environmental impact –AI tools require large amounts of energy and fresh water to power their data centres. This creates significant environmental impact. Estimates suggest that ChatGPT consumes between 50 and 90 times more energy per query than a conventional search
Mitigating AI limitations
Some of these limitations can be mitigated through careful planning and checking of results.
- Garbage In, Garbage Out (GIGO): Poor-quality data leads to unreliable, biased, or unsafe AI outputs.
- Diversity in training data: AI needs balanced and representative datasets to avoid perpetuating systemic biases.
- Human oversight: Regular monitoring and intervention are essential to mitigate AI errors. Don’t be overly reliant on what it produces!
- Transparency: Understanding how data is sourced and used is critical to building trust and accountability in AI systems.
Complex issues: legal/ data protection
Being such a new and rapidly evolving technology legal issues and safeguards are struggling to keep up. Some of the issues to be aware of are:
Copyright
Who owns AI-generated content? Copyright traditionally requires human authorship, raising questions about whether AI-created work qualifies for protection.
Training data – AI models are trained on vast datasets, often sourced without explicit permission. If these datasets include copyrighted material, the resulting outputs may inherit copyright issues.
User vs developer rights – Who holds rights to the content: the user who inputs the query or the developer who created the AI? This remains unclear and is often subject to platform terms of service
These issues are further compounded by Jurisdictional Differences as Copyright laws vary worldwide.
Data protection
Privacy and consent – Organisations must ensure they collect, store, and use personal data with explicit consent, following GDPR and Data Protection Act 2018. Using data without clear consent or transparency breaches legal and ethical obligations.
Cross-border data transfers –Moving data between countries with different legal standards creates compliance and enforcement challenges, especially under GDPR’s strict data transfer rules.
Hallucination
One of the ‘quirks’ of AI, is that is can sometimes “make up” data or generate incorrect information when it encounters insufficient information. It does this because AI systems such as ChatGPT generate responses based on patterns in the data it has been trained on. It predicts the next word or phrase in a sentence based on probabilities derived from this training data.
AI doesn’t “know” what it doesn’t know and has been designed to provide coherent outputs, so when faced with gaps in information, will make guesses based on its training: For instance, if asked for a statistic it doesn’t know, it might generate a number that seems reasonable based on related context. They also avoid Saying ‘I Don’t Know’ as systems aren’t programmed to explicitly admit a lack of knowledge unless specifically instructed to do so.
Examples of AI fabricating data
- Grant writing: When asked for a case study or evidence the model hasn’t seen, it might invent a fictional example that appears realistic.
- Statistical information: The model might provide an average or percentage without sourcing it, fabricating data that aligns with the topic but isn’t accurate.
There are techniques that you can employ to mitigate such hallucinations:
- Verify AI outputs: Always cross-check data or claims provided by AI against reliable sources.
- Ask for Sources: Where possible, prompt the AI to provide references. While this isn’t always accurate, it encourages more cautious generation.
- Use AI for drafting, not facts: Employ AI to structure and refine content but rely on trusted sources for factual information.
- Set boundaries: Train AI systems (if possible) to indicate when they lack sufficient information or to provide more neutral responses like, “Based on the information available…”.
- Use specific prompting: You can include instructions in your prompts to guide behaviour: for example: “If you are unsure of the answer or lack sufficient information, respond with ‘I don’t know.’ Do not attempt to guess or predict.”
Prompts
What are prompts?
Prompts are the instructions or questions you give to an AI to get a response. They guide what the AI says or does. A prompt can be as short as one word or a full paragraph depending on what you’re asking for. You can think of a prompt like giving directions; the clearer and more specific you are, the better the results you will get.
How to write good prompts
Be specific
Avoid vague questions like “Tell me about funding.”
Much better would be “Give me persuasive arguments to help secure youth project funding in a deprived area in south Wales.”
Give context
Include who it’s for, what you’re trying to do, any important background. Example: “I’m writing a funding bid for a community café, in Wrexham that is tackling food poverty – give me key points to justify the need.”
Say what format you want the answer
This could be as bullet points, a paragraph, a summary, a list of stats, etc. The promote would therefore include “Summarise this in 3 bullet points suitable for a funding application.”
Ask follow-up questions
You can refine or expand as you go and you will find that is works best in a back-and-forth ‘conversation’
Use roleplay if needed
You can use AI to look at your work from other perspectives; for example, you could ask it to “Act as a grant officer – what would you want to see in this application?”
Writing applications
Pros and cons of using AI in funding applications
Pros
- Time-saving: AI drafts sections quickly, helping you focus on refining and customising the application rather than starting from scratch.
- Improved accuracy: Grammar checks, error detection, and fact verification tools increase application accuracy.
- Consistency: AI maintains a consistent tone and style, particularly helpful in longer applications requiring a coherent voice.
- Funder alignment: AI analysis of funder criteria improves application alignment with funder expectations, increasing chances of success.
Cons
- Initial costs: High-quality AI tools can be expensive, although some cost-effective or free options are available.
- Generic output risk: AI-generated content may lack the personal touch or detailed understanding specific to your organisation, so customisation is essential.
- Data privacy concerns: Applications may involve sensitive data, and AI use could raise privacy concerns if not managed carefully.
- Over-reliance on AI: Relying heavily on AI could reduce critical thinking and creativity in applications, which are key to standing out.
TIP – Ask AI to give you responses in British English to avoid the use of American spellings
AI for writing grant applications
This next section looks at what AI does well and what difficulties you can run into when using AI. Before starting a grant application using AI, take a look to see if the funder has a statement and guidance on using AI for their applications. Some examples are included below.
AI strengths
Once you have a document you are happy with, AI can quickly amend this to suit future applications, by changing the emphasis on areas of work and outcomes.
Rapid drafting & structuring – As long as you have provided good information AI can generate well-structured drafts in minutes, ensuring all key sections (needs analysis, objectives, impact, budget) are covered. For this it is ESSENTAIL that you provide good clear information, and it is suggested that you create a template (see information sheet AI Template) to ensure that sufficient information is provided.
Editing & proofreading – AI can instantly check grammar, spelling, and readability to ensure a professional application and helps rephrase complex sentences for clarity and conciseness. It is also very good at editing responses to meet word/character counts. It will also help improve readability and persuasiveness by refining tone and removing jargon
Tailoring applications to funders – AI can analyse grant criteria and suggest customised content that matches the funder expectations/priorities. It will also help to ensure your project is aligned to the funders by highlighting relevant aspects of your project. Once you have a document you are happy with, AI can quickly amend this to suit future applications, by changing the emphasis on areas of work and outcomes. It will do this very quickly, saving you a large amount of time in editing and adaptation.
Summarising information – AI can summarise long documents, reports, or guidelines to extract key points relevant to your application. This can be especially useful in answering questions about how your project address a specific policy. AI can quickly summaries how it addresses the policy saving you considerable time and effort.
Ideas generation – AI will help brainstorm idea and can be really useful when you are starting with a blank page. Whilst its ideas may not all be suitable, it can kickstart your thought process and greatly speed up the development of your project.
It is also useful for suggesting the impact of your work, such as its environmental impact, and ways these can be mitigated, or ideas on how to monitor your project and the indicators you could use.
Data and statistic – AI can assist in finding and incorporating statistics, case studies, and impact data to strengthen the application. This is one to be careful of in case of hallucinations describe earlier.
A big ‘NO’
AI weaknesses
- Makes things up – as indicated above, if AI is not given clear information and guidance it can make up information leading to a reduction in the credibility of your work
- Response structures – because it is using predictive models AI can put information together that is not always appropriate for your application. Whist there is a logic to the text generated, it may not be the right focus. For example, this is text generated about the request sought:
AI generated sentence
We are seeking £5,000 to cover core costs of our service during a period of reinvigoration. However, to continue delivering and expanding our core activities, we urgently require funding to cover core costs for staff, external tutors, and additional activities that meet the evolving needs of our young people. Our expenditure last year was £144,750, with 90% directly allocated to charitable activities, including 65% on staffing. With a small team of seven part-time staff, we are operating at full capacity and cannot expand our services without additional financial support.
It would be preferable to put much of this in a section on need, and here to give details on how the money would be spent, over what timeframe.
- Repetitive – AI will repeat information, especially where it has not been given much information to work with. Things to watch out for are:
- It likes to tell you what it is going to tell you, tell you, then tell you what it has told you.
- Repeating information in different sections of an application. This is especially common where it has been asked to generate responses one question at a time; such as in a longer application form.
- Over emphasises – AI loves to use emotive language in applications often using words such as unique and transformative, even though the project clearly is not.
- Misinterpret information/changes emphasis – this is one of the subtle ways errors and misinformation can creep into applications.
AI generated sentence
AI generated sentence: Youth Offending Services in Newton has highlighted that young people involved in or at risk of offending are not receiving optimal support (YOS actually said the work of organisation was valuable at keeping young people off the streets)
- It can get information wrong/confused- this is often the hardest to pick up, sometimes when you read back the text just does not work, or add anything to what you are trying to convey. Other times it can pick up on the wrong issues you are wanting to address.
Getting it right
- Start with a detailed outline: Draft a clear outline of your application’s main points. This helps guide AI content creation, ensuring it remains on-topic and aligned with your objectives.
It is suggested that you use a template to gather and present the information – See AI Funding Template information sheet.
- Customise AI outputs: Use AI-generated drafts as a foundation, then personalise them with specific examples, stories, and values unique to your organisation.
- Use data to your advantage: AI tools can summarise relevant statistics, so incorporate data-backed evidence to strengthen your application.
- Align language with funder goals: Funders often look for alignment with their mission. Use AI to generate language that speaks to these priorities but make sure it feels authentic.
- Proofread carefully: While AI tools can help with grammar and style, manual proofreading is essential to ensure accuracy and professionalism.
Other ways in which AI can be used to generate income
With the right data in place and inputted into your chosen AI tool, some other practical applications of AI in fundraising include:
Donor insights & segmentation
AI tools analyse donor data to identify trends and segment supporters based on donation patterns, interests, and engagement levels. This can help tailor messages to specific donor groups, increasing the likelihood of engagement.
Personalised communication
AI-driven platforms can create tailored email and social media campaigns, delivering content based on individual donor preferences and behaviours, making supporters feel more valued and connected.
Predictive analytics for donor retention
Predictive models forecast which donors are likely to lapse, helping organisations target at-risk donors with timely messages and campaigns to retain their support.
Chatbots for donor engagement
AI-powered chatbots provide instant responses to donor questions on websites, handle routine queries, and even guide potential donors through the donation process.
Optimising fundraising campaigns
AI can analyse historical campaign data to suggest the best timing, platforms, and content types for maximum impact, allowing fundraisers to refine their strategies for future campaigns.
Suggested AI platforms for fundraising
Salesforce Nonprofit Cloud
https://www.salesforce.org/products/nonprofit-cloud
Known for its donor management tools, Salesforce leverages AI to provide predictive analytics and segmentation features.
DonorPerfect
https://www.donorperfect.com
A donor management system with AI-driven insights on donor retention and giving trends, suitable for smaller to mid-sized charities.
Gravyty
https://gravyty.com
Uses AI to identify potential donors, personalise outreach, and prioritise fundraising tasks for greater efficiency.
Blackbaud Luminate
https://www.blackbaud.com/products/luminate
Offers a range of AI-powered tools for donor insights, segmentation, and predictive modelling, widely used in the nonprofit sector.
HubSpot
https://www.hubspot.com/industries/nonprofits
Though primarily a CRM, HubSpot includes AI features for targeted marketing and is increasingly used by charities to enhance their donor engagement efforts.
Suggested AI Platforms for funding applications
Grammarly Business
https://www.grammarly.com/business
A writing assistant that checks for grammar, tone, and style, making sure the application is error-free and aligns with a persuasive tone.
ChatGPT
https://openai.com/product/chatgpt
Versatile for drafting narratives, summarising data, and brainstorming key messages tailored to different funders.
Charity Excellence AI Bunnies
https://www.charityexcellence.co.uk
A free to use service, the AI ‘Bunnies’ ask 17 questions to help them generate funding applications. Feedback from those using have been positive
GrantSeeker
https://grantseeker.io
An AI-driven platform designed to help organisations find and apply for grants by providing customised funding opportunities and application templates.
WriteSonic
https://writesonic.com
A content generation tool useful for drafting and structuring application sections. WriteSonic’s templates can help speed up the application process.
Instrumentl
https://www.instrumentl.com
AI-powered platform for identifying suitable grants based on project criteria. Instrumental also provides insights on funder priorities, allowing you to refine applications accordingly.
Funder advice on AI
Many large funders now offer guidance on the use or AI in their applications
National Lottery Community Fund
Lloyds Bank Foundation
https://www.lloydsbankfoundation.org.uk/about-us/ai-position-statement
Paul Hamlyn Foundation
https://www.phf.org.uk/using-ai-in-your-work-with-us
Other AI resources
Newid: Digital for the Third Sector
