Artificial intelligence has arrived in the charity world faster than almost anyone expected. Used well, it can translate a message into a family’s own language, route supplies to where they are needed, and flag a suspicious invoice before it is paid. Used carelessly, it can waste money, leak private data and quietly shut out the very people a charity exists to serve.
This is a practical look at AI tools for nonprofits in 2026: where they genuinely help a gift travel further, where they do not, and the guardrails we believe every charity should put in place first.
Where nonprofits stand with AI in 2026
Adoption is no longer the question. In Virtuous’s 2026 Nonprofit AI Adoption Report, based on a survey of 346 nonprofit organisations in December 2025, 92% said they were using AI in some capacity. Most of that use is everyday work: drafting emails, generating ideas, research and brainstorming.
The same report found something more sobering. Only 7% of those organisations reported major improvements in their ability to achieve their mission, and 47% had no AI governance policy at all. In other words, using AI and benefiting from AI are not the same thing.
That gap is where the real opportunity lies. A tool that writes a newsletter faster is pleasant. A tool that helps a field partner deliver water filters to the right village a week sooner is transformative. The difference is not the technology. It is whether the technology is pointed at the mission, and whether anyone is checking the results.
Translation: meeting people in their own language

Language is one of the quietest barriers in humanitarian work. According to UNESCO, 40% of people globally lack access to education in the language they speak and understand fluently. The World Atlas of Languages lists around 7,000 languages, yet only 351 are used as a medium of instruction in schools.
Machine translation and speech tools can help close part of that gap. A health leaflet, a water-safety message or a consent form can be drafted in a regional language in minutes rather than weeks. Voice tools can read information aloud for people who do not read comfortably.
But translation quality drops sharply for languages with little digital text, which are often the very languages spoken by the families with the greatest need. The responsible pattern is simple: let AI produce the first draft, then have a fluent local speaker review it before it is used. When a mistranslated dosage or address could cause harm, a human reviewer is not optional.
Logistics: getting supplies to the right place

Much of what makes aid expensive happens between the warehouse and the doorstep. Stock expires on a shelf in one place while another clinic runs short. Trucks travel half empty. Orders are placed by guesswork because last season’s records sit in a dozen spreadsheets.
This is where simpler forms of AI often shine. Forecasting models can estimate demand from past distributions and seasonal patterns. Route planning can cut fuel and time. Even a well-designed spreadsheet assistant that spots duplicate orders or approaching expiry dates can save real money. None of it is glamorous, and that is the point: the most useful AI in a charity is often invisible to donors.
The caution here is about data. Forecasts are only as good as the records behind them, and records from places facing emergencies are often incomplete. A good team treats every model’s output as a suggestion to be checked against what people on the ground are actually seeing.
Fraud checks and verification
Every charity has a duty to make sure money reaches the work it was given for. AI can help by doing the tedious first pass: matching invoices to receipts, flagging unusual payment patterns, spotting duplicate beneficiary records, or checking that photos of completed work are not reused from elsewhere.
What AI should not do is make the final call. An algorithm that flags a partner as “high risk” can be wrong, and it can be wrong in biased ways, for example by treating unfamiliar names or informal bookkeeping as suspicious. Flags should open a conversation, not close a partnership. Our own approach to checking causes, which puts people and documents ahead of automation, is set out in how we verify causes.
If a tool told you a long-standing local partner looked suspicious, what evidence would you want before you acted on it?
Design for the people who are offline
It is easy to design a brilliant app for people who already have fast internet. It is harder, and more important, to design for those who do not. The International Telecommunication Union’s Facts and Figures 2025 estimates that 74% of the world’s population is online, which still leaves about 2.2 billion people offline. The gap between richer and poorer countries is stark.
Source: ITU, Facts and Figures 2025
For a charity working across regions such as Sub-Saharan Africa, South Asia and Afghanistan, this changes how technology should be built. Tools should work on basic phones, over text messages or voice, with low data use and offline modes that sync later. If an AI service only works for people with a new smartphone and a strong connection, it risks serving the easiest-to-reach rather than those in greatest need. You can see where we hope to help on our Where we help page.
Privacy and ethics: guardrails before gadgets
People who receive help are often in vulnerable moments. Their names, locations, health details and photos deserve the same protection you would want for your own family. Pasting that information into a public AI tool can expose it in ways nobody intended.
Helpful frameworks already exist. The OECD AI Principles, first adopted in 2019 and updated in 2024, set out values such as human rights and fairness, privacy, transparency, robustness and accountability. As Jamila Charity takes shape, these are the commitments we intend to hold ourselves to:
- A written AI policy before wide use. Which tools are allowed, for what, and who is responsible.
- No personal data in public tools. Beneficiary and donor details stay out of systems we do not control.
- Humans make decisions about people. AI can suggest; staff and local partners decide.
- No synthetic images of need. We will not use AI-generated pictures presented as real families or real places.
- Open about where AI is used. If a tool shaped a decision or a report, we will say so.
Would you trust a charity more if it told you exactly where it used AI, and where it chose not to?
What this means for donors
You do not need to be a technologist to judge whether a charity uses AI wisely. A few plain questions go a long way:
- Does the organisation have a published AI or data protection policy?
- Can it name a concrete result AI helped achieve, not just time saved on emails?
- Who checks the output, especially in translation and partner checks?
- How does it reach people who are offline?
The best answer is rarely “we use the latest model”. It is usually something more modest and more honest: a small tool, carefully chosen, measured against the mission, and switched off if it is not helping. How much further could each gift travel if every charity held its tools to that standard?
If you are still deciding how to fit causes like this into your year, our guide to pledging before you give walks through a simple plan, and our programmes page explains the work we are preparing for.
Frequently asked questions
What are the most useful AI tools for nonprofits?
The most useful tools are usually the least flashy: translation with human review, demand forecasting and route planning for supplies, and assistants that check documents and records for errors or duplicates. Writing aids save time too, but they rarely change outcomes for the people a charity serves. Start with the problem you need to solve, then look for the simplest tool that addresses it.
How many nonprofits are using AI in 2026?
In Virtuous’s 2026 Nonprofit AI Adoption Report, 92% of the 346 organisations surveyed said they used AI in some capacity. Only 7% reported major improvements in their ability to achieve their mission, and 47% had no AI governance policy. Adoption is widespread, but meaningful impact is still rare.
Is it safe for charities to put beneficiary data into AI tools?
Only with great care. Personal details about people receiving help should not be entered into public AI services that a charity does not control. Organisations should set a clear data policy, remove identifying details where possible and use tools with strong privacy terms.
Can AI replace human translators in humanitarian work?
Not for anything important. Machine translation is a fast first draft, but quality falls for languages with little digital text, and errors in health or safety information can cause harm. A fluent local speaker should review translated material before it is shared with families.
How can AI help prevent fraud in charities?
AI can scan large volumes of invoices, payments and records to flag duplicates and unusual patterns for a person to review. It works best as an early warning, not a judge. Every flag should be checked by people who understand the local context before any decision is made about a partner.
Does Jamila Charity use AI?
Jamila Charity is still in formation and is not yet registered, so we are setting our policies before our programmes begin. We expect to use AI only where it clearly helps the work, under a written policy, with personal data kept out of public tools and with people making the decisions. We will report openly on where AI is used.
What should a nonprofit AI policy include?
At a minimum it should list approved tools and uses, set rules on personal and sensitive data, require human review for decisions about people, and name who is accountable. Published frameworks such as the OECD AI Principles are a helpful starting point. The policy should be short enough that every staff member and volunteer actually reads it.
From Jamila Charity
See how we plan to earn your trust
Technology should make generosity more accountable, not less. Our transparency page sets out how we intend to check partners, protect people’s data and report openly on where every pledge goes once we begin.








