A few weeks ago, we hosted the latest in our Social Impact Champions webinar series – a series where we showcase and celebrate specialists who are doing brilliant work in support of the third sector. This session featured Jack O’Connell from EBM, a consultancy that works almost exclusively with charities and nonprofits on AI and automation. Jack walked us through some of the real-world work EBM are doing with charity clients – from large national organisations through to small teams with limited resource – and shared some really practical advice for anyone trying to figure out where AI fits in their organisation. You can watch the full recording here.
Here are the key things that came out of it.
Most charities are already using AI – just not at an organisational level
Jack opened by asking attendees what AI tools they were already using day-to-day. The answers – ChatGPT, Copilot, Gemini, mostly for writing and research – were pretty much what he expected. His point was that while individual use of AI tools is now widespread, there’s a gap between that and using AI to solve challenges at an organisational level. That’s where EBM focus most of their work: not just helping individuals be more productive, but identifying where AI can free up team capacity, automate repetitive processes, or fundamentally change how a service is delivered.
Case study: Kinship – turning a 150-page PDF into a conversational tool
Kinship is a charity that supports kinship carers – grandparents, close friends and family members who step in to care for children when parents aren’t able to. They produce a comprehensive care guide, but at 150 pages, carers were struggling to find the right information quickly. Many were turning to ChatGPT instead – which gave coherent-sounding answers but couldn’t be trusted to be accurate or up to date.
EBM built a custom AI tool trained specifically on the Kinship care guide, creating a ChatGPT-like experience but with the knowledge base locked to that one document. The tool handles complex queries, pulls relevant sections from across the guide, applies Kinship’s tone of voice automatically, and flags safeguarding concerns to the appropriate team member. As an unexpected bonus, running the guide through the AI process actually helped identify sections that didn’t conform to Kinship’s own style guidelines(!)
The tool has so far been rolled out to Kinship’s advice team, who were being overwhelmed with enquiries. Feedback has been overwhelmingly positive – including from a self-described “AI sceptic” on the team who said the quality of information matched what was on the website and that they’d happily recommend it. The next phase is putting it directly into the hands of carers themselves.
Case study: Turn to Us – handling thousands of conversations a day
Turn to Us are a benefits charity whose online tools – a benefits calculator and a grant search tool – were seeing enormous traffic spikes during the cost of living crisis. Their team simply couldn’t keep up with the volume of enquiries, particularly out of hours, and they were paying significantly for a third-party out-of-hours human support service.
EBM have built several AI solutions for them over time. The first helped users navigate the benefits calculator itself, which involves complex financial questions that not everyone finds straightforward. The second used generative AI to improve their grant search – moving away from basic keyword matching to understanding the context of a user’s situation, so people are only shown grants they’re actually eligible for. The third is a conversational site search tool that blends traditional search with an AI assistant experience, similar to what Google are now doing with their AI summaries.
Across those three tools, Turn to Us now handles over 30,000 conversations or requests per month through AI – and was able to turn off its costly out-of-hours human support entirely, because the AI was accurate and reliable enough to handle those queries. Human live chat is still available during working hours for queries that genuinely need it.
Case study: Northwest Air Ambulance – a low-cost starting point
Not every AI solution needs to be complex. Northwest Air Ambulance had a single team member responsible for handling all supporter enquiries – on top of their other responsibilities. Out-of-hours enquiries were piling up overnight as emails, because there was no way to respond to them in real time.
EBM implemented Chat2Impact, their low-cost AI assistant tool built specifically for charities. It’s now handling around 300 conversations a month, and the volume of overnight emails waiting in the inbox has dropped from 10-15 down to one or two. The tool can update CRM records, sign people up to newsletters, and take donations directly from the chat interface. When something changes in the organisation – like a Christmas card campaign that ran into problems – new content can be added to the assistant in around 30 seconds.
What is Chat2Impact?
Jack did a live demo of Chat2Impact during the webinar. It’s essentially a safe, low-cost first step into AI assistance for charity websites. Unlike tools like ChatGPT, it doesn’t generate answers on the fly – all responses are written and curated by the team, which means there’s no risk of hallucination or inaccurate information. The AI is used to help speed up setup (it can scan a webpage and generate question-and-answer pairs automatically) and to match user questions to the right pre-written answers, even when the wording isn’t an exact match.
Jack is offering a three-month free trial of Chat2Impact – if that’s of interest, you can get in touch with him directly via EBM.
How to identify AI opportunities in your organisation
Jack shared a practical approach EBM use when starting work with a new charity client. It starts with a simple internal survey – asking teams what AI tools they’re already using, what they’d like to use, and where they think AI or automation could make the biggest difference. The answers are often surprising, and the best ideas frequently come from within the team rather than from the top.
From there, the approach is to map potential use cases against an impact versus effort grid – prioritising things that are high impact and relatively low effort, and deprioritising anything that would cost a lot for limited return. Use cases tend to fall into two broad categories: internal efficiency (content creation, reporting, email drafting, meeting summaries) and external or service-facing tools (assistants, search, automated workflows).
A few practical principles Jack kept coming back to
- Phase your rollout. Don’t launch AI tools straight to end users without testing internally first. Every case study Jack shared followed a phased approach, and it made a real difference to the quality of the final product.
- Guard rails matter. Whether that’s locking a model to a specific knowledge base, building in safeguarding alerts, or curating all answers manually as Chat2Impact does – think carefully about how you’re protecting against inaccurate or harmful outputs.
- AI isn’t about replacing people. Jack was clear on this throughout. The goal is to free up human colleagues to focus on higher-value work, not to remove them from the picture.
- Start small. Low-stakes internal tasks are a good place to build confidence and familiarity before moving to anything more complex or public-facing.
- Think about policies. Many charity staff are keen to explore AI tools but aren’t sure what they’re allowed to use or what information is safe to share. Having even basic AI guidance in place makes a big difference.
Tools worth looking at
Jack also shared a handful of tools EBM rate highly. For productivity and research, he highlighted NotebookLM (Google’s deep research tool) and N8N, a workflow automation tool with lots of pre-built integrations. For writing, he recommended trying Claude alongside ChatGPT – noting that it’s generally considered stronger for copy and has a slightly more considered approach to AI development from its maker, Anthropic. For learning and inspiration, he pointed to AI Recipes for Charities and AI for Non-Techies as good starting points. And for AI-powered fundraising intelligence, he flagged a company called Daro as worth looking at.
If you’d like to watch the session back in full, the recording is available here. And if you’d like to talk through what any of this might mean for your organisation, feel free to get in touch with me at All Things Equal.