To AI or Not to AI? Whatever — just keep it human.

Reflections on AI, the third sector, and what a brilliant day in a room full of fearless people taught me.

FROM THE ROOM

I had the pleasure of speaking at ChEW's Festival of Impact and Evaluation this June. Whilst the occasion gave me a chance to introduce more people to the JRF Insight Infrastructure programme, I was asked to share our experience with AI. Before getting into that, though — some things I'm still thinking about from what was, without question, a brilliant event all around.

"Getting together in person takes things to a whole different level. It goes beyond our intrinsically social nature as a species."

In-person events allow us to establish meaningful connections, entertain constructive conversations, and amplify our ability to influence cultures and practices in ways that remote exchanges simply can't replicate.

To ChEW and others working in this space, we need you!

WHAT THE EVENT TOLD ME

The third sector is not afraid. The prohibitive economy hits this space as hard — if not harder — than others. And yet that doesn't stop a rich ecosystem of people and communities from wanting to learn, adapt, and innovate. There's a genuine propensity for knowledge exchange and a willingness to improve not just what we do, but how we go about solving the issues we care about.

It is not afraid of AI — rather, it's keen to interrogate it fully and thoroughly: the advantages, and the unintended impact on places and communities across the country.

It is not afraid to ask uncomfortable questions. Something I personally find very refreshing. And positively challenging to assumptions that third sector organisations are risk averse, or willing to maintain the status quo. Often unsung for its role and contributions, the third sector is the always-present social and financial infrastructure that keeps together the fabric of our democratic society.

WHAT WE'RE ACTUALLY WORKING ON

Spoiler: I am no expert on AI. But it's a tool we're actively exploring at JRF — both internally and with partners. Three areas are front of mind right now:

DATA, TECH & BLACK COMMUNITIES Exploring the impact of AI on Black communities and where bias manifests in practice.

ACCESS SOCIAL CARE Bridging the gap between quantitative and qualitative evidence pertaining complex statutory care experiences.

GROUNDED VOICES Testing whether AI can help process large volumes of qualitative insight without flattening it.

We haven't reached a fully informed conclusion — and we're honest about that. But we've been asking ourselves one key question: can the third sector harness the power of AI? Here's what the picture looks like so far.

RISKS	OPPORTUNITIES
Embedded bias AI trained on existing datasets inherits their gaps. Marginalised groups are frequently under- or misrepresented — and those distortions carry forward into outputs.	Bridge qual & quant LLMs can help connect lived experience with statistical data, creating a richer, more complete picture of need and impact.
Unclear community impact The data we collect refers to real people and places. The downstream effects of AI-driven decisions on individuals remain poorly understood.	Align funders & grantees AI can help build shared language around success criteria and data — reducing friction and building more equitable partnerships.
Widen exclusion Without deliberate design choices, AI risks deepening existing inequities — excluding the communities it should serve and entrenching power imbalances.	Reveal intersectionality AI can support triangulation of large, complex datasets, surfacing overlapping needs in ways manual analysis rarely can.
Quantitative dominance AI systems trained on pattern recognition in structured data implicitly privilege numerical, measurable evidence over lived experience.	Democratise knowledge AI can dramatically widen access to information, challenge elitism in how knowledge is produced, and amplify grassroots voices at a local level.

BEFORE YOU COMMIT TO ANYTHING

On the whole question of whether the sector should adopt — and pay for — AI tools: I don’t know about you, but I'm getting CRM flashbacks. The most pressing risk, in my view, is committing to something that does a multitude of things, but not quite the specific thing you actually need.

A FEW THINGS WORTH DOING FIRST

→  Take advantage of free tools, trial periods, and bespoke licences before committing long-term.

→  Ask the same question to different LLM models and compare the answers — this is something I like doing myself. You'll learn a lot quickly!

→  Be very clear on what outcome you're looking for before you start any of the above.

So, whatever you will decide on AI, whatever you will use it for, never forget to keep it human. The data and information we deal with and generate in the sector is first and foremost data about people and, as such, it should serve the people the data is about in a way that is proportionate, compassionate and truly reflective of their experience of dealing with hardship, inequality, stereotypes and prejudice on a daily basis. 

Rosario Piazza,

Chief Insight Architect