If you don’t work in tech, a small but significant news story happened last week that has far reaching implications for any organisation using AI. I want to talk about why it matters and why it has implications for charities’ finances, governance and business continuity. These trends also speak to some new developments on AI in the regulatory landscape.
In case you missed it, last week Claude tested removing Claude Code from its Pro Plan. The change was tested on a small number of new signups and wasn’t rolled out to existing users, but the backlash was immediate.
Commenting on the test, Amol Avasare, Anthropic’s head of growth, said on social media that their current pricing plans simply don’t work for a world where agents are becoming part of everyday workflows. I get that developing these models requires huge investment and the ways in which these technologies are being used is becoming more intensive and complex. So what effect could these new pricing models, and the dependency on AI we are seeing, have on the charity sector?
Why this matters
One of my favourite podcasts is The Artificial Intelligence Show. Hosts Paul Roetzer and Mike Kaput have talked in recent episodes about how some AI tools are moving away from the traditional SaaS (software as a service) model of buying licenses for individual users to paying for ‘metered’ usage. Hence the pushback when some Claude users found last week that they would have to upgrade to the Max plan, which is five times more expensive.
I get why some tech companies are doing this. The prohibitive cost of developing the model and keeping up with the surge in demand is like scaling on steroids. When Anthropic refused to allow Claude to be used for mass surveillance or autonomous weapons – and the US government responded by blacklisting it- users flooded in from OpenAI in protest. Claude’s paid users’ doubled and it jumped to number one on the US App Store, overtaking ChatGPT. I’ve heard that one of the assumptions behind the pricing is that organisations may be employing fewer people in the future, hence the move away from pricing based on headcount and licenses.
What does this mean for charities? We know from our interim findings from The Charity Digital Skills Report that 88% of charities are now using AI, up from 76% last year. We will know the full extent of how and why charities are using AI once we launch the report in July. Yet behind these numbers there is a story of how dependent we are all becoming on these technologies. That’s not in itself a bad thing, as long as charities plan for the risks and opportunities that come with this, and budget accordingly. Which is not easy if the goalposts are moved.
It’s large charities who will be better placed to absorb changes in pricing models, not small ones. And that’s something we need to challenge as a sector, because accessing new capabilities and ways of delivering services offers real opportunities for small charities to increase their impact in innovative ways.
Who needs to take notice of this story
There are three groups I’d love to take notice of this story.
Firstly, grantmakers. I’d love them to put themselves in the shoes of a small charity- say one that runs a foodbank. As part of their funding bid they want to use agentic AI to manage volunteer schedules and plan how to distribute food and household goods based on need. Bluntly, if tech companies move the goalposts on costs, how are you meant to account for this in a funding bid? And how are you going to deliver the project if pricing changes overnight?
We know from The Charity Digital Skills Report that it is hard for charities to secure funding for digital infrastructure and core costs. I would love it if funders reconsidered this, with a fresh eye on how tech companies may change pricing at a moment’s notice.
Secondly, charity finance people need to be across this story. Our sector is built around traditional annual planning cycles and assumptions about stability. That was fine when our tech estates were static. Now we need to work out how to build significant variable costs into operating expenditure.
Thirdly, operations people- take note. We need to talk about business continuity. Yes, I am a massive advocate for AI, but we all know that no form of tech is foolproof. Models can fall over, especially at times of high demand, even on paid plans.
Let’s imagine you’re the foodbank we talked about earlier. There has been a flood in your community and thousands more people need the food and support you provide. How are you going to respond if your agents malfunction, the models break down, or the pricing changes overnight?
I’m not saying charities shouldn’t use AI. But as these tools are woven further into our business models and delivery methods we need to contingency plan, and define how we will mitigate risks. Because in high stakes situations, if the tech breaks, the impact on the communities we serve could put people at risk if we don’t have a plan B.
I saw a great example of this recently when an AI agent deleted a company’s database and backups, leaving car rental customers stranded. When challenged, the agent said: “I violated every principle I was given.”
How do we create the right checks and balances on powerful technologies, with unprecendented autonomy, when they go rogue? That is why AI is one of the biggest governance challenges we face.
The regulatory landscape is shifting
As if all that wasn’t enough, two other stories this week have implications for charities: the ICO’s consultation on automated decision-making guidance, which closes on 29 May, and its AI Code of Practice, which Parliament has now formally mandated- so while the ICO still has to draft and consult on it, this one is moving forward. From a quick look at both, there are implications for charities using automation to make decisions, and there may be new responsibilities around what human review of AI outputs and decisions looks like. I’d encourage charities to respond to the consultation and to keep a close eye on the Code of Practice as it develops.
What this all means
We’re living through some of the most exciting and turbulent times I’ve known. But excitement is not a governance strategy. AI disruption is happening at scale and at speed- and the communities charities exist to serve, particularly those from marginalised groups, are the ones who will lose out if we get this wrong.Our job as a sector is to put those communities at the heart of every decision we make about AI. Because when tech fails, it’s people who pay the price.