What really mattered to me was the value actually being added by AI. What is the real-world utility, what can we trust it to do, and does it ultimately improve people’s lives?
We are already seeing genuine utility for PocketSmith customers through our MCP server. People can gain new insights into their finances while still remaining in the driver’s seat, and the administrative burden of being the Household CFO can be greatly reduced. Long-time customers have clarified their financial position, while new customers can organise and understand their data much more quickly.
This is where AI gets genuinely interesting to me. It can help us get important things done that we either can’t do, don’t have time to do, or find too tedious to face. Plenty of the work involved in running a household’s finances fits into those categories. AI can help a Household CFO clarify their options, identify patterns and understand the possible consequences of a decision.
At the same time, AI is moving from simply helping us find and understand information to acting on our behalf. Most people seem reasonably comfortable using AI for research, but less comfortable allowing it to make purchases or financial decisions. That distinction matters because we need to understand whether an AI is aligned with our interests and genuinely capable of making the decision we are asking of it.
An AI agent could soon compare and purchase insurance, manage subscriptions or make increasingly complex financial decisions. These applications could be enormously useful, but the consequences of getting them wrong could also be significant.
As Household CFOs, we should think of AI as part of our financial team rather than a replacement for our own judgement. It can help us become more informed, capable and confident. Before allowing AI to act on our behalf, we need to understand its limitations, check its assumptions and remain engaged with the decisions it makes, because it is our life, after all.
Olav Nielsen, Head of Growth
AI gets genuinely interesting in finance when it stops being another place to look at your money and starts becoming a way to actually work with it. That was the idea I kept coming back to at our Intersekt panel.
For years, PFMs have done a great job of giving us more visibility: better dashboards, better categorisation, better reporting, better ways to understand where our money went.
AI helps us go beyond visibility, towards financial agency.
Instead of hunting through menus and reports, you can now just ask the question that’s actually on your mind: Can I afford this? Why am I spending more than usual? What patterns are you seeing about my money that I should be aware of?
And then, the truly interesting part: asking AI to help you organise your money: Please categorise my transactions; Make me a meaningful budget; Schedule a weekly financial check-up with me.
What excites me is what happens when you shorten the distance between seeing your finances and doing something useful with them.
We’ve been exploring this through our MCP server, which lets people securely connect their PocketSmith to tools like ChatGPT and Claude, giving them the capabilities of a great household CFO.
Until now, that level of help has mostly been available to people who can afford professional advice or who are highly engaged with their finances. AI is changing that.
Of course, the more useful it becomes, the more trust matters too. Accuracy, context, consent and transparency need to be built into the architecture of any software that handles someone’s financial life.
For me, that’s where AI gets properly interesting: not when it gets better at talking about money, but when it helps more people feel capable of making better decisions with it.
Jason Leong, CEO and co-founder
For me, it’s a question of what taking responsibility for AI actually looks like. A gnarly one, I know!
When AI gets something wrong with someone’s money or data, who answers for it? That’s a pressing question that is undoubtedly being worked on. But I came home from Intersekt thinking about how much earlier that responsibility starts, like when a business decides whether something needs AI in the first place.
There are people in fintech taking this seriously, and it would be great if that care became an expectation across the industry. Among the useful tools, there’s plenty of AI for AI’s sake. “We can build this” needs a follow-up: “Who does this actually help, and what does it cost?” Privacy, copyright and environmental impact belong in that conversation too.
At PocketSmith, there’s a healthy spectrum of opinions about AI, from enthusiastic to rightfully uneasy. I value that, and that respect for different comfort levels shapes how we’re approaching AI for our users, too. We’re working on ways for people to use it as a tool on their own terms, with control over what they share and how they engage. For me, that agency is part of that wider responsibility. When given a choice, an answer should be able to be a “no thanks”.
That’s what I think the industry will need to grapple with — taking responsibility for the whole thing, from the decision to build something to the choices available to the person using it. When someone’s money is involved, “look what it can do” only gets us so far.
Chloe Adams, Content and Partnerships Marketing Manager
The increasing capabilities and accessibility of local LLM models are very interesting. Over the past year, open-source models have started to rival frontier models (like Claude and ChatGPT) for some use cases — if you have the hardware to run them.
What this means is that more work is able to be done in-house, without reaching out to frontier models to provide answers. Fine-grained or deep work is still not there yet, however things are moving at such a rapid pace that I’d expect that in the next 12 months the capability of open-source models to replace frontier models will be that much further on.
James Wigglesworth, CTO and co-founder
