Christine, a 44-year-old children’s nurse, who lives in the Wirral with her partner and her three-year-old daughter, describes herself as ‘not much of an early adopter’.
Beyond a bit of ChatGPT – to research holidays and answer everyday questions, whether or not a service is AI-enabled "never enters my head as a consideration."
Except Christine also uses an AI agent to manage her household finances and life admin. It's called Nous.
Silicon Valley can’t get enough of personal agents right now.
Early adopter types are queuing up to tell the world that Instinct booked them a table at the hottest NYC restaurant, or that Muse ordered them groceries for a week’s worth of high-protein dinners.
For the rest of us, exactly how these general purpose agents are actually going to be helpful isn't quite so clear yet.
So far, they’re mostly geared towards a narrow category of tech natives.
As a result, they’re not actually that good at solving the problems that most people routinely face, like getting a good deal on energy bills or renewing car insurance.
That’s why I’m convinced vertical agents are going to matter so much.
Nous isn’t a general purpose agent. We don’t promise to do a bit of everything. We’re unashamedly specialised at doing one thing really well: managing household spending and admin.
That’s why our customers are people like Christine. People who are a long way from your mental picture of a ‘tech early adopter’, but who are delighted to hand off their everyday admin tasks to an outsider – AI or otherwise – so long as it's competent.
Most of them are women–Mums in fact. They live all over the country.
And they’re not handing responsibilities to Nous because they love playing with new tools.
They're handing over some of the mental load of running a household – groceries, bills, forms, calendars, laundry, childcare – which is endless and mostly thankless, and falls disproportionately on women.
On a practical level, building a vertical agent for the household means distinguishing between things that should be accomplished by probabilistic systems (with the right safeguards) and things which require deterministic systems.
We've opted for a hybrid architecture that combines LLMs where flexibility is paramount with rules-based logic for the really consequential moments – like managing your money – to avoid the risk of costly mistakes or hallucinations.
We think this combination (chat technology for ease of interaction, with robust deterministic flows for high stakes processes) brings together the benefits of AI with the safeguards needed for this class of technology to meet consumer needs reliably.
And because our core flows are rules based, it also means we can work fully within regulatory requirements–which is why we can help out with categories like car insurance.
All of this means the agent really works, at the task it’s designed to do, for people who really need it.
But for Christine, what goes on underneath the hood doesn’t really matter.
As she put it: “Nous just makes things more efficient. Where I am right now in my busy life, I'm up for that.”
