Vision¶

Where Maud is heading, and why we think it lasts. This is direction of travel and first principles — not a plan with dates. For what's actually queued, see the Roadmap; for what's already live, see status.md at the repo root.

The job, stated plainly¶

The purpose isn't to file documents. It's to get a household ready — ahead of the moments when readiness actually matters: selling a house, a mortgage, an insurance claim, probate, a bereavement, an emergency, a parent who suddenly needs help. The filing is just how readiness gets built.

The test we want to pass is simple: when one of those moments arrives, everything you need is already there — complete, current, structured, and easy to hand to a solicitor, a broker, a hospital, or a sibling.

That's a deliberate move away from the "filing cabinet" frame. A cabinet is passive: it holds what you put in it and does nothing with it. Maud is meant to be active — to understand a household, notice what's missing or about to lapse, and keep quietly moving you toward ready.

The hook: not "get organised"¶

Nobody wakes up wanting to file documents. "Get organised" is a New Year's resolution — a chore that fades by February. It's a vitamin, and vitamins don't sell.

The real pull is curiosity and a little low-grade anxiety about your own affairs: Is my gas safety certificate actually current? Does my mortgage require buildings insurance I'm not sure I have? Is my passport inside the six-month rule for that trip? So the promise isn't "we'll help you tidy up." It's "we'll find out about your household and tell you what you're missing or exposed on." Same mechanism — much stronger hook. A series of small painkillers, not a vitamin.

How it works: diagnose, don't interrogate¶

Most onboarding either drops you on an empty page or interrogates you with a form. Answering twelve questions feels like work and pays nothing back per answer.

Maud flips it. After scanning your inbox (read-only) she produces a verdict, not a pile:

"Vehicles: strong — MOT, tax and V5C all tracked. Property: thin — I found your mortgage and council tax, but you're missing an EPC and a gas safety certificate, and I can't see buildings insurance. Identity: a passport expires in seven months."

From there the guidance is a short, adaptive conversation rather than a form — and it's always grounded in what the scan already found, so it feels like a knowledgeable friend, not a cold questionnaire:

1. Ground in what we already know — never ask what we can infer. "I can see a Halifax mortgage and an OX postcode — is this a house or a flat, and do you own it or rent it?"
2. Establish the type — a few questions that fix the shape of things (owned or rented, leasehold or freehold, who's in the household, what they drive).
3. Map the obligations that type implies — this is the curated UK-knowledge layer. A let flat needs gas safety, an EPC, landlord insurance; a leaseholder has a service-charge and freeholder-insurance chain; a country house probably has a septic tank and no mains gas.
4. Diagnose — match obligations against what's on file: strong / weak / missing / expiring.
5. Act — every gap resolves to upload this, we'll watch this date, or here's how to get it.
6. Keep it live — readiness drifts. Things expire, life changes, you buy a car. Maud reopens the relevant conversation when it does.

That last point is the answer to the "it's a one-time job" problem. Tidying your inbox once is a great front door, but it's not a reason to come back. Readiness that drifts — and a Maud who notices — is. And it stays strictly inside the frame: we're telling you what's lapsing, not managing your money.

What we won't do¶

The fastest way to kill a product like this is to become horizontal. "AI that does useful things with your inbox and documents" is exactly the ground Gemini-in-Gmail, Apple Intelligence, and ChatGPT-with-connectors will own — they have the inbox, the OS, and infinite scope. Our defensible move is the opposite: be the narrow, trusted, structured system of record for a specific job that a generalist won't bother to do well, because doing it well takes curated UK domain knowledge, durable custody, and an output a generalist won't produce.

So we hold a hard line:

• In: anything a household needs to keep, prove, and act on at a life-admin moment. "Your buildings insurance renews on 14 November, here's the policy" is dead centre.
• Out: ongoing money-optimisation and bookkeeping. We are not freelancer expense capture (that's Xero / FreeAgent) and not subscription-spend cutting (that's Rocket Money / Emma / Snoop).

The line is action: we remind, store, and produce packs; we do not do accounting or optimise spend. Narrowness isn't a limitation here — it's the strategy.

Who we're up against¶

The closest thing to a category leader is the US product Trustworthy — the "family operating system" for a household's important documents. It's useful context in three ways:

1. It validates the category. People will pay to keep their family's documents in order. We're localising proven demand, not inventing it.
2. It leaves the UK wide open. Trustworthy will never build TA6 forms, gas-safety logic, MOT/DVLA lookups, leasehold service charges, council tax, NHS or HMRC admin. UK depth is ours to own.
3. Its name nails the emotional core: trust. For a thing holding your most important papers, "can I trust it?" is the central feeling.

But we don't beat Trustworthy at the word trust — we beat it on register. Trustworthy is institutional, literal, American, a corporate vault. Maud is the opposite: warm, British, human, conversational — a capable, discreet person who keeps your household in order. That contrast is the positioning. "Hey Maud, am I on top of everything?" versus a login to a vault.

The longer game¶

Two things are where the durable value really sits. Neither is something we can claim on day one — a moat here is something you manufacture over time. (We're honest about this: UK domain-depth is a head start a competent team could copy in a quarter; trust must be earned; being narrow is a survival strategy, not a barrier. The real moats are below.)

Pillar 1 — Preparedness intelligence¶

The interesting engine isn't classification. Anyone can point an AI at an inbox, and accuracy plateaus quickly. The interesting engine is learning, across many households, what good preparation actually looks like for a given situation — and using that to make each household excellent before it needs to be.

Concretely: what should a leasehold-flat-with-a-new-baby household have on file, keep current, and pre-assemble, so a remortgage or sale is effortless? What do people wish they'd kept when a parent died? What's the difference between "has the documents somewhere" and genuinely "ready"? Today this exists only as its shadow — junk-suppression, learning what to ignore. The bigger, better version is the opposite: learning what to do, and keeping every household continuously sale-ready, claim-ready, executor-ready.

This gets better the more households use it — a real but helpful edge: the product gets smarter, rather than the exit getting harder. It needs the data-protection groundwork done first (a DPIA before any cross-household learning switches on), and it stays personal-by-default — advice shaped by patterns, never your documents shared with anyone.

Pillar 2 — Getting there through institutions¶

Direct-to-consumer is the starting point, but the same readiness product is far more powerful delivered through organisations that already sit at these moments and have a stake in their people being ready:

• Embedded with adjacent professionals — conveyancers, mortgage brokers, estate agents, insurers, will-writers, accountants. They get a better client experience and a tidy pack exactly when it's needed; we get distribution and a relationship that's genuinely hard for a generalist to replicate.
• White-labelled by larger players — banks, insurers, and the personal-finance incumbents (Quicken and similar have something here, but nothing strong) that want a household-readiness layer without building it.
• As a public good — longer-term and more speculative: a service like Citizens Advice offering it to the people it helps, or something a government or local authority sponsors as everyday civic infrastructure for life admin. This is upside, not the plan — but it's the kind of reach and trust a product like this could earn, and a reason to build to a standard an institution would put its name to.

This is where "being UK" stops being a copyable feature and becomes a real edge: the domain knowledge can be replicated, but the relationships with UK institutions can't be, quickly.

We won't win by locking people in¶

Plenty of software defends itself by making leaving painful. We'd rather not. We want people to stay because Maud is genuinely excellent and trusted — and we want it to be easy to take your documents and go. Clean export isn't a risk to manage; it's a feature, and it's the precondition for the partnership and public-service routes above. Nobody hands a public-good service to a company that holds its users hostage.

What this asks of us now¶

Both pillars only work on genuinely excellent, trustworthy foundations — accurate classification and rich extraction, a reliable proactive layer, security and privacy an institution would stake its name on, and clean data export. None of that is a far-future item; it's the near-term work on the Roadmap, done to a higher standard than an MVP strictly needs.

So more than anything, this vision is a reason to get the fundamentals right.