Option 1 · the full picture
When we last met, the plan was to build you a product. Since then we've done the deeper work — on the market, not just the tech — and two things have moved the direction. Here's the whole breakdown.
We pressure-tested the products you'd expect us to build, and looked at whether the market works — not just whether AI can make the thing.
A price war — Farewill sold in a fire-sale, rivals at ~£20. Demand is passive, so winning a customer costs £150–330 on ads, and margin turns thin-to-negative once you add the legal safety net. No moat.↗ click for the full breakdown
The government portal is already free. AI only helps at one of the five stages, and buyers are older and want a person. It works as a cross-sell after wills — not on its own.↗ click for the full breakdown
The strongest need we found — but free charities anchor willingness-to-pay at £0, the community is actively warning parents off AI, and a failed case for a child is catastrophic to the brand.↗ click for the full breakdown
What just happened · June–July 2026
Anthropic launches Fable 5 — the most powerful model it had ever released.
The US government pulls it. A national-security export-control order, three days after launch — because it can find security flaws in essentially any software it's pointed at.
Back online, with new guardrails. Its more powerful sibling, Mythos 5, stays locked to ~50 vetted defenders under "Project Glasswing".
This is the kind of capability we're now building on — and it's the slowest it will ever be. Against that, the room to charge for a simple AI product shrinks to nothing.
This guts the product moat.
Any product whose pitch is "the AI does it for you" can be rebuilt in a weekend once the model is free and good enough — and Fable already is (powerful enough that the government briefly pulled it). The value has left the tool; it now sits with whoever people trust to point them at it — the whole reason the direction moves to you.
Of everything, this is the product we like most. The demand is real and urgent — a defendant with a claim form has a 14-day clock, and there are ~1.4m county-court judgments a year — and it's cheap to reach because it's a moment of need.
But two catches. The commoditisation point still applies — anyone can build it. And its whole value rests on being airtight and truth-based: right now the AI is essentially grading its own work against an AI-written rubric. To stand behind it — with your name on it — we'd need external legal validation and ongoing checking (effectively lawyers). That's the wall we've hit.
Here's the positive: the one thing AI is unambiguously brilliant at is force-multiplying you — turning a little of your time into a lot of trusted content across many channels. We're genuinely confident in the top of the funnel — your reputation, audience and network. What we're missing is the product to land that traffic.
So the proposal: build the audience and brand first, then point that traffic — starting with the affiliates and network you already command, staying flexible on exactly where it lands. It's the MoneySavingExpert playbook: one trusted expert, genuinely useful content, an owned audience, monetised downstream.
Option 2 · the commoditisation case
The biggest thing that's changed isn't our opinion of the products — it's what AI can now do. That reframes the whole plan.
What just happened · June–July 2026
Anthropic launches Fable 5 — the most powerful model it had ever released.
The US government pulls it. A national-security export-control order, three days after launch — because it can find security flaws in essentially any software it's pointed at.
Back online, with new guardrails. Its more powerful sibling, Mythos 5, stays locked to ~50 vetted defenders under "Project Glasswing".
When a model is powerful enough for the government to suspend it on national-security grounds — and this is the slowest it will ever be — no simple AI product has a moat.
The AI product moat is disappearing.
You can't charge for a wrapper around a tool that's free and this good. Any "the AI does it for you" product can be rebuilt in a weekend — so the moat was never the software. It's trust, brand and audience. That's what you bring.
These products already had weak foundations, and commoditisation makes each worse:
If anyone can build the product, the durable advantage isn't the software — it's trust, brand and audience, exactly what you bring. And that's not a consolation prize: it's the one asset AI makes more valuable, not less. Here's what it looks like in practice.
AI turns a little of your time — your take on the week's legal stories — into a lot of content across every channel. You stay the voice; AI does the volume.
That builds what a wrapper never can: a trusted audience that's yours. Trust compounds where products commoditise — this is the moat.
Once the audience exists, we direct it — starting with the affiliates and network you already command, staying flexible on exactly where it lands.
It's the MoneySavingExpert playbook applied to law: one trusted expert, genuinely useful content, an owned audience — monetised downstream, once the trust is built.
Option 3 · the map
Since we last met, two things moved — the market turned against the products we discussed, and AI took a leap (Fable, below). So rather than debate each product, we put every one through the same honest test: what's the real blocker, and does AI actually remove it?
| Product | The real blocker | Does AI fix it? | Verdict |
|---|---|---|---|
| Wills | Price war + passive demand (high cost to win a customer); no moat | No — AI can't lower the cost of acquisition | Weak |
| LPA | Gov portal already free; only 1 of 5 stages helped; older buyers want a person | Barely | Cross-sell only |
| SEND tribunal | Free charities anchor willingness-to-pay at £0; community hostile to AI; catastrophic brand risk | No — none of the blockers are capability | Strongest need, hardest economics |
| County Court Defence | Real, urgent demand — but must be airtight/truth-based, and needs external legal validation | Partly — but AI grading itself isn't enough | Warmer, but gated |
The answer is the same almost every time: the product isn't the problem — the cost of winning customers and the lack of a moat are, and AI fixes neither.
The closest call is County Court Defence. The demand is real and urgent — a defendant has a 14-day clock, and there are ~1.4m county-court judgments a year. But its whole value rests on being airtight, and right now the AI is grading its own work against an AI-written rubric. To stand behind it — with your name on it — we'd need external legal validation and ongoing checking. That's the wall we've hit.
And the ground moved under all of them.
Fable (out ~a week) lets ordinary people do this themselves — a non-technical mum just won her bank-fraud refunds with free AI. It's exponential; by launch it's stronger again. So even a good product gets commoditised — the moat was never the tool.
Run the same test on your reputation and the answer flips. It's exactly what AI is brilliant at — force-multiplying a trusted expert into reach — and it's the one asset that gets more valuable as AI spreads, not less.
A little of your time — your take on the week's legal stories — becomes content across every channel. You're the voice; AI does the volume.
That builds a trusted audience that's yours — and trust compounds where products commoditise. It's the moat AI can't erode.
We're confident in that top of the funnel; we just choose what it lands in — starting with the affiliates and network you already command.
It's the MoneySavingExpert playbook, applied to law.
Legalhelp × Rob Rinder · internal draft for discussion · three framings of the same argument — pick one.