Compare

Cast & Rule vs a public AI chatbot

A general AI chatbot is fast, familiar and often already open in a browser tab. That convenience is real. So is what it means for confidential client information once it's been typed in.

Where a public chatbot is genuinely good

It needs no setup and gives a first answer in seconds. For drafting that touches no client data (a covering email, a general explanation of a rule, a first go at wording) it's a reasonable tool. ICAEW's own guidance doesn't tell members to avoid AI altogether. Business plans from the same providers can offer contracts, data-retention and no-training settings that personal accounts don't. A managed business assistant is a different decision from a personal chatbot account.

Where the concern sits

ICAEW's generative-AI guidance says confidential client information shouldn't be put into public AI tools, because once it's typed in you have little control over how it's shared, secured and kept. A public chatbot's consumer terms are written for anyone, not for a UK practice's confidentiality and data-protection duties.

Compare

Side by side

Where Cast & Rule differs

Client data processed on local hardware
A named partner rules on each piece of prepared work
A locked Ledger of what the AI did

How the two approaches handle client data, sign-off, logging, firm rules, and where the work runs
	A public AI chatbot	Cast & Rule
Where client data goesDetail	Sent to the provider's service Sent to the provider's service. Personal accounts run under consumer terms; business plans usually offer different contract, retention and training settings. Check which one staff actually use. Tell me more about finding which AI tools staff already use: the shadow-AI checklist (opens in a new tab)	Local hardware the practice controls Processed on local hardware the practice controls. Results the rules allow to leave are encrypted before they leave. With cloud AI off, no client data leaves, and a report shows what left through Cast & Rule. Each client's records sit in their own encrypted store. Tell me more about where client data lives, on the security page (opens in a new tab)
Who signs offDetail	Nobody, by default Nobody by default: the output can be copied straight into client work.	A named partner signs off each piece A named partner rules on each piece of prepared work before it goes anywhere. Tell me more about the partner's ruling, on How it works (opens in a new tab)
What is loggedDetail	Whatever the provider logs Whatever the provider's product logs, for its own purposes; some business plans offer audit logs to administrators.	What the AI did, in a locked Ledger The Ledger records what the AI did, with each entry locked to the one before it. We're extending it to every AI action now. Tell me more about the Ledger, on the security page (opens in a new tab)
Which firm rules applyDetail	None, unless the practice enforces its own None, unless the practice writes and enforces its own separately; see the AI-use policy template. Tell me more about setting the practice's own AI rules: the AI-use policy template (opens in a new tab)	The practice's own, on each task The practice's own rules apply as permissions on each task. Connection controls limit what each connection can reach. Tell me more about Rules and the Strongroom, in the product tour (opens in a new tab)
Who controls the AI's instructionsDetail	Whoever types the prompt Whoever types the prompt.	Signed, versioned tasks We sign every task, and the machine checks the signature before it runs. Tasks are versioned: your staff can't edit them, and a change needs your approval. Tell me more about signed tasks, in the security architecture (opens in a new tab)
Where it runsDetail	The provider's infrastructure The provider's infrastructure; data-residency options depend on the plan.	A local model on the practice's hardware A model running on local hardware the practice controls, on macOS. Each task is tested against named model versions and upgrades only with your approval. Your practice gets the list of suppliers who handle data, and where it flows. Tell me more about install and cloud options, on Security and governance (opens in a new tab)

In summary

A public chatbot is convenient because it has no rules of your practice's own. That's also why confidential client information shouldn't go into a personal account.Cast & Rule works the other way round: narrower than general chat, and built around your practice's rules, a named reviewer and a record of what the AI did.

Read the security overview

Questions partners ask

Frequently asked questions

Is using a public AI chatbot against the rules?

Not automatically, but ICAEW guidance is clear that confidential client information shouldn't go into a public AI tool. The AI-use policy template sets out what that means in practice.

Does Cast & Rule replace general AI chat entirely?

No. It's built for work that touches client data and needs a record. General drafting that touches nothing confidential is a separate question for the practice's own policy.

Has anyone measured how often practices paste client data into public tools?

Not with a checked figure for UK accountancy. Wider workplace surveys show real risk and a gap in the rules; see the shadow-AI checklist for the evidence and its limits.

One practice first

Start with one workflow, over six weeks, with us alongside

For a UK practice of 5 to 20 staff, with a partner who'll own the review. We run one workflow, measure it before and after, and agree in writing how it ends. Send an enquiry or email us today.

Discuss a scoped pilot or get the AI-use policy template hello@castandrule.co.uk