Senshin holds the plans, budgets, risks and politics of real programmes — often for people who work across several client organisations at once. So isolation and restraint aren't features here. They're the foundation.
The hardest case is the honest one: an independent consultant or a delivery lead running several clients inside the same account. Their clients must never bleed into each other.
Every project lives inside a workspace, and every single data operation — reading a risk, loading a budget line, running an AI analysis — is scoped to that workspace before it returns a thing. There is no shared pool a query can accidentally reach across. This isn't a promise on a page: we run multi-tenant isolation tests on every code change, and a release that could let one workspace see another's data does not ship.
Encrypted at rest and in transit, with access limited to your workspace.
Senshin runs in the browser. Your client's data is shown to you; the browser app does not save a copy to the machine beyond what your browser holds in its normal cache and the files you choose to export. (The optional mobile app keeps an encrypted offline copy — see the Privacy Policy.)
We never sell personal data and never share it for advertising. Our privacy notice describes the kinds of service provider we use. Aggregated industry statistics, if published later, contain no one identifiable.
The AI reads your project to help you. We don't use it to train foundation models, and every AI request is routed only to model hosts that don't store or train on it. Learning from programmes is off unless your organisation chooses it.
This is where most platforms get vague. Here's the specific mechanism.
Senshin reasons in two stages. First, a deterministic engine on our own servers extracts the signals — schedule slippage, budget variance, RAID velocity, and so on — with nothing leaving our infrastructure. Then, to write that up in plain English, your project context is passed to a language model through our AI gateway. Direct calls to model companies are blocked in code, and every request asks the gateway to use only model hosts that do not store or train on prompts. We ask for that on each call rather than leaving it to a provider's default setting. Our privacy notice describes the kinds of provider we use.
When you drop files to start a project, Senshin reads the workbook or plan itself first. A language model is used only for running prose that still needs one.
Data controller: Extreme Gravity Ltd, trading as Senshin (senshin.io), company number 09590911, registered in England and Wales. Full data-subject rights, honoured within 30 days.
A Data Processing Agreement is available for every customer. Sub-processor information available on request.
Data changes, AI runs and sign-ins are recorded in an audit trail.
Encrypted at rest and in transit.
We do not hold SOC 2, ISO 27001 or Cyber Essentials certification, and we do not claim to. If your organisation needs one, tell us before you buy.
If you believe you've found a security issue, email security@senshin.io. We read every report and respond quickly. Responsible disclosure is always welcomed, never penalised.