Security & trust
Built so you can trust it with real work.
The questions your IT lead, your finance lead and your lawyer will ask before an AI system touches your customers, your data and your accounts. Here's how we handle each one.
Security & trust
Security designed in from the first workflow.
Before anything is built, you sign off what the system can read, write and send, and who can change it.
Your data stays in your systems.
We work inside the tools you run and move the minimum the task needs.
Access is the narrowest that works.
In your accounts, with named people and credentials kept in a secret store.
A person signs off what matters.
Decisions with consequences stay with a named owner on your team.
Every run is logged.
Inputs, decisions and outputs are recorded, so any result can be traced.
How we handle each question.
01Data handling
Before we build anything, you get a list of the data each part of the system will read, write or send elsewhere, and you sign it off. Where that data is stored, which third-party services process it, and for how long, is decided then and written into the proposal.
We work inside the systems you already run and move the minimum data the task needs. If a model provider is involved, you're told which one, what it receives and what its retention terms are.
02Access controls
Each connection gets the narrowest access that lets it do its job, in your accounts rather than ours wherever the platform allows it. Credentials live in the platform's secret store, never in a workflow, a document or a chat.
You decide who can see, run and change each system, and the people with access are named.
03Human review
Every system has a line between what it does on its own and what it prepares for a person. Decisions with commercial, legal or customer consequences stay on your side of that line until you choose to move them.
Exceptions, low-confidence results and anything outside the agreed cases go to a named owner on your team, with the context they need to decide.
04Testing
We test with your team on your examples: the typical case, the awkward case and the case the system should refuse. The test sets are kept, so any later change can be checked against them.
Nothing enters daily work until the people who'll rely on it have watched it handle their own cases.
05Monitoring and logging
Each run is logged with its inputs, the decisions it took and its outputs, so you can trace and explain any result. Failures and exceptions go to the owner you've named.
What is reviewed, how often and by whom depends on what the system does, and is agreed with you for the engagement.
06Integration with your systems
We connect through the options your systems already offer, such as their APIs, webhooks and exports, and choose the most dependable route over the quickest. If a connection is missing or restricted, you hear about it before the design is fixed.
Whether a specific product can be connected is confirmed during the strategy step.
07Ownership
Workflows, prompts, configuration, documentation and the accounts they run in are set up so they stay yours. Licensing for any third-party service, and anything we keep the rights to, is set out in the engagement terms.
You can take the system to another team or supplier without needing us.
08Support after launch
After launch we agree what you need from us: support, maintenance, performance reviews, further improvements or a handover to your own team. Responsibilities, response arrangements and fees for that scope are written into the engagement.

What you sign off before we build.
- 01The data each part of the system will read, write or send elsewhere.
- 02Where it's stored and processed, and for how long.
- 03Which model provider, if any, receives it.
- 04Who can see, run and change the system.
- 05Which decisions the system makes, and which it prepares for a person.
Questions
Security questions.
Is our data used to train AI models?
Not by us. Where a model provider is involved, we tell you which one, what it receives and what its terms say about retention and training, and you sign that off before we build.
Can the system run entirely in our own cloud?
Often, yes. It depends on the tools involved. Wherever the platform allows it, we set the system up in your accounts under your security controls, and tell you early where that isn't possible.
What happens when the system isn't sure?
It stops and asks. Low-confidence results and anything outside the agreed cases go to a named person with the context they need.
Will you complete our security questionnaire?
Yes. Send it over or bring it to the call, and we'll walk through how a specific system would be built for you.
Make your business AI-native.
Start with a short call about where your team's hours go and what you want to change.