How we use AI
AI on the volume, people on the decisions
An AI-native marketing service should be able to tell you exactly where the AI is and where it is not. This page does: what the models do, what they are never allowed to do, where a person checks the work, and what happens to your data.
What AI does
- Research. Reads a prospect's website, recent news and hiring signals and summarises what the company does and why now might be the right time.
- Enrichment. Fills in firmographics and verifies business contact details against the sources we use, so lists are built on checked data.
- First drafts. Opening lines, follow-ups, article drafts, LinkedIn posts and nurture emails — always as a draft for a person to edit.
- Segmentation. Groups contacts by fit and behaviour for scoring and nurture, against rules a person wrote and you approved.
- Ad variants. Generates creative and copy variations for testing. A person picks what runs.
What AI never does
- Decides who is contacted. The ideal customer profile, the exclusions and the list are approved by a person and by you.
- Sends anything without approval. Every sequence is approved before it goes live, and nothing sends from your domains on a model's say-so.
- Replies to a prospect unread. Every reply is read by a person before anyone answers it.
- Invents facts, claims, statistics or case studies about you or about a prospect. Every claim in an email or article is checked.
- Makes a commitment on your behalf — a price, a promise, a meeting time you have not offered.
Where a person checks the work
Five checkpoints, at the points where a wrong call would cost you a domain, a prospect or your reputation.
01 · Ideal customer profile
Agreed with you in writing before any list is built, including who is excluded.
02 · List sample
You see a sample of the list before launch and can strike any account or role.
03 · Every sequence
A person edits each AI draft; you approve the sequence before it sends.
04 · Reply handling
A person reads every reply and passes interested ones to you within one business day.
05 · Monthly report
Numbers are counted from the sending tool and your CRM, then checked by a person before you see them.
Models and tools
What the engine runs on
Models are used through their business APIs, from providers whose terms exclude training on the data sent to them. The current sub-processor list is in the Privacy Policy.
Your data
In the marketing service
- We process your prospects' business contact details as your processor, under a Data Processing Agreement sent with your proposal.
- Prospect and client data is never used to train or fine-tune a model.
- Your CRM, ad accounts and sending domains are registered to you; we work inside them.
- When a plan ends, prospect data is returned or deleted under the DPA. The suppression list of people who asked not to be contacted is kept, so they are never contacted again.
When we build automation
The commitments on our build tooling
Separate from the marketing service above: these cover the AI tooling we use to build and maintain the automation the engine runs on, and every custom build.
Your data does not go into our AI tooling
Machine assistance operates on workflow structure, code and documentation. Your live records, your customers' personal data and the contents of your systems are not fed into it. Automations are built and tested against sample or synthetic data until the logic is right.
Nothing you give us trains a model
Client data is never used to train or fine-tune AI models, by us or through us. This has been our position since before we described how we build, and describing it has not changed it.
A named engineer is accountable for the whole build
However a line was drafted, a person reviewed it, tested it and put their name to the handover. AI-assisted is not AI-decided, and there is no point in the process where nobody is answerable for what shipped.
The same rules apply to anyone we bring in
Contract engineers work to the same encoded standards and the same data boundary, under written agreements that include IP assignment. The standard does not loosen because capacity was added.
Straight answers
Is the outreach written by AI?
The first draft usually is. A person then edits it for accuracy and tone, checks every claim about the prospect, and you approve the sequence before anything sends. AI makes the research and drafting fast; it does not decide what goes out under your name.
Which AI models do you use?
Mainly OpenAI and Anthropic Claude, through their business APIs, orchestrated in n8n. We only use model providers whose API terms exclude training on the data sent to them. The current list of sub-processors is in the Privacy Policy and in the Data Processing Agreement sent with your proposal.
Is our data or our prospects' data used to train AI models?
No. Neither your data nor your prospects' data is used to train or fine-tune a model, by us or through us. When we run outbound for you we process prospects' business contact details as your processor, under a Data Processing Agreement.
Can we see and approve what goes out?
Yes, and before launch you must. You approve the ideal customer profile, a sample of the list and every sequence. Nothing sends from your domains until you have.
What happens when the AI gets something wrong?
That is what the checkpoints are for: a person reviews every draft, every list and every reply. If something incorrect still goes out, we tell you, correct it, and change the step that let it through.
See the checkpoints in a real plan
The how-it-works page walks through a plan from the first call to the monthly report, including every approval you give.