How Graft handles your data

Graft reads your sales conversations to build your pipeline. That’s the only thing it does with them. Here’s what it reads, what it never does, and who owns what.

NDPA 2023

Nigerian law, NDPC regulated

Encrypted

In transit and at rest

Yours to take

Export or delete anytime

You set the scope

Exclude any domain any time.

Our compliance as it stands

Graft is new, so here’s exactly where it stands. Every requirement, including the ones we don’t hold yet.

NDPA 2023

In progress

Built to the Act as much as we could from launch. Registration with the NDPC follows once we're processing data at the scale the Act sets.

NDPC compliance audit return

Not due yet

Filed every year through a licensed DPCO, once registered.

Data processing agreement

Included

Part of our terms, with Graft as the processor and your company as the controller.

Privacy policy

Published

Covers our lawful bases and how AI processes your data.

Data residency in Nigeria

Not offered

Some processing happens outside Nigeria, including by the AI providers Graft runs on. The Act allows that with the right transfer terms, which sit in our vendor contracts.

How Graft actually reads your data

Graft looks at every thread once. Only the sales ones get extracted. Everything else is dropped before any of it is pulled apart.

What comes in

Your work mail, minus any domains you exclude and whatever WhatsApp messages forwarded or pasted to the Graft bot.

Every mail thread meets one question

Is this sales work? Asked once, before anything else happens.

Where it goes

Internal threads, invoices, calendar invites, personal mails and newsletters are discarded immediately. Deals, contacts, tasks and signals are extracted and stored.

Our security practices

Permissions live in the database

Every deal, contact and message carries an owner and a visibility scope, enforced on the data itself, so the rule holds however the data is reached.

The AI only receives your rows

Retrieval is filtered by your role and id before anything reaches the model, so it is never handed the whole team's data and asked to be discreet

Your inbox stays yours

Raw mail is only read when you're the one asking, or when a message is on a shared deal. A manager's view is built from deal records.

Built for privacy

Graft processes your threads to keep your pipeline up to date, not to expose your inbox. Production access is restricted and logged. Your data is never sold.

Encrypted, with keys kept apart

In transit and at rest, backups included. The tokens that let Graft read your inbox sit under their own key.

Leave whenever

Export anytime. Disconnect or delete and access ends at once, with the data gone from Graft and the AI's working context within 30 days.

Questions about your data

Graft may use deal data to improve the models that read your pipeline. You can switch that off for your whole workspace in settings, and opting out changes nothing about how Graft works for you.

Not yet. Registration, and the yearly audit return filed through a licensed DPCO, apply once a company is processing personal data at the scale the Act sets, so they come after launch, and the groundwork is in progress. Until then we build as closely as we can to the Act's rules.

No. Prompts and Graft's outputs aren't retained by our providers after the response comes back, and nothing is used to train their models.

It's classified at the first step and dropped. Nothing is extracted from it and nothing is kept.

We use relevant professional data for B2B prospecting, relying on legitimate interest where appropriate. You can request removal, and our suppression list prevents re-importing your details across customer accounts.