Your data is yours. Full stop.

Alfred connects to your marketing stack read-only, synthesises what matters inside your own workspace, and delivers decision-ready intelligence — without taking control of your accounts, and without your data ever training anyone else's Alfred. This pack explains exactly how, in plain language.

Read-only by default

Alfred connects over secure read-only APIs. It can see your data; it cannot change it. Write-back is off until you explicitly authorise it, per integration.

Encrypted throughout

Every connection and every synced record is encrypted in transit and at rest. Credentials and OAuth tokens are stored encrypted and are revocable at any time.

Everything is logged

Every read, recommendation and authorised write lands in a full audit trail — who, what, when — exportable for your own review at any time.

No cross-customer training

Your workspace is isolated. Your data, your patterns and your outcomes are never used to train models for any other customer.

Where we are on certifications — the honest answer

SOC 2 Type II is in progress: the audit window is underway and the report is expected [DATE — confirm with security before sending]. A GDPR-aligned data-processing agreement is available on request. ISO 27001 is on the roadmap and not yet certified. Alfred does not display badges for certifications it has not earned — if a claim is not in this pack, do not make it.

In this pack

How your data moves through Alfred — connect, synthesise, deliver
Where your data lives, and for how long
Access controls: roles, audit trail, retention and deletion
Subprocessors and frequently asked questions

Security pack · data flow

The data-flow narrative

How your data moves through Alfred

Three stages, one direction of trust: Alfred reads, reasons, and reports back. Nothing changes in your accounts unless you authorise it.

Connect — read-only

You connect your stack over secure read-only API and OAuth integrations — ad platforms, analytics, CRM, billing. Typical setup is under an hour with zero engineering, syncing up to a year of history. Alfred can see; it cannot touch.

Synthesise — in isolation

Synced data is normalised into your workspace's isolated store, where Alfred detects what changed, explains why, and recommends what to do next. Every recommendation carries its sources and a confidence score — never a black-box answer.

Deliver — and only act if asked

Briefs and alerts arrive in the app, email and Slack. If you enable write-back, Alfred acts in your tools only after explicit per-integration authorisation and an approval step — and every action is logged.

Where your data lives

Data category	What it includes	Where it lives	Retention
Connected-platform data	Campaign, spend, performance and pipeline records synced read-only	Encrypted at rest in Alfred's cloud infrastructure, [REGION — confirm current hosting region]	While your account is active; deleted within 30 days of termination
Workspace content	Briefs, recommendations, decisions, annotations, audit log	Your isolated workspace store — never pooled across customers	While active; exportable any time; deleted within 30 days of termination
Credentials	OAuth tokens and API keys for your integrations	Encrypted secret storage; never logged in plaintext	Revocable instantly — by you in Alfred, or at the platform side
Account & billing	Users, roles, invoices	Alfred's application database and payment subprocessor	As required for legal, tax and accounting obligations

Subprocessors

The current subprocessor list is maintained by our security team and provided with the DPA. Confirm this table is current before sending.

Subprocessor	Purpose	Region	DPA status
[CLOUD PROVIDER]	Hosting & encrypted storage	[REGION]	[SIGNED / DATE]
[MODEL PROVIDER]	Model inference (no training on customer data)	[REGION]	[SIGNED / DATE]
[EMAIL PROVIDER]	Brief & alert delivery	[REGION]	[SIGNED / DATE]
[PAYMENTS PROVIDER]	Billing	[REGION]	[SIGNED / DATE]

Security pack · access & FAQ

Access controls

Who can see what — and the record of it

Role-based access control

RBAC is available on the Max and Enterprise plans. People see and do only what their role permits — nothing further.

Role	Can do
Admin	Manage integrations, users, write-back authorisation and billing
Member	Read briefs, ask Alfred, propose and approve actions within their scope
Viewer	Read-only access to briefs and the KPI cockpit

Audit trail

Every read, recommendation and authorised write is logged with who, what and when. The trail is exportable for review at any time — by you, without asking us.

Retention & deletion

Alfred syncs up to a year of history to build context. When you leave, you can export your workspace, and all customer data is deleted within 30 days of termination. Deletion earlier than that is available on request.

Frequently asked questions

Does Alfred train models on our data?: No. Your workspace is isolated, and your data is never used to train models for any other customer. Alfred's memory of your business compounds for you alone.
Can Alfred change our campaigns without us knowing?: No. Connections are read-only by default. Write-back requires explicit per-integration authorisation plus an approval step, and every action is recorded in the audit trail.
What certifications does Alfred hold today?: SOC 2 Type II is in progress — the report is expected [DATE — confirm before sending]. A GDPR-aligned DPA is available on request. We will not claim a certification before it is earned.
Who at E902 AI Labs can access our data?: Access is restricted to named engineers for support and reliability, under least-privilege controls, and that access is itself logged.
How do we off-board?: Export your workspace, revoke the integrations (from either side), and terminate. Data is deleted within 30 days; written confirmation is provided on request.
Where can we send security questionnaires?: hello@seekalfred.ai — we respond to standard questionnaires (CAIQ, SIG Lite) as part of every Enterprise evaluation.

Prepared by E902 AI Labs Private Limited. This pack describes current architecture and practice; it is not a contractual commitment — contractual terms live in the MSA and DPA. Bracketed values must be confirmed with the security team before this document is shared externally.