Founding offer for our first 20 clients: launch prices locked for 6 months and 50% off setup.Join the programme
Security

Control is a design decision, not a setting.

This page is maintained by WorkForge to answer common security and privacy questions about the platform. It describes current practices, not an independent audit.

Access control

An AI employee is given named accounts and scoped permissions for its role. Access is granted by your business and can be withdrawn at any time. It does not receive blanket access to your systems.

Approved knowledge

Answers are drawn from the documents, policies and templates you approve. Where information is missing or contradictory, the employee is instructed to ask rather than guess.

Human approvals

You define which actions require sign-off. Discounts, refunds, complaints, contractual language and payment matters are typical approval points.

Activity logging

Tasks, drafts, sent messages and escalations are recorded so you can review what was done and why.

Privacy and data practices

WorkForge is built for POPIA. We collect the information required to run the role and no more, and we agree data handling in writing before service begins. WorkForge does not hold ISO 27001, SOC 2 or any POPIA certification.

Responsible AI

An AI employee should never make an unauthorised promise, discount, payment, booking change or contractual commitment. Uncertainty is escalated, not improvised. Our assistants introduce themselves by name and your business, mention call recording, and never pretend to be human if asked.

Human control

Your business stays in control.

WorkForge employees only use the tools, information and permissions assigned to their role. Sensitive or uncertain matters can be routed to a human before any action is taken.

  • Role-based access

    Each employee is given only the accounts and systems its role requires.

  • Approved knowledge sources

    Answers come from documents and policies your business has approved.

  • Human approval rules

    You decide which actions must be reviewed by a person before they happen.

  • Activity logs

    Every task, reply and escalation is recorded and reviewable.

  • Controlled integrations

    Connections are scoped to specific mailboxes, folders and systems.

  • Escalation for uncertainty

    When information is missing or conflicting, the task goes to a person.

An AI employee should never make an unauthorised promise, discount, payment, booking change or contractual commitment.

Built for POPIA.

Discuss your security requirements
Access on your terms

Passwords and data: how we keep you safe

  • We never ask for passwords on WhatsApp or email.

  • For your existing Microsoft 365, you approve access on Microsoft’s own sign-in page. You can remove that access at any time.

  • If we create a mailbox for your AI employee, you get the login once through a secure one-time link. You change the password and turn on two-step sign-in.

  • Your AI employee only sees the mailboxes, calendars and folders you choose.

  • Each client has a private server. Nothing is shared between clients.

  • Connections are encrypted. We use business AI services set not to train on your data.

  • We sign a POPIA data processing agreement with you.

  • If you cancel, we export your data on request and delete our copies within 30 days.

Your next employee is one WhatsApp away.

Live in 48 hours for standard set-ups. You get your time back.

Start your free trial

Prefer to talk first? Book a call

Founding offer for our first 20 clients: launch prices locked for 6 months and 50% off setup.