AI · Responsible AI
What we commit to, and what we do not claim.
AI on Spark ERP runs inside the same tenant isolation, permissions and audit trail as the rest of the platform. Here is what that means in practice, and where the boundaries honestly are.
Our commitments
Tenant data isolation
AI features operate within the same per-tenant data isolation as the rest of the platform — one tenant's data is never used to answer another tenant's query.
No training on customer data without consent
We do not use a tenant's data to train or fine-tune AI models without that tenant's explicit consent. This is a commitment we hold ourselves to, not a certification.
Human-in-the-loop for consequential actions
AI Agents create drafts, recommendations and follow-up tasks for a person to review — a consequential action (a payment, an approval, a send) is carried out by a human through the platform's normal workflow, not silently by an agent.
Transparency about AI vs. deterministic logic
The platform is explicit, in its own documentation and configuration, about which behavior comes from AI and which comes from deterministic rule-based logic in the Workflow Engine — the two are not blended in a way that hides which one made a decision.
What we do not claim
We have not obtained a third-party AI-specific audit or certification, and we do not display one. Our broader compliance posture — including the frameworks Spark ERP is designed for and the ones we have not yet certified against — is stated plainly on the Compliance page, with the same discipline applied here: no framework is claimed until it is actually obtained, with evidence.
Read the full security and compliance posture.
Tenant isolation, encryption and the compliance frameworks Spark ERP is designed for — stated honestly, not marketed.