XceedNexus

Agents you can actually deploy

The control layer for every AI agent in Xceed365HR: what each one can see, do and approve.

How it works

Agents prepare. People approve.

  1. 1A request

    Someone asks in Microsoft Teams, on the web or on their phone.

  2. 2An agent prepares

    The right agent does the work, across suites, with only that person’s permissions.

  3. 3A person approves

    Anything consequential, like a payroll run, waits for a named approver.

  4. 4It’s recorded

    Every step is logged permanently, ready for an auditor.

Five guarantees

Control you can verify

Every agent in the core and add-on suites runs through XceedNexus. None operates outside it.

Acts only as you

An agent carries the asker’s permissions and nothing more: no credentials of its own, no direct database access.

A person approves what matters

Payroll and other consequential actions wait for a named approver. An agent’s mistake is a proposal someone declines, never a payment already made.

Every action is recorded

Each step is logged permanently, so any decision can be traced for an auditor, a regulator or a dispute.

Your data stays in your region

Agents run inside your regional data centre in Nigeria, Kenya, South Africa or the UAE.

Never trains on your data

Nothing is kept after a task, and no customer data trains any model.

Security and trust

For your security team

The technical detail behind each guarantee: every claim is a control, at a stated layer, that an auditor can test.

Read the technical detail
  • Agents do not store your data, and no customer data trains any model. Agents retain nothing after completing an instruction. Models run inside our own environment, process only the payload required to carry out the instruction, and retain nothing once the response is returned.
  • Agents act only on instruction, through the product's own API. An agent holds no credentials and no standing database access. It calls the same documented API a human calls, under the identity of the person who asked, inheriting that person's permissions and the same row-level policy. It cannot query the database directly or widen its own scope.
  • Isolation is enforced by the database, not by application code. The platform is multi-tenant with logical separation, deployed as regional shards in Nigeria, Kenya, South Africa and the UAE, with your data resident in your region's shard. Row-level security policies in PostgreSQL bind to a tenant context taken from the authenticated principal, database roles are least privilege so no application path can bypass them, and cross-tenant access tests run on every build. Encryption is applied in transit and at rest, with per-tenant keys.
  • Consequential actions require a person. Agents prepare and recommend. Actions with real consequence, a payroll run being the clearest case, stop at an approval gate until a named human accepts them. When an agent is wrong, it is wrong in a proposal a person reviews, not in a payment already made.
  • Everything is recorded, permanently. Every agent action is logged with its tenant context and retained indefinitely, so any decision can be reconstructed long after the fact.

Certifications and evidence

ISO 27001:2022 Certified Company SOC 2 Type II Certified GDPR Compliant

We hold SOC 2 Type II and ISO 27001:2022 certification, comply with GDPR, and publish an AI Security Policy and a Trust Center. Our SOC 2 Type II report and penetration test summary are available under NDA.

See XceedNexus running your payroll

Bring your security team. We will walk the controls, the logs and the approval gates in the product, not in a slide.