Governance

We bring the governance, not just the agents.

Most agent projects don't fail on technology. They fail because nobody knows who may approve what, what it costs, and what to do when it goes wrong. Every pilot leaves with this solved — in documents the client adopts and in rules the platform enforces.

21%

of companies have mature agent governance

Deloitte, State of AI in the Enterprise 2026

> 40%

of agentic projects will be cancelled by 2027 — inadequate risk controls is a named cause

Gartner, June 2025

1 · Policy pack

An adopted governance policy, not a forgotten PDF.

Included in every pilot (S8). Also available standalone for those who already have agents installed elsewhere, without governance.

The client leaves the pilot with an agent governance policy approved by management, answering:

  • Who may hire and retire agents — and who approves each hire.
  • Approval tiers per action type — automatic, one tap, two admins, prohibited.
  • Prohibited actions — for example: issuing or altering fiscal records, evaluating people's performance, emotion recognition.
  • Incident procedure — who stops what, how it is communicated, what is logged.
  • Worker information — the duty under EU AI Act Art. 26 and Portugal's Lei 13/2023, handled with a ready notice.
Sample

Sample governance policy (redacted)

A real policy, names removed, to see the level of detail before talking to us. We send it by email — no form, no marketing sequence.

Request by email →

2 · Approval matrix

Designed per client. Enforced by the platform.

During the audit we classify every action of every agent. The matrix doesn't live in a folder: it lives in the Approve / Reject cards the platform sends. A rule people can't forget.

ActionTierNote
Read tasks, documents and the shared mailbox auto Within the charter's scope
Remind an internal colleague by DM auto Logged; no new content leaves the company
Email a client or supplier one tap Area manager approves the text
Grant a user access to an agent one tap The agent's admin
Hire or retire an agent one tap Chief of Staff proposes; manager approves
Alter a posting or fiscal record prohibited Never: the certified accountant keeps responsibility
Access health or sensitive HR data two admins With access notification (Lei 58/2019 Art. 29)
Change an agent's model or limits two admins IT + area lead
Quote Builder IA
The proposal for Clínica Sol is ready (PDF). Per the matrix, sending to clients needs one tap from the sales manager.
Send proposal: Clínica Sol
To
geral@clinicasol.pt
Value
€4,860 + VAT
Tier
one tap — sales manager
Approve Reject
Illustrative excerpt: the “email a client” row of the matrix, happening.

3 · Operating rhythm

Governance that runs itself.

A policy is only worth something if someone executes it. The Chief of Staff's daily brief, the monthly review with us and the audit logs form a documented, repeatable oversight process the client can show an auditor.

Chief of Staff IA

Daily brief — Wednesday

  • Activity: 58 tasks by 4 agents; 3 escalated to humans.
  • Costs: €3.60 yesterday; €41 this month; seat 2 at 88%.
  • Refused access: 1 (external@… tried to DM Quote Builder).
  • Pending decisions: 2 — proposal send (Pedro), HR access (Ana + João).
  • Health: backup 03:00 ok; no model being deprecated.
Illustrative daily brief. The format is fixed; the numbers come from the logs, not from the agent's memory.

Daily

Chief of Staff's brief in the management channel: activity, costs, refused access, pending decisions, installation health.

Monthly

Review with us: what the agents did, cost per task, seat usage, incidents, what to hire next. Minuted.

Always

Audit and usage logs, retained ≥ 6 months. Auditable by construction.

4 · Compliance records

What we prepare. What your DPO signs.

We don't replace the data protection officer or the certified accountant. We hand them the work done, to review and sign.

RecordWe prepareThe client decides / signs
Impact assessment (DPIA)Template filled in during the audit: data, purposes, risks, measures.The DPO reviews, adjusts and signs.
Register of AI systemsInventory of agents, roles, models, data processed, risk classification.Management approves the classification.
Log-retention statementRetention ≥ 6 months, minimisation, reconciliation with GDPR.The DPO validates.
Worker information (Lei 13/2023)Notice template for workers, works council and union delegates.Management delivers it before go-live.
Approval matrixClassification of every action; configured in the platform.Area leads approve the tiers.
Subprocessor list and DPAEnlogical's DPA and the model providers' DPAs.Management signs.

Already using AI?

A governance review of the AI you already use.

70% of adults use generative AI personally — and many paste company data into free chatbots. A governance review (S8 standalone) maps what is happening, what the law requires and how to bring it in-house.

Book a governance review of your current AI use Back to the Trust Center

Pilot in 4 weeks

Four weeks, a fixed price, an agreed KPI. At the end, the first coworker is in Teams doing real work.

A 20-minute call, a real agent at work, no slides.