Skip to content
AgentIQ AgentIQ
Trust and governance

AI that shows its evidence, boundaries and actions.

AgentIQ keeps AI-generated interpretation visibly distinct from deterministic results — and records how each recommendation was produced.

Explore a design partnershipSee how the decision is made
Agent investigation record

Every investigation leaves a record like this.

Evidence with its source. Signals with their confidence. The hypothesis, and what argues against it. The rule version that applied, and who still has to approve. This is the artefact, not a summary of it.

INV-2417-8ac9 Account Northstar Labs Created 12 Mar 2026, 08:14 CET

Source evidence

Weekly active users down 34% across the last six weeks

Product usage data · 10 Mar 2026

Three unresolved P2 tickets on the reporting module, oldest open 19 days

Support data · 09 Mar 2026

Renewal date 62 days out; no expansion discussion logged

Commercial data · 01 Mar 2026

Executive sponsor has not attended the last two QBRs

Account notes · 24 Feb 2026
View all evidence (12) →

AI-derived signals

Adoption decline concentrated in one team, not account-wide High confidence
Support friction is blocking a core reporting workflow Medium confidence
Sponsor disengagement may indicate an internal ownership change Low confidence
How signals are derived →

Risk hypothesis

Renewal is at risk because the reporting workflow that justified the original purchase has degraded, and the sponsor who championed it is no longer engaged.

Medium confidence

Why this hypothesis →

Contradictory evidence

  • Seat count is unchanged and two new users were provisioned last month.

  • Invoices are paid on time, with no billing disputes on record.

  • A second team increased usage 18% over the same period.

Governance metadata

Rule version Playbook rules · published rev 42
Configuration tenant-config rev 118
Governed status Policy applied
Human approval Pending — required before action
Next review 19 Mar 2026

Action history

  1. Investigation created 12 Mar 2026, 08:14
  2. Evidence collected from 4 sources 12 Mar 2026, 08:14
  3. Policy applied — routed to renewal-risk playbook 12 Mar 2026, 08:15
  4. Recommendation generated 12 Mar 2026, 08:15
  5. Awaiting human approval No action taken until approved

Illustrative record. The account, the evidence and the source categories are fictional; the structure, the confidence levels, the version stamping and the approval gate are the ones the platform produces.

Six properties we hold to

These are architectural commitments, not settings. They are true of every investigation the platform runs.

Evidence grounding

Every conclusion carries the source records it was drawn from, with the system and date they came from. No claim without evidence behind it.

Deterministic control

Policies and rules determine what can happen. AI recommends; configured logic decides.

Human authority

Actions that touch a customer wait for a person. The approval gate is part of the pipeline, not a preference someone can switch off.

End-to-end audit

Investigation, applied policy, recommendation and approval are recorded as events, stamped with the rule and configuration versions in force.

Stated data quality

Every week of data is graded before anything is calculated from it. Where the inputs do not support a defensible score, the platform says so and routes the account for review, instead of producing a confident number on weak evidence.

Enforced reflection

The system cannot act again on the same account until someone records what was done about the last recommendation and what came of it. Acting is rate-limited by learning, not by a quota.

AI investigates and recommends. Configured policy controls what happens next.

Where your data lives, and what we do with it

We are a European company building for European B2B SaaS. That shapes the architecture, not just the paperwork.

EU data residency

The platform, your data and the deterministic engine run in the EU. Where the AI layer is used, the account context is sent to a model provider for inference, and that step is disclosed, region-configurable and covered by the data processing agreement — including the option to run without the AI layer at all. Deployment and sub-processor detail is agreed per engagement.

GDPR by construction

Tenant isolation, scoped access and a documented processing basis for each connected source. Data minimisation is a design constraint: we connect the sources a decision needs, not everything available.

No model training on customer data

We do not train models on customer data, and we do not pool one tenant's data into another's results. The domain model is ours; your data stays yours. That commitment is contractual with our model provider, not only our own policy.

Security and privacy controls will be documented against the implemented product and deployment model. We hold no certifications today and do not imply any.

Bring us a real Customer Success decision.

The fastest way to judge any of this is to point it at an account you already have an opinion about, and compare.

Explore a design partnershipSee how the decision is made