Solutions · Operating AI

Keep every AI deployment safe, reliable and compliant after go-live.

Vendors update their models, the business extends the use, regulations change. Each change can invalidate the basis on which the deployment went live. AIQURIS shows which controls a change affects, re-assesses at the depth it calls for, and turns monitoring signals into decisions.

AI in operation · Changes since go-live
Customer service AI assistantLive since March · on a vendor model
WHAT CHANGEDWHAT IT MEANS
APR
Vendor model update
Re-assessment. 3 controls depend on the model.
JUN
Use extended to customers in Germany
New controls. EU AI Act and GDPR requirements added, 6 controls.
AUG
Complaints about refund promises above threshold
Owner decision. Refund questions routed to human agents.
SEP
New regulator guidance on AI chatbots
No action. Existing controls already meet it.
Every change is linked to the controls it affects and the decision taken.
The challenge

Go-live is one decision. The AI keeps changing.

01

The assessment goes out of date

The assessment that cleared the deployment describes it on the day it went live. Within months, the model, the use or the rules have moved on.

02

Changes come from every side

Vendors update their models without asking. The business extends the use to new users and countries. Regulations and standards change. None of it waits for the next review.

03

Monitoring without a decision

Dashboards show drift, errors and complaints. They do not say whether a control is now missing, or who has to act.

Control does not end at go-live.

Five changes after go-live

Every change gets the assessment it calls for.

A minor change gets a check. A major one gets a full re-assessment. Each one ends in a documented decision.

Vendor update

The model behind the deployment changes. AIQURIS shows which controls depend on it and what must be assessed again.

Extended use

New users, countries or purposes. AIQURIS adds the risks and requirements that come with them.

New regulation or standard

A regulation or standard is published or amended. AIQURIS shows which of your deployments it affects.

Incident

Something goes wrong in operation. AIQURIS traces it to the risk, the control that should have prevented it, and its owner.

Retirement

The deployment ends. AIQURIS sets out what must happen to its data, its users and the processes that relied on it.

You record changes to your deployment. AIQURIS tracks regulations and standards, and works out what every change means.

Monitoring

Monitoring that leads to a decision.

Monitoring tools collect signals. They do not know which risks matter for this deployment, what level is acceptable, or who acts. AIQURIS works out what each deployment needs monitored, from its actual risks, and sets the thresholds with your accountable team. When a threshold is breached, it shows which controls and requirements are affected, and who decides.

Every threshold has an owner. Every breach ends in a decision.

1
Define what to monitorAIQURIS
The signals each risk calls for, and the risks monitoring cannot cover.
2
Set the thresholdsAIQURIS, with your team
Agreed with your accountable team, with a named owner for each.
3
Collect the signalsYour tools
Your monitoring tools and processes collect them.
4
Act on a breachYou, with AIQURIS
AIQURIS shows which controls and requirements are affected. The owner decides.
Traced, not templated

A known risk, caught in operation.

The risk was identified at go-live, rated moderate and accepted with two controls: an output filter and a complaint threshold. In operation, the threshold caught what testing could not.

INCIDENTComplaints about refund promises cross the threshold set at go-live
RISKCommitments the AI assistant must not make · identified at go-live, rated moderate
REQUIREMENT · SOURCENo unauthorised commitments to customers · your customer service policy
RE-ASSESSMENTSeverity raised from moderate to high. The output filter is not enough: refund answers must now be checked against the refund policy.
OWNERThe customer service lead routes refund questions to human agents until that control is in place.

Every change and incident is traced this way.

What you get

One record per deployment, from go-live to retirement.

  • ✓Every deployment in one AI register, with its status and what needs attention
  • ✓The controls each change affects, linked to their requirements and sources
  • ✓Every incident traced to its risk, its control and its owner
  • ✓Re-assessment at the depth the change calls for
  • ✓What to monitor, the thresholds, and who acts on a breach
  • ✓Updates when regulations and standards change
  • ✓A documented decision for every material change
Start free

Your first impact screening (IMPACT+) is free.

Describe one live deployment. See who could be affected today, how badly, and whether its last assessment still goes deep enough. No sales call needed.

  • ✓Impact tier for the deployment as it runs today
  • ✓Affected stakeholders and potential harms
  • ✓Whether a deeper re-assessment is due
  • ✓The deployment registered in your AI register
Start your free screening
Get started

Bring one live deployment. See what has changed since go-live.

In one call we look at one deployment in operation: what has changed, which controls those changes affect, and what to monitor from now on.

Privacy Overview

We use cookies to operate our website, ensure its proper functioning, improve performance, analyse traffic, and support our marketing activities. Some cookies are strictly necessary and cannot be disabled. Others can be enabled or disabled below according to your preferences. For full details, please see our Privacy Policy.