apertix

We build AI you can open up

Apertix designs and builds AI systems, and leaves a record of what each decision was based on, in a form an auditor can check. The same team does the building and the reviewing.

Kept for every AI decision

  • Where the input came from
  • Documents it relied on
  • Model and prompt version
  • Policy check result
  • Who approved it
f/1.4

Move the aperture. Closed, it is a black box. Open, you can read the decision record.

What we do

Wherever you start, you end up in the same place: AI whose behavior you can check afterwards.

AI governance assessment

We find out where AI is used in your organization, who is accountable for it, and which controls are missing. We compare your current state, item by item, with what regulations and standards require.

  • AI system inventory with risk tiers
  • Missing controls, in priority order
  • Mapping to Korea's AI Basic Act, the EU AI Act, ISO/IEC 42001 and NIST AI RMF

AI built with controls inside

We design and build AI agents and workflow automation. Permissions, logs and human approval steps go into the structure from the start, not on top of it at the end.

  • A system that runs in production
  • A record of the basis for every decision
  • Operating and change management procedures

AI review and testing

We examine AI that is already running, as an outside party. The evidence is organized so you can use it as is for internal audit, external audit and regulators.

  • Test results and supporting evidence
  • Findings and recommendations
  • A report for management and the audit committee

We do not take on independent review of systems we built ourselves. Reviewing your own work is not a review.

What happens when builders and reviewers are different teams

The development team focuses on shipping, and controls get added just before launch. The audit team arrives after everything is built and finds that no record was kept. In between, the AI runs without anyone being able to explain it.

Apertix starts design with the auditor's questions. The answers have to live in the system, not in a document, if you want to open it up later.

What we ask on the first day of design

  • Where is the basis for this decision recorded?
  • Who approved it, and how do we prove that?
  • When the model or prompt changes, who knows?
  • When it is wrong, who stops it, and how?

How we work

  1. Open it up

    We map where and how you use AI today, and where the risk sits.

  2. Design

    We set the control requirements and the system structure together, in one design.

  3. Build

    We build a working system that records its decisions automatically.

  4. Test and hand over

    We hand the test evidence and operating procedures to the team that will run it.

Company

apertus
Latin, adjective
Open, uncovered, unconcealed. It shares a root with aperture, the opening in a camera lens.

That is where the name Apertix comes from. An image only forms when light gets in, and we think AI can only be trusted when you can see inside it.

So the systems we build keep a record from the start, and when we review a system, the first thing we check is whether that record exists.

Tell us where your AI stands today

It does not matter whether you are still evaluating or already in production. Write to us about what you use AI for and what worries you, and we will reply by email.