EU AI Act deadline · August 2, 2026

AI governance for
regulated IT functions

The platform that turns informal AI use into auditable AI operations — across NIST AI RMF, FDA CSA, SR 11-7, HIPAA and the EU AI Act. Use-case registry, risk evaluation, evidence trail and monitoring in one place.

Proprietary platform · Provider-agnostic · Evidence from day one · Co-founded with David Luquin
NIST AI RMF FDA CSA SR 11-7 HIPAA EU AI Act
your-domain.com/kairos/dashboard
Executive panel
Last 14 days · updated 5 min ago
Export14d ▾
Use cases+2 mo
12
AI systems+3 mo
24
High risk-1
3
Compliance+4pp
76%
ActivityRequests Compliance
Key metrics
Compliance gap76%
Fairness score0.87
Policies OK6/8
XAI reports9
Open incidents2
Use case "Clinical Assistant" classified High Risk2h
FRIA completed for use case "Submission Review"5h
SR 11-7 validation evidence captured · model v41d
5
Frameworks operationalized
6
Operating capabilities
3
Vertical editions
100%
Decisions with evidence trail
Why act now

Regulation, operational risk and AI in the shadows are the same conversation.

August 2026

High-risk AI systems must comply with the EU AI Act by August 2, 2026. The obligation reaches any organization doing business with European clients, regardless of where it is headquartered.

Up to €35M

EU AI Act penalties for non-compliance reach €35M or 7% of global annual turnover. SR 11-7, HIPAA and FDA CSA carry their own examination, audit and inspection exposure on top.

5 frameworks

NIST AI RMF, FDA CSA, SR 11-7, HIPAA and EU AI Act. They are not theoretical anchors — they are constraints that determine what an operator can produce in inspection, in court, in client review.

The problem Kairós solves

An IT function using AI without governance operates without an evidence trail.

The capability is already in the building. What is missing is the layer that makes that use visible, attributable, auditable and defensible. Four problems show up in every function we walk into.

01 · Shadow AI

Engineers use AI tools the organization can't see

Copilot, Cursor, Claude and ChatGPT are inside the workflow whether or not IT approved them. Productivity accrues to individuals, risk accrues to the company. The CIO has no map.

02 · Evidence gap

AI-assisted decisions can't be reconstructed

Which prompt produced which output, which human reviewed it, which decision followed. Without that trail the work is not defendable before an inspector, an auditor or a regulated client.

03 · Compliance exposure

The frameworks already apply

NIST AI RMF, FDA CSA, SR 11-7, HIPAA, EU AI Act. Operating without instrumented governance carries unbounded legal exposure, even before anyone enforces it.

04 · Drift

What works in pilot stops working in production

Prompts evolve, models update, engineers leave with their patterns, agents spawn sub-agents that wander from spec. What was governed in week one isn't governed in week thirty.

The organization either over-restricts AI and loses the gain, or runs unrestricted and carries the exposure alone. Kairós is the third option.

Six pillars · the platform

Six operating capabilities, integrated from week one.

Kairós is not a wrapper around LLMs. It is the layer that sits across them — registering what happens, attaching ownership, and producing the evidence that makes AI-assisted work defensible.

Capability 01

Use-case & agent registration

Every AI use case is registered with purpose, scope, risk classification and ownership. The CIO finally has a map of what is happening with AI inside the function.

Capability 02

Decision provenance capture

Every prompt, output, model version, retrieval source and human review logged with timestamp and identity. Decisions reconstructable end-to-end, weeks or years later.

Capability 03

Risk gates by use case

Each use case carries its gates: what needs human review, dual sign-off, or can run agentically with monitoring only. Enforced in operation, not added in audit.

Capability 04

Evidence trail for auditors

Inspection-ready evidence packages on demand: FDA submission support, HIPAA access logs, SR 11-7 model documentation, EU AI Act conformity files. A defendable artifact, not a dredged log.

Capability 05

Multi-agent orchestration governance

When workflows spawn sub-agents, Kairós tracks the hierarchy and attaches ownership at each level. The piece that makes agent swarms operable in regulated sectors.

Capability 06

Drift & continuous monitoring

Detects when prompts evolve, models update, or behavior shifts away from spec. Surfaces drift before it becomes an incident. Continuous, not point-in-time.

The companion · how it shows up in your org

Kairós doesn't replace your tools. It rides with them.

The analyst stays in their IDE (Cursor) and GitHub. The companion rides alongside, as an extension, surfacing notices from two autonomous agents. Three layers, lightest first: a silent badge, a tab opened at will, and a dashboard per level.

01 · In the flow

A silent badge

By default, just two counters in the IDE status bar: “G 1 · K 2”. Nothing opens, nothing interrupts. Zero friction over daily work.

02 · On demand

Two tabs at the base of the IDE

G for governance (mandatory) and K for knowledge (optional). K is autonomous: it spots the case and offers what's reusable without being asked. G interrupts only on hard controls; everything else is logged silently.

03 · When expanded

A dashboard per level

Personal for the analyst, group for the group lead, general for the board — the latter carrying the four metrics that test the thesis. The same Kairós substrate, aggregated by level.

Walk through it step by step
Open full screen ↗
Multi-framework from day one

Five regulatory frameworks, mapped to operating workflows.

Kairós does not interpret regulation; it operationalizes it. Each framework becomes specific gates, specific evidence requirements and specific audit-ready outputs.

NIST AI RMF
US base layer
Framework · 2023
FDA CSA
Life Sciences · GxP
Guidance · 2022
SR 11-7
Financial · model risk
Fed · 2011
HIPAA
Healthcare · PHI
Statute · 1996
EU AI Act
EU · cross-border
Regulation · 2024
NIST AI RMFUS base layer
What it requires
A function-level governance posture covering identification, measurement, management and mapping of AI risk. The voluntary baseline for any US organization with AI in operation.
How Kairós covers it
Use-case registry maps directly to the four AI RMF functions. Evidence trail produces the documentation NIST expects per use case, on demand.
FDA CSALife Sciences · GxP
What it requires
Computer Software Assurance for any AI-enhanced workflow producing artifacts that enter an FDA-regulated process. Validation evidence, change control, inspection-readiness.
How Kairós covers it
Vertical instance configured for protocol drafting, submission review, adverse-event detection and GxP validation. Inspection-ready evidence library out of the box.
SR 11-7Financial · model risk
What it requires
Federal Reserve guidance on model risk management: independent validation, ongoing performance monitoring, governance documentation for AI-assisted decisions in banking.
How Kairós covers it
Validation evidence captured per model and per use case. Continuous monitoring built in. Documentation produced on demand for OCC and Fed examination.
HIPAAHealthcare · PHI
What it requires
Protection of PHI at every AI touchpoint. Access logs, de-identification enforcement, audit-ready reports of who accessed what PHI through which AI tool.
How Kairós covers it
PHI boundary enforcement at the prompt level. Access logging by user, model and use case. Audit packages structured for HHS Office for Civil Rights inquiries.
EU AI ActEU · cross-border
What it requires
Risk classification, conformity assessments for high-risk applications, transparency obligations, fundamental-rights impact assessments. Mandatory for any organization doing business with European clients.
How Kairós covers it
Risk classification built into registration. Conformity files maintained continuously. Transparency artifacts produced for affected parties. EU representative reporting supported.
Kairós vs. the status quo

The difference between governed and ungoverned AI.

Most IT functions are running AI on logging, policy decks and good intentions. Here is what changes when governance is instrumented instead of recommended.

Property
KAIRÓS
Ungoverned AI
Use-case visibility

Every AI use case registered with owner, scope and risk class.

Mapped
Shadow AI
Decision provenance

Prompt, output, model, reviewer and decision reconstructable end-to-end.

Captured
Lost
Risk gates

Human review, dual sign-off or monitored-autonomy enforced in operation.

Enforced
Manual
Audit evidence

Inspection-ready packages per framework, produced on demand.

On demand
Dredged logs
Multi-agent accountability

Ownership tracked through every sub-agent in the hierarchy.

Traceable
Collapses
Drift detection

Continuous monitoring of prompt, model and behavior change.

Continuous
Undetected
Provider lock-in

Provider-agnostic layer; switching models doesn't break governance.

None
Per-vendor
Architecture

Where Kairós lives in your IT function.

Kairós sits between the AI tools your engineers use and the organizational layer that must govern them. It does not replace the tools, the engineers or your compliance function. It instruments what they do.

Above the tools

Provider-agnostic by design

Sits across OpenAI, Anthropic, Google and open-source models. No lock-in; the governance layer stays stable when underlying tools change.

Inside the workflow

Instruments use cases, not platforms

Engineers keep using Copilot, Cursor, ChatGPT and internal agents. Kairós captures use-case context, not keystrokes. Low friction, high coverage.

Beside the data

Tenant & boundary aware

Critical for IT-services firms across many clients. No prompt, output or learned pattern crosses tenant boundaries. SOC 2 passes with evidence, not assurances.

Under the organization

Ownership & accountability native

Every use case has a human owner. Every decision carries identity at each level. Even when swarms spawn sub-agents, the chain stays traceable.

Detailed technical specifications and product documentation live at kairos.luquin.com.

Inside a PRAGMAGROUP engagement

Kairós across the three engagement phases.

In a PRAGMAGROUP installation, Kairós is not bolted on at the end as a compliance afterthought. It is present from Discovery and instrumented through every phase.

Phase 1
2 weeks · fixed fee

Discovery

We map current AI use across the organization: highest-leverage workflows, governance gaps, applicable frameworks. Surfaces the use cases that become Kairós registered entries.

Kairós role
Governance gap map. Use-case candidates for registration.
Phase 2
1 week · signed criteria

Roadmap

A six-month installation plan against Discovery priorities. Case selection, signed binary criteria, gates per case, instance scoped. Approved by your executive committee before kickoff.

Kairós role
Instance scoped. Gates defined per use case. Evidence templates calibrated to sector.
Phase 3
180 days · binary outcome

Installation

Operational execution under signed criteria. Kairós installed against the Phase 2 gates; your team enters the operating practice from week one. By the end, the function operates under governance whether we stay or leave.

Kairós role
Live instance capturing evidence from day one. Owned by your team at handover.
Agent swarms

The piece that makes agentic workflows operable in regulated sectors.

Multi-agent architectures — one human directing a hierarchy of agents that create sub-agents — are already in production at limited scale. The capability is here. What is missing in every deployment is the governance layer.

Problem 01

Accountability collapses

A sub-sub-agent drafting protocol language, flagging an adverse event or committing code to a banking client cannot be the accountable party. Without a granular trail, accountability becomes unenforceable.

Problem 02

Auditability fails

HIPAA, FDA CSA, SR 11-7 and EU AI Act all expect a verifiable chain of evidence. A swarm without instrumented governance produces output the auditor cannot trace.

Problem 03

Drift goes undetected

Sub-agents creating sub-agents over months wander from spec. A traditional company solves drift through culture and reviews; a swarm has none of that by default. It has to be engineered in.

The swarm future the market describes is real. What it leaves unsaid is that no version of it works in regulated sectors without something like Kairós underneath. That is the position we occupy.

Vertical instances

Configured per sector, not generic.

Each vertical arrives with its use cases preconfigured, its framework mapping done and its evidence templates calibrated to that sector's regulatory reality. Same motor; the wrapper speaks the buyer's world.

Vertical 01 · Available

Kairós Life Sciences Edition

For CROs, biotechs, regional health systems and pharma IT functions. Preconfigured for protocol drafting, submission review, adverse-event detection, clinical-trial documentation, GxP validation and patient-facing workflows.

FDA CSA · HIPAA · ICH GCP
Vertical 02 · Available

Kairós for IT Services Companies

For software houses, integrators and MSPs in the 50–500 range. Preconfigured for developer-productivity governance, code-review traceability, client-deliverable provenance and multi-tenant segregation.

NIST AI RMF · SOC 2 · sectoral overlays
Vertical 03 · In development

Kairós for Manufacturing

For industrial IT functions, OT/IT boundary cases and manufacturing-adjacent IT services. Use cases under development around predictive maintenance, QC vision systems and ERP-AI integration.

NIST AI RMF · sector-specific overlays
What Kairós is not

Boundaries set by design.

Kairós does what the name says: it instruments AI governance. It does not pretend to do four things the market sometimes asks of it.

Not

An AI firewall or content filter

Kairós does not block prompts on content rules in real time. That is a different category. Kairós captures, attributes, governs, audits.

Not

A wrapper around LLMs

Provider-agnostic. It works above whichever models your engineers use. Switching models does not break the governance layer.

Not

Logging-as-a-service

Logs are an output, not the product. Kairós registers use cases, attaches ownership, enforces gates, detects drift and produces evidence.

Not

A replacement for compliance

Kairós instruments compliance; it does not eliminate compliance officers. Human accountability stays human. What it removes is the gap between policy and operation.

Two ways forward

Don't wait for the inspection to start governing your AI.

Kairós is installed through a PRAGMAGROUP engagement. It starts with a two-week Discovery: governance gap map, signed criteria, no further commitment. Or look at the product directly first.