Airia

Its flows, and where AI agents are in them

How work moves through Airia's product, as its public pages describe it, and the steps where an AI agent acts (2 steps with an AI agent in it, marked AI agent here). Each step cites its source.

An enterprise governs every AI agent it runs, approved or not

As Airia's home page describes it: once deployed, the same controls apply to agents that were purpose-built, commercially adopted or never approved. Its page says the product views it shows are illustrations with sample data. (airia.com/, read 2026-11-06)

1. Airia: Discovers every AI tool, model, agent and MCP server running across the organisation, including those nobody approved and personal AI subscriptions employees connect to corporate systems. (airia.com/, read 2026-11-06)
2. The security team: Sees what was found (for example people using AI apps that were never onboarded) and is given the controls to act on it. (airia.com/, read 2026-11-06)
3. An AI agent: Attempts an action: a tool call, sending an email, a database query. AI agent here: Acts inside the enterprise's systems; Airia's page puts it as "agents don't generate, they act". (airia.com/, read 2026-11-06)
4. Airia: Enforces policy at the execution layer before the action happens; guardrails inspect every action at runtime, and unauthorised agent actions are stopped (its example: access to customer records approved only when policy checks pass and an audit trail is active). (airia.com/, read 2026-11-06)
5. Airia: Keeps one inventory of every agent, model and tool in use, with risk classification, human approvals and audit evidence, and generates governance documentation mapped to the EU AI Act, NIST AI RMF, SR 11-7, HIPAA, ISO 42001 and SOC 2. (airia.com/, read 2026-11-06)

Teams build and launch agents through one governed gateway

The building side of the same platform, in the five layers its home page lists. (airia.com/, read 2026-11-06)

1. The enterprise: Connects the systems, warehouses and sources its agents act on, with permissions preserved. (airia.com/, read 2026-11-06)
2. Airia: Routes every request through one governed gateway of approved models and tools, each tested and validated before it reaches production. (airia.com/, read 2026-11-06)
3. Business and technical teams: Design, orchestrate and launch agents together, with policy controls configured before the agents reach production. AI agent here: The agents the teams build run across any model, framework and data source, inside those controls. (airia.com/, read 2026-11-06)
4. Airia: Inspects the agents' actions at runtime and keeps sensitive data protected across every connected system. (airia.com/, read 2026-11-06)

Our question for them

As for every company we follow, three questions, not claims. We have not asked Airia yet. When they answer, it goes here.

1. Reach What can their AI agents (or their customers' agents) actually reach?
2. Mandate What were they authorised to do?
3. Gap Who controls the gap between the two?

Where this comes from

• websummit.com/sessions/lis26/0753d94e-3227-408e-b04d-a095ee327061/ai-…, read 2026-11-01 by the team's curator: LIS26 session page (conferenceSlug lis26): 'AI sprawl is exploding. Get ahead of governance and security challenges before it’s boo late', 12 Nov, Growth Summit, Stage 9, names Airia as the speaker's organisation.
• websummit.com/attendees/lis26/770add22-7f4e-4d16-97f2-88e9696d503b/ke…, read 2026-11-01 by the team's curator: LIS26 attendee page: Kevin Kiley, CEO, Airia AI; country and industry as listed by Web Summit; public bio.
• airia.com/, read 2026-11-01 by the team's curator: what the company does, from its own site (title and description).
• airia.com/, read 2026-11-06 by the team's curator: Airia's home page, read 6 November 2026: the platform's cycle (discover, secure, govern, optimise), its five layers (enterprise data, models and tools, agents and orchestration, runtime security, governance and compliance) and enforcement at the execution layer (the flows govern-every-agent and build-and-launch).

Re-checked by newsroom@web-summit.dinis on 2026-11-05 at 01:49 UTC: the summit's page answered and is still live; the company's website answered and is live.

Taken from the team's company map (data/companies.json on web-summit.sgit.ai), public facts only. The team's own notes on each company (why it matters, whether to meet, who owns it) are not facts about the company and are not published here. Something wrong? Tell us and we correct it on this page, dated.

About PrompTugal.news

PrompTugal.news: People and AI agents, reporting together. Every fact sourced and dated, for the people who are there. Edited by Dinis Cruz, Indi Gupte and Vicky Wang, made with AI agents. Web Summit is named only to say where we report from. Text CC BY 4.0, code Apache-2.0.

• Why this page exists. We keep a page on every company we follow at the summit, built only from public pages, so that anyone, Airia included, can see what we know and where it comes from. Nothing on it is an endorsement.
• Independence. Independent. Not affiliated with Web Summit. The team that makes PrompTugal.news also works on RiskMandate, so we are independent of Web Summit, not of RiskMandate. Our disclosure says it in full; please weigh what we say with it in mind.
• Corrections. If something here is wrong or out of date, we correct it on this page, dated. We publish no contact details and nothing about a person that is not already public and professional.
• More. About PrompTugal.news, all the companies, the summit.

The same, as data: companies/airia.json and its graph, graph/companies/airia.json (how to read them: the data, as JSON).

All companies

Open Airia in the graph