The RiskMandate X-ray
Type a name. Watch the agents X-ray it. An orchestrator sends the name to four search agents (Perplexity, OpenAI, Gemini and OpenRouter web search); a collector gathers what they find into a vault; the graph grows as the evidence arrives. At the end: what the subject's AI agents can reach, what a deployer would plausibly authorise (the mandate), and the gap between the two.
- Amex GBT
A travel management company whose agents now search, book and service business trips from inside Claude, Google Chat and Teams, on the traveler's own credentials. the deck - Atlassian
Atlassian's AI layer, Rovo, whose agents search, write and act across Jira, Confluence and connected apps, on a user's account or their own. the deck - Booking.com
A travel platform that runs AI agents for travellers (trip planning, support, voice, car rental) and for its partner properties (guest messaging). the deck - Cloudflare
The connectivity cloud that hosts AI agents: the Agents SDK on Durable Objects, Workers AI inference, AI Gateway and managed MCP servers. the deck - Feedzai
AI-native fraud and financial crime prevention for banks, now with Farol, an always-on agent for fraud and AML analysts, and a stated plan to tell good customer agents from bad ones. the deck - Fidelidade
Portugal's largest insurer, whose in-house AI reads motor claims, answers the phone and, its AI director told the press, drafts customer messages with language models the deck - Klarna
A payments and shopping company whose OpenAI-powered assistant answers most of its customer-service chats, with humans brought back as a choice. the deck - Manus AI
A general AI agent that plans and carries out tasks on its own cloud computer, browsing, running code and acting in the apps you connect. the deck - Primitive
Email infrastructure for AI agents: each agent gets its own address, an API to send and receive, and hosted code that runs on every inbound message. the deck - Replit
An AI platform whose agent builds, tests and publishes apps from plain language, with its own database, hosting and connectors. the deck - RiskMandate
Give agents access. Not free rein: a written record of what one AI agent can reach, what you authorised, and the gap. the deck - Salesforce
Salesforce's Agentforce: autonomous AI agents that read and act on CRM data across service, sales, marketing and commerce, inside the Einstein Trust Layer. the deck - SGit
Git for encrypted vaults, for humans and AI agents: the server stores ciphertext it cannot read, and agents hold vault keys, seal secrets to a person's key and sign what they produce. the deck
How to read an X-ray
- What a model says is a lead, not a fact. Each search agent comes back with claims and citations. We read the sources; only facts whose sources we read reach the page, each with its URL and the date read. Every lead is listed with its status: confirmed, not found or unverified.
- The mandate is illustrative. It is our judgement of what a deployer would plausibly authorise for that use case, so the gap can be seen. It is not the company's own policy.
- The run is a replay. The animation replays the real run: the agents in the order they ran, their timings scaled to a few seconds, their citation counts, and what the run cost.
- The simulation places, it does not stop. RiskMandate is never in the request path: read-only by design. It measures what your agents can actually reach, evidences the controls that contain them, prices the exposure, and produces that record.
A name that is not X-rayed yet plays the dispatch and says so honestly: our agents X-ray new names between runs. The request path from the booth is being built (WS-194). Nothing you type here is stored or sent.
How an X-ray is made, step by step, with what every run cost and how often each model's claims held up: How it works. Each X-ray has its own story too: "What happened: how this X-ray was made", linked from its page.
Booth mode, for the screen on the stand: ?booth=1.
Built by xray@web-summit.dinis for the booth at Web Summit, 9 to 12 November 2026.