Indi GupteProfileBriefDashboardMeet page

PolyAI: the factory infographic

ai-agentsUnited Kingdomat Web Summit: speaker's companyour decision: describe, priority 3

Voice AI agents that answer customer phone calls for large companies, from start to finish.

1

What the company does

PolyAIVoice AI forcustomer serviceAgents that talkand act on callsFor largeenterprises

In normal terms

Enterprise voice AI: agentic dialog platform for customer service.

PolyAI builds an agentic voice platform: AI agents that hold customer-service phone conversations and act on them for enterprises.

In lay terms

PolyAI makes AI that answers a company's phone: it talks to customers and sorts out what they called about.

2

The risks it faces, especially from AI agents

A voice thatcan actRegulatedcallersCoached, notwritten downAgent risks forPolyAI

In normal terms

  • A voice that can act. A voice agent that can change bookings, accounts or orders acts on a caller's word, which can be faked or misheard.
  • Regulated callers. In banking and other regulated services, every action on a call must be within what the firm has authorised, with evidence.
  • Coached, not written down. Agents are tuned and coached over time; without a written policy, nobody can show what the agent may and may not do today.

In lay terms

  • A voice that can act. Someone rings pretending to be you, and the AI changes your account.
  • Regulated callers. A bank must be able to show exactly what its phone helper is allowed to do.
  • Coached, not written down. The helper was trained like a new employee, but there is no job description to check it against.
3

How RiskMandate helps

Reachwhat it can touchMandatewhat it may dothe Gap
  1. Reachwhat it can actually access or affectEvery system a voice agent can act on during a call.
  2. Mandatewhat the business authorisedThe tasks the enterprise authorised it to do on a call.
  3. Gapthe difference: the riskActions on a call beyond that authority.
  4. Barrierswhat really stands in the wayCaller checks, limits and hand-offs to people, with evidence.
  5. Provenancewho owns it, who reviewed it, which versionWhich client owns each agent, who reviewed it, which version is live.

In normal terms

Their own session calls it coaching an AI agent like human talent. A behaviour policy is that coaching written down: a job description per voice agent that a bank's risk team can review and an insurer can read.

In lay terms

RiskMandate writes the job description for each phone helper, so the bank knows exactly what it may do.

A first line for the conversation: You coach your agents like human talent. Human talent has a job description. We write one for each of your agents.

Sources and method

The facts in grey (what it does) are from the company map, which cites:

  • websummit.com, read 2026-11-01: LIS26 session page (conferenceSlug lis26): 'Coaching an AI agent like human talent', 11 Nov, AI Summit, Stage 13, names PolyAI as the speaker's organisation
  • websummit.com, read 2026-11-01: LIS26 attendee page: Nikola Mrkšić, Co-founder & CEO, PolyAI; country and industry as listed by Web Summit; public bio
  • poly.ai, read 2026-11-01: what the company does, from its own site (title and description)

The risks and how RiskMandate helps are our analysis, written by dev@web-summit.indi on 3 November 2026 from those facts and RiskMandate's model as this site explains it (Reach, Mandate, Gap; an Agent Behaviour Policy in practice). They are not the company's own statements and name no incident. A person reviews each before it is used.

All the factory's companies

For an agent: this page is https://web-summit.sgit.ai/people/indi/factory/polyai/index.html (markdown: https://web-summit.sgit.ai/people/indi/factory/polyai/index.md). Site index for agents: https://web-summit.sgit.ai/llms.txt. Source: https://github.com/SGit-AI/SGit-AI__Website__Web-Summit (release branch dev). Identity of the agent that runs this site: https://web-summit.sgit.ai/agent/index.html.