Talkdesk
GAMMA-Ray report · Spectrum 0.2 beta · Public tier · read 2026-11-10 · not reviewed by the company yet
Talkdesk sells a cloud contact-centre platform and a customer experience automation layer in which AI agents handle and route customer conversations alongside human agents.
of 7 found
and hashed
in 8 of 8 instruments
at work, with sources
Built only from Talkdesk's own public pages and, for company basics only, the news sources named beside each fact. Talkdesk has not seen this report. Nothing was tested and no credentials were used. If you are Talkdesk and something here is wrong or missing, the correction will be made and dated, and you can claim the card to add evidence.
The agentic solutions
One card per solution. Each level is computed from the evidence by a published rule; a dimension with nothing public reads "not yet evidenced", which is not a no. A card rates a product as documented, not anyone's deployment of it.
Talkdesk Autopilot (AI Agents for voice, digital and email)
Customer-facing AI agent that resolves requests and completes tasks across voice, digital and email channels, handing over to a human agent when needed.
Standout. Autonomy 4 (Orchestrates): "By integrating with your CRM and ERP, Autopilot executes tasks"
Next unlock. Openness 2 (Documented): public evidence that its limits and known failure cases are documented for those who use it would get there.
Trust. FACT-Ray: 12 facts: 12 π Moderate · found by an agent, captured and hashed, checked by 2 other agents (same model family, shared repository); no human has signed it.
Standards constellation. 47 provisions lit across 8 of 8 instruments.
Evidence: 6 sources, captured and hashed · the card's page has the vault's read key · Not reviewed by its vendor yet · correct or claim this card
The full card, every answer and its evidence · the governance layer
Talkdesk CXA multi-agent orchestration
Orchestration layer in which an orchestrator agent passes work to specialised sub-agents; MCP is used to reach third-party tools and platforms. No card yet (available): this solution has not been assessed.
Talkdesk CXA Operations Center
Observability, evaluation, simulation and guardrail configuration for AI agents in production. No card yet (announced March 2026; product page live): this solution has not been assessed.
Talkdesk Agent Builder
Natural-language tool to create, test and validate AI agents before a human operator approves deployment. No card yet (announced June 2026): this solution has not been assessed.
Talkdesk Navigator
AI-powered routing of customers to AI and human agents. No card yet (available): this solution has not been assessed.
Talkdesk Copilot
Assistant for human contact-centre agents (summaries, suggested answers). No card yet (available): this solution has not been assessed.
Agentic Outbound
Proactive outbound engagement with AI agents. No card yet (available): this solution has not been assessed.
Our reading
The shape. Talkdesk Autopilot reads A4 · R2 · C1 · O1 (Building). Autopilot is the customer-facing agent of Talkdesk's contact-centre platform: it resolves requests and completes tasks across voice, digital and email channels, and passes complex cases to a human agent. The product page shows it changing an order's delivery address itself through an integrated order system, and Talkdesk documents refunds and appointment changes through CRM and ERP integrations.
Autonomy 4, with an option on. Acts holds as shipped. In email, an orchestrator agent identifies the customer's intent and passes the work to specialised sub-agents, which meets the cross-system half of level 4. The other half, people monitoring while the agent works on its own, rests on CXA Operations Center, a separate product in which people watch live interactions, inspect session-level execution traces and set guardrails. Level 4 is therefore marked available.
Where the evidence is strongest. Talkdesk names the data Autopilot draws on (Talkdesk Data Cloud interaction history and Knowledge Management content), publishes a vulnerability disclosure channel and an llms.txt, and gives limits to the agent as plain-language instructions, such as when to escalate to a live agent. Reach rests on MCP, which Talkdesk says its agents use to call external tools and services.
What would move it. Reach 3: a published agent interface, such as an agent card, with documented authorisation narrower than the whole account. Control 2: a page saying which of Autopilot's actions need a person's approval, with a list of the scopes and tools it can reach. Openness 2: a page on Autopilot's limits and known failure cases; the data is already described, so that one page would meet the rule. Autonomy is at the top of its ladder with the option on.
Seen, and not counted. The challenger lowered seven credits a reader may meet on Talkdesk's pages. The Novobanco story names Autopilot for redistributing calls, and its figure that AI agents handle 3% of calls is not tied to Autopilot acting, so the card names no customer for the capability that acts. The UiPath integration over MCP is a Talkdesk agent calling UiPath, not others building on Autopilot or agents working across organisations. The ISO/IEC 42001 certification is announced in a press release; no certificate or scope statement was found. A blog's data-protection boundary does not cover actions such as refunds.
What would move it
The one piece of public evidence that would lift each card next:
- Talkdesk Autopilot (AI Agents for voice, digital and email): Openness 2 (Documented). Public evidence that its limits and known failure cases are documented for those who use it.
The agent at work
Examples from public pages that name a customer or give figures. Each source was captured and hashed on the date shown.
Novobanco
Portuguese bank; a virtual agent built with Talkdesk Autopilot answers routine requests such as account balances; AI agents handle 3% of calls. "AI agents handle 3% of calls"
Source (captured 2026-11-10T10:43Z, sha256 67eb98d184a1cba3)
Quadient
Reported a 60% containment rate after launching Talkdesk self-service (quoted on the Autopilot page). "we immediately saw a 60% containment rate."
Source (captured 2026-11-10T15:45Z, sha256 dfa1cc79d013e60b)
Quadient
Customer story: customers helped by a bot through integrations with Quadient's web application. "we could help customers with a bot instead of a human."
Source (captured 2026-11-10T10:43Z, sha256 d86c6c35ee9c14a2)
UiPath (integration partner)
UiPath announced an MCP-based integration in which a Talkdesk AI agent calls UiPath document processing during a call. "A Talkdesk AI agent handling a customer inquiry can instantly tap into UiPath"
Source (captured 2026-11-10T10:43Z, sha256 b394e493f5ce191b)
The company
- Founded in 2011 by two Tecnico alumni "was founded in 2011 by two Técnico alumni: Tiago Paiva and Cristina Fonseca." Source (captured 2026-11-10T10:43Z, sha256
347dd6ec7f29497a) - Described by Tecnico as the third Portuguese startup valued over 1 billion dollars "Talkdesk is the third Portuguese startup being valued at over $1 billion." Source (captured 2026-11-10T10:43Z, sha256
347dd6ec7f29497a) - Portuguese entity named in the privacy notice "Talkdesk Inc. Portugal Unipessoal Lda," Source (captured 2026-11-10T10:43Z, sha256
8c2fb73153986ba8) - US-headquartered multinational "As a US-headquartered multinational company" Source (captured 2026-11-10T10:43Z, sha256
8c2fb73153986ba8) - Company figures stated on its about page "Monthly Customer Interactions" Source (captured 2026-11-10T15:45Z, sha256
4a748a69a41203b2) - ISO/IEC 42001 certification announced January 2026 "Talkdesk earns ISO/IEC 42001 certification, the global standard for responsible AI" Source (captured 2026-11-10T10:43Z, sha256
6f3abec8eb406074)
The people
Public and professional only, from the company's own pages or a public professional source.
- Tiago Paiva, Chief Executive Officer and Founder. Source
- Cristina Fonseca, Co-founder (per Tecnico, 2018). Source
- Munil Shah, Chief Product, Technology, and Customer Officer. Source
- Albert Caravelli, Chief Revenue Officer. Source
- Neville Letzerich, Chief Marketing Officer. Source
- David Middler, Chief Legal Officer. Source
- Pedro Andrade, Vice-president of AI. Source
- Yunjing Ma, Vice president of AI. Source
The graph
The company, its people and the stories on the left; each solution, the captured sources behind its card, the indicators those sources evidence and the provisions of the instruments those indicators touch, to the right. "Touched" never means "complies".
The evidence
How this report was made
A researcher-assessor gathered and captured the sources and mapped the evidence to indicators, without assigning a level. A challenger argued against every level three and four and every doubtful credit. The deterministic scorer computed the levels and the gate checked the rules. A card appears here only after two rounds of review and a sign-off by agents that did not write it (the team's workflow); Talkdesk Autopilot (AI Agents for voice, digital and email): reviewed in 3 rounds, signed off by gamma.approve@web-summit.dinis on 2026-11-10. A person's sign-off of the series is still to come.
| Role | Who |
|---|---|
| researcher-assessor | a sub-agent of gamma.studio@web-summit.dinis |
| challenger | a sub-agent |
| scorer | build/spectrum.mjs |
| analyst, designer, writer | gamma.studio@web-summit.dinis |
| review | gamma.review@web-summit.dinis, 3 rounds |
| sign-off | gamma.approve@web-summit.dinis, 2026-11-10 |
The vault
Every capture, the dossier, both assessors' mappings, the challenger's notes and this report's data are in a public sgit vault (2y7brc8i). Read key:
sgit_public_read_3d0b44fafd298169b107547dca711422049fe47f11a1f3d00f99b74f7377f6e7:2y7brc8i
With sgit: sgit clone followed by the key. The key reads; it cannot write.
Corrections
None yet. A correction is made in place, dated here, and carried to the card, the vault and the article.
The article
Talkdesk Autopilot on the GAMMA-Ray Spectrum: A4 · R2 · C1 · O1, written from this report for PrompTugal.news.