AI Agents & Assistants
PolyAI
UK / wider Europe Report an errorPanel rating · 6 judges · How to read the stars
Category median
Sovereignty: 1 of 4 dimensions proven
0–5 in half steps. 5 means the rubric's top anchor is met on the evidence.
by PolyAI Limited · poly.ai
Compare with Parloa → Report an error on this page Is this your product? →
Link to this page
Add ?theme=dark to the image address for dark backgrounds.
The badge shows the category and the month of the latest evaluation, and EU-Made where
that is the product’s provenance. It never shows a score.
PolyAI at a glance
- Vendor
- PolyAI Limited
- Headquarters
- Not stated on the vendor’s pages as of 26 Oct 2026
- Data hosting
- Not stated on the vendor’s pages as of 26 Oct 2026
- Free plan
- Not stated on the vendor’s pages as of 26 Oct 2026
- Panel verdict
- The AI panel rates PolyAI 2/5, the mean of 6 judges. PolyAI is an enterprise voice AI agent platform for customer service, powered by the vendor's proprietary Raven model, and the bench's scores are tight: no criterion spans more than one point. how scores are made
- As of
- Vendor pages captured 26 Oct 2026; evaluated by the panel 28 Oct 2026. all sources
Read this page as one judge. Each weighs the same scores by what they care about.
The panel's verdict
PolyAI is an enterprise voice AI agent platform for customer service, powered by the vendor's proprietary Raven model, and the bench's scores are tight: no criterion spans more than one point. It is strongest on integrations and knowledge and agent capabilities, with 130+ integrations across telephony, CRM and vertical systems, named partners including Salesforce, Zendesk and Twilio, OAuth 2.0, a self-serve Connect Portal, an Agent Developer Kit, and test calls, simulated conversations and A/B tests on live traffic at no extra charge. It is weakest on pricing transparency, where the pricing page publishes no figures and routes buyers to "Talk to our team", and on admin and access, flat at 2, where we found no public information on single sign-on, roles or audit logs. Data use and training and sovereignty stay low: we found no public information on whether conversations train models, even as Raven is advertised as "trained on 1B+ enterprise conversations", and hosting is stated only as AWS, with no server region or published subprocessor list. The judges show no genuine split.
Speaks for it
- 130+ pre-built integrations across telephony, CRM, productivity and vertical systems, with named partners including Salesforce, Zendesk, NICE, Five9, Twilio and Verizon Business.
- Managed OAuth 2.0 with automatic token refresh, custom objects and field mappings, a self-serve whitelabelled Connect Portal, open APIs and an Agent Developer Kit (CLI tool and Python package).
- Test calls, simulated conversations and A/B tests on live traffic are included at no extra charge, with real-time latency and cost monitoring per agent.
- Use is priced per minute and billed per second of connected time, with no minimums and no rounding up to the minute, and new Dialog-RSN-1 API accounts get $100 in free credit for a limited time.
- SOC 2 and ISO 27001 certifications are published, with regular audits and third-party penetration testing.
Held against it
- We found no public information on whether conversations, prompts or outputs are used for training, while Raven is advertised as 'trained on 1B+ enterprise conversations'.
- The pricing page publishes no figures — 'Flexible pricing … Talk to our team' — and we found no public information on per-minute rates, tiers, usage limits or minimum commitments.
- We found no public information on single sign-on, SCIM provisioning, roles, group-based permissions or audit logs.
- Hosting is stated only as AWS with no server region given, we found no published subprocessor list, and the privacy policy mentions service providers with operations 'including in the United States'.
- Retention is publicly stated only for client administrator records — kept for six years after the agreement — with no public information on retention of conversations or call logs.
Best for
- You run an enterprise contact centre on telephony and CRM stacks such as Salesforce, Zendesk, NICE, Five9 or Twilio and want voice agents that act inside those systems.
- You want to test before scaling — test calls, simulated conversations and A/B tests on live traffic are included at no extra charge.
- You route calls over SIP or PSTN and want context handed back to human agents via SIP headers.
- Your team wants to build agents without code from a plain-language description, or programmatically via the CLI tool and Python package.
Avoid if
- Your finance team needs to forecast an annual cost from public pages — the pricing page publishes no figures and routes buyers to 'Talk to our team'.
- You need written answers, before signing, on whether customer conversations train models and how long transcripts are kept.
- You need contractual accuracy guarantees — the terms provide AI Outputs 'as is' and state they 'may contain inaccuracies'.
- Your rollout cannot proceed without documented single sign-on, provisioning, roles and audit logs.
The scores
| Criterion | Panel median |
|---|---|
| Data use, training & retention | 1 of 5 |
| Models & processing location | 2 of 5 |
| Admin, identity & access controls | 1 of 5 |
| Integrations, connectors & knowledge | 3.5 of 5 |
| Agent capabilities & guardrails | 3 of 5 |
| European sovereignty | 1.5 of 5 |
| Pricing transparency | 1 of 5 |
Data use, training & retention
Show reasoningHide reasoning
How this is scored
What happens to prompts, uploaded files, connected data and outputs: whether any of it trains a model (the vendor's or a model provider's), how long it is kept and where logs live, whether zero-retention is available, and whether a data processing agreement covers it.
0 — No statement on training or retention; the privacy policy is a website policy that does not cover the product.
3 — A general statement that business data is "not used for training", with no word on retention, on the model providers behind the product or on logs.
5 — Business data excluded from training by contract, retention periods stated for chats and files, and a DPA offered to business customers.
8 — No-training commitments stated for the vendor and every model provider it routes to, retention configurable by the admin (including deletion of chat history), zero-data-retention terms with the model providers named, and abuse-monitoring retention described.
10 — Data use fully accounted for: training, retention, logging and abuse monitoring stated per model provider, customer-controlled retention and deletion with an audit trail, outputs and connected data treated as customer data in the DPA, and an independent audit (SOC 2, ISO 27001 or ISO 42001) covering the product.
The Head of IT
The privacy policy does cover the product and records that the vendor is contractually committed to processing client personal data only on the client's instructions, with six-year retention stated for client administrator records, and SOC 2 and ISO 27001 are advertised. But we found no public information on whether conversations train the Raven model, how long call content is kept, zero-retention terms, or abuse-monitoring retention — and Raven is advertised as trained on 1B+ enterprise conversations, which makes that silence a real problem for me. 2 4 6
The Data Protection Officer
The privacy policy commits the company to processing client personal data only on the client's instructions as processor and states that client administrator records are kept "for as long as we have an agreement with the relevant client and for 6 years thereafter", but we found no public information on whether conversations, transcripts or recordings train any model, on retention for call logs, or on zero-data-retention terms with model providers. SOC 2 and ISO 27001 certifications are published, which is welcome but does not answer the training and retention questions a data processing agreement review needs. 3 4 6
The Automation Lead
The privacy policy commits PolyAI to process client conversation data only on the client's instructions and states one retention period — six years after the agreement ends, for client administrator records — but I found no public information on whether conversations or outputs are used for training, on retention for call and chat content, or on a zero-retention option. SOC 2 and ISO 27001 certifications are published, yet they sit on top of a training and retention story the captured pages never tell, and the vendor's own Raven model is described as trained on 1B+ enterprise conversations. 6 4 2
The Customer-Service Director
The privacy policy commits PolyAI to processing client conversations as a processor on the client's instructions, gives a six-year retention figure for client administrator records, and the security pages state SOC 2 and ISO 27001 certification. But we found no public information on whether recorded calls and transcripts are used for training, on retention of conversation data, or on zero-retention terms — and while Raven is advertised as "trained on 1B+ enterprise conversations", we found no public information on whether live client calls feed that training. That gap I would need closed before letting customer conversations flow. 2 4 6
The Works Council Member
The privacy policy commits the vendor to process client-conversation data "only on the instructions of the relevant client" and gives one concrete retention period — administrator data kept six years after the agreement ends — but we found no public information on whether conversations or staff input are used to train models, on transcript retention, or on zero-retention terms. The vendor's own pages say Raven was "trained on 1B+ enterprise conversations", and without a written exclusion that is precisely the training-on-input question a works council needs answered in writing before any rollout. 6 2 8
The Skeptic
The privacy policy commits PolyAI to processing client conversation data only on the client's instructions and states one retention figure — administrator contact data kept six years after the agreement ends — but we found no public information on whether conversations, prompts or outputs are used for training, and the vendor advertises Raven as "trained on 1B+ enterprise conversations" without a word on whose. Zero-retention options, logs and abuse-monitoring retention are likewise undocumented. 6 2
Models & processing location
Show reasoningHide reasoning
How this is scored
Which models sit behind the product and who runs them, where inference and storage happen, whether the customer can pin processing to the EU, choose or exclude model providers, bring its own model or key, or run the product in its own cloud or on premises.
0 — The models behind the product and the processing location are not stated.
3 — The model family is named, but where inference runs and through which provider (the model maker, a hyperscaler, the vendor) is unstated.
5 — Models and their providers named, with an EU processing option stated for storage and at least the default model.
8 — Each available model listed with provider and inference region, EU-only processing selectable by the admin and enforced across models, model providers excludable, and bring-your-own-key or own-cloud deployment documented.
10 — Processing fully under the customer's control and documented: EU-only or on-premises inference for every feature including agents and connectors, open-weight or self-hosted models supported, data flows per feature published, and a residency commitment in the contract rather than on a marketing page.
The Head of IT
The product is powered by the vendor's own Raven model — versions, training method and latency claims are documented — on AWS infrastructure, so I at least know who runs inference. But we found no public information on server region, an EU processing option, provider pinning or exclusion, or any bring-your-own-key, own-cloud or on-premises deployment, so residency rests on trust rather than documentation. 2 3 8
The Data Protection Officer
The product runs on Raven, the vendor's proprietary model trained on enterprise customer service conversations, on infrastructure the security page describes as hosted on AWS; we found no public information on the inference region, an EU-only processing option, excluding providers, or bring-your-own-key and own-cloud deployment. As a data protection officer I cannot sign a DPA for processing whose location is unverified. 2 3 8
The Automation Lead
The model family is named — Raven, a proprietary model purpose-built for voice customer service — but the pages say nothing about where inference runs or which provider operates it, with hosting stated only as AWS and no server region given. I found no public information on EU processing, provider exclusion, bring-your-own-key or self-hosting, so I cannot pin processing for a European deployment. 2 8 3
The Customer-Service Director
The model behind the product is named — Raven, a proprietary model the vendor runs itself, with AWS named as the host. We found no public information on which region inference and call data live in, whether processing can be pinned to the EU, or any bring-your-own-model or on-premises option, which is what I would need in writing before routing regulated conversations. 2 3 8
The Works Council Member
Raven, the vendor's own proprietary model, is named as what powers the agents, and hosting is stated as AWS, but we found no public information on the inference region, an EU processing option, provider exclusions, or bring-your-own-key deployment. The "gated generative AI" used selectively in conversations is described without naming the provider behind it. 2 3 8
The Skeptic
Raven, a proprietary model, is named as what powers the agents, and the security page says infrastructure is "hosted on AWS" with no server region given — where inference actually runs is unstated. The "gated generative AI" the security page describes inside conversations is attributed to no provider, and we found no public information on an EU processing option, excludable providers, bring-your-own-key or own-cloud deployment. 2 3 8
Admin, identity & access controls
Show reasoningHide reasoning
How this is scored
What the IT department controls: SSO and SCIM provisioning, roles and groups, which models, agents and connectors each group may use, retrieval that respects the permissions of the source system, audit logs, usage reporting, and policies such as data-loss prevention — and whether reporting on staff use is scoped and documented.
0 — No admin features documented; accounts are individual sign-ups.
3 — A team workspace with an owner and member invitations, but no SSO, no roles and no audit log.
5 — SSO, admin and member roles, user management, and an admin view of usage.
8 — SAML/OIDC SSO with SCIM provisioning, group-based permissions for models, agents and connectors, retrieval that inherits source-system permissions, exportable audit logs, and usage analytics with documented scope.
10 — Governance a regulated company can operate: fine-grained policies (DLP, content filters, prompt and output logging rules), audit logs streamed to a SIEM, per-agent ownership and approval, usage reporting that can be aggregated or pseudonymised for works-council agreements, and every admin action logged.
The Head of IT
We found no public information on single sign-on, SCIM provisioning, roles, group-based access to models, agents or connectors, audit logs, or data-loss-prevention policies. The evidence stops at the existence of client administrators, a console showing real-time latency and cost per agent, and a self-serve whitelabelled Connect Portal for configuring integrations — not enough for me to govern a rollout to eight hundred staff or feed a SIEM. 4 6 9
The Data Protection Officer
A console with "Latency and cost per agent on the console, in real time" and integration dashboards showing real-time sync status, error tracking and automatic retry are documented. We found no public information on single sign-on, SCIM provisioning, roles and group permissions for models or connectors, audit logs, or how reporting on staff use is scoped. 4 9
The Automation Lead
I found no public information on SSO, SCIM provisioning, roles, group-based permissions for models or connectors, or audit logs — for an agent platform that acts in CRM and ticketing systems, that is the gap that matters most to me. What is documented is operational rather than governance: a console showing latency and cost per agent in real time, a dashboard with sync status, error tracking and automatic retry, and managed OAuth 2.0 on connectors. 4 9
The Customer-Service Director
A console showing latency and cost per agent in real time and a whitelabelled Connect Portal inside Agent Studio are documented. We found no public information on SSO, role-based access for supervisors, user provisioning, group permissions on agents and connectors, or exportable audit logs — controls I cannot run a forty-seat operation without. 4 9
The Works Council Member
The captured pages document operational dashboards — real-time sync status and error tracking, and "latency and cost per agent on the console, in real time" — but we found no public information on single sign-on, SCIM provisioning, roles, group-based permissions or exportable audit logs. Likewise we found no public information on whether usage reporting on staff can be scoped, aggregated or pseudonymised for a works-council agreement. 9 4
The Skeptic
We found no public information on single sign-on, SCIM provisioning, roles, group-based permissions or audit logs. The closest documented evidence is an operator console showing latency and cost per agent in real time, a dashboard for integration sync status and errors, and the privacy policy's category of "client administrators". 4 6 9
Integrations, connectors & knowledge
Show reasoningHide reasoning
How this is scored
How the product reaches company knowledge and systems: connectors to file stores, wikis, mail, CRM, ticketing and ERP, how fast and how completely they sync, the API, support for open protocols such as MCP, and how the assistant cites the sources it answers from.
0 — No integrations; users paste text or upload files one at a time.
3 — File upload and one or two connectors (for example a drive), with no API and no statement on sync or citations.
5 — A connector library covering the common file stores, wikis and collaboration tools, answers citing sources, and a documented API.
8 — Connectors to file stores, wikis, mail, CRM and ticketing with incremental sync, custom connectors or MCP support, an API and webhooks for embedding assistants and agents in other tools, and source citations on every grounded answer.
10 — The product sits in the company's systems: read and write connectors with documented scopes, MCP servers and clients, contact-centre or telephony integration where relevant, a knowledge layer that can be audited and corrected, and the data, agents and configuration exportable when the customer leaves.
The Head of IT
This is the strongest area: more than 130 integrations across telephony, CRM and knowledge, productivity and vertical systems, with named partners such as Salesforce, Zendesk, NICE, Five9 and Twilio, open APIs, an Agent Developer Kit for custom integrations, managed OAuth 2.0, and real-time sync status with error tracking and automatic retry. We found no public information on MCP support, webhooks, incremental sync, or how answers cite the sources they are grounded in. 2 3 9
The Data Protection Officer
More than 130 integrations span telephony, CRM and knowledge, productivity and collaboration, and vertical systems, with a self-serve Connect Portal, managed OAuth 2.0 with token refresh, custom field and object mapping, and an Agent Developer Kit for programmatic control of custom functions and conversation flows. We found no public information on MCP support, webhooks, or incremental sync cadence. 7 9
The Automation Lead
This is the strongest-documented area: 130+ pre-built integrations across telephony, CRM and knowledge, productivity and vertical systems, named partners like Salesforce, Zendesk, Twilio and NICE, open APIs, a self-serve Connect Portal with managed OAuth and custom field mapping, and an Agent Developer Kit as a CLI tool and Python package for custom functions into any enterprise system. Telephony is first-class — SIP or PSTN in, with custom SIP header information transferred back to the contact centre for handoff. I found no public information on MCP support, webhooks, or how answers cite their sources. 9 2 7
The Customer-Service Director
This is where the product speaks my language: 130+ integrations across telephony, CRM and knowledge, productivity and collaboration, and vertical systems, with named partners including NICE, Five9, Salesforce, Twilio, Verizon Business and Zendesk, and a SIP or PSTN connection to route calls into my existing contact centre team. Managed OAuth 2.0 with automatic token refresh, custom fields, objects and user-defined filters, an Agent Developer Kit for custom functions into any enterprise system, and real-time sync status with error tracking and automatic retry visible from the dashboard make the order and ticket side credible. 2 4 7 9
The Works Council Member
This is a broad documented library — 130+ integrations across telephony, CRM, productivity and vertical systems, managed OAuth 2.0 with token refresh, custom field and object mapping, a self-serve Connect Portal, and an Agent Developer Kit (a CLI tool and Python package) for custom connections, with SIP or PSTN enough to route calls. We found no public information on MCP support, webhooks, or how answers cite their sources, which keeps it short of the top level. 9 2 7
The Skeptic
This is the strongest documented area: 130+ integrations across telephony, CRM and knowledge, productivity and vertical systems, named partners including Salesforce, Zendesk, NICE, Twilio and Five9, managed OAuth 2.0 with token refresh, custom objects and user-defined field mappings, a self-serve Connect Portal, open APIs, and an Agent Developer Kit to build custom functions against any enterprise system. We found no public information on MCP support, webhooks, or export of data and configuration when a customer leaves. 9 2 7
Agent capabilities & guardrails
Show reasoningHide reasoning
How this is scored
What an agent can actually do and where it stops: multi-step tasks, tool use and actions in other systems, scheduling and triggers, human approval before consequential actions, run logs, evaluation and testing before release, and the limits the vendor states rather than hides.
0 — No agents; the product is a chat window, whatever the marketing says.
3 — Saved prompts or custom assistants with instructions and files, but no actions in other systems and no run history.
5 — Agents that call tools or APIs and run multi-step tasks, built without code, with a log of each run.
8 — Agents that act in connected systems with scoped permissions, human approval configurable per action, triggers and schedules, test and evaluation tooling before release, and documented limits and failure handling.
10 — Agents a business can put in production and account for: versioning and staged rollout, evaluation against test sets, monitoring of quality and cost per agent, handover to a human with full context, every action attributable to the agent and the person who deployed it, and the vendor's own statement of what agents must not be trusted with.
The Head of IT
Agents can be built without code from a plain-language description or programmatically via the CLI and Python package, they act in connected systems, and test calls, simulated conversations and A/B tests on live traffic are included at no extra charge, with SIP-header context handed to human agents at handover. We found no public information on per-action human approval, scheduled triggers, or per-run logs, though the terms state plainly that AI outputs are informational only and "may contain inaccuracies". 2 4 5 7
The Data Protection Officer
Agents are built from plain-language descriptions with no code, act in connected systems for use cases such as billing and payments and account management, and "Test calls, simulated conversations and A/B tests on live traffic, at no extra charge" are documented alongside real-time cost and latency monitoring per agent and SIP-header context transferred back to contact-centre agents at handoff. We found no public information on human approval before consequential actions, triggers and schedules, or attribution and run-level audit of who deployed which agent. 2 4 7 9
The Automation Lead
Agents act in connected systems through the integration library and Agent Developer Kit custom functions, and the testing story is genuinely good: test calls, simulated conversations and A/B tests on live traffic at no extra charge, per-agent latency and cost monitoring on the console, and SIP-header handoff to human agents. The terms also state a limit plainly — AI outputs are provided as-is and may contain inaccuracies. What I found no public information on is human approval before consequential actions, per-action scoped permissions, versioning, and an auditable run log — the guardrails I need before an agent touches production systems. 9 2 4 5 7
The Customer-Service Director
The pieces I need for a controlled launch are documented: agents built from a plain-language description with no code, custom functions into back-end systems, test calls, simulated conversations and A/B tests on live traffic at no extra charge, real-time latency and cost per agent, and custom header information in the SIP header transferred back to the contact centre to assist agents with handoff. We found no public information on configurable human approval before consequential actions, on triggers and schedules, or on per-run logs, and the terms state AI Outputs "may contain inaccuracies" — a caution I take at face value for a voice channel. 2 4 5 7
The Works Council Member
Agents act in connected systems — billing and payments, bookings, authentication — are built without code, and the vendor documents both pre-release testing ("test calls, simulated conversations and A/B tests on live traffic, at no extra charge") and real-time per-agent cost and latency monitoring, with a SIP-header handover carrying context back to contact-centre colleagues. We found no public information on a configurable human-approval step before consequential actions or on agent versioning and run logs, and the only stated limit we found is the terms' warning that AI Outputs "may contain inaccuracies". 2 4 7 5
The Skeptic
Voice agents are documented acting in connected systems — billing and payments, order management, authentication, bookings — with test calls, simulated conversations and A/B tests on live traffic included at no charge, real-time latency and cost monitoring, automatic retry on integration errors, and SIP-header context passed to human agents at handoff. We found no public information on configurable human approval before consequential actions, scoped per-agent permissions or versioned rollout; the terms do at least state that AI Outputs "may contain inaccuracies". 4 9 7 10
European sovereignty
panel opinion
Show reasoningHide reasoning
How this is scored
Who the contracting entity and the parent company are, where customer data is hosted, and who the subprocessors are — model providers above all, then hosting, search, telephony and email. Independently sourced by the sovereignty pipeline; for an AI product the model provider is a subprocessor, so a European interface over a non-European model counts as such.
0 — Non-EU vendor and contracting entity, hosting unstated, subprocessors (including model providers) unnamed.
3 — A privacy policy exists, but hosting location is unstated or non-EU and the model providers behind the product are not named as subprocessors.
5 — EU hosting stated and an EU contracting entity, but prompts reach non-EU model providers or other subprocessors without a stated safeguard or EU region.
8 — EU hosting on named infrastructure, EU contracting entity, every subprocessor listed with its location including model providers, EU inference available, and a DPA offered.
10 — Sovereign end to end and evidenced: European ownership, entity, hosting and model processing, every subprocessor European or optional, data export and deletion documented, and a security or certification statement published rather than implied.
The Head of IT
The contracting entity is PolyAI Limited under the exclusive jurisdiction and law of England and Wales, the UK Information Commissioner's Office is named as the complaints authority, and AWS (hosting) and Stripe (payments) appear as named service providers, with SOC 2 and ISO 27001 published. But we found no public information on hosting region or model-processing location, and no published subprocessor list, while the privacy policy mentions service providers with operations "including in the United States". 3 5 6
The Data Protection Officer
The contracting entity is PolyAI Limited under the exclusive jurisdiction of England and Wales, the infrastructure is described as hosted on AWS, and Stripe is named as payment processor; the privacy policy's mention of service providers with operations "including in the United States" gives no regional control. We found no public information on the hosting region, data residency, ownership, or a published subprocessor list naming model providers with their regions — for an AI product that list is where my review starts. 3 5 6
The Automation Lead
The contracting entity is named — PolyAI Limited, terms governed by the laws of England and Wales, the UK Information Commissioner's Office as the data protection authority — with hosting on AWS at no stated region and Stripe named for payments. I found no published subprocessor list, no EU inference or hosting commitment, and the privacy policy's own mention of service providers with operations including in the United States leaves the exposure question open; the model is the vendor's own Raven, but where it runs is not said. 5 3 6
The Customer-Service Director
The contracting entity is PolyAI Limited under the exclusive jurisdiction of England and Wales, AWS is named as host, and SOC 2, ISO 27001 and NCSC guideline alignment are stated. But no server region is given, the only named subprocessor is Stripe for payments, the policy refers to service providers with operations "including in the United States", and we found no public information on EU data residency or an EU inference option. 3 4 5 6
The Works Council Member
The contracting entity is the UK company PolyAI Limited, governed by the laws of England and Wales with exclusive jurisdiction there; hosting is stated only as AWS with no region, and the privacy policy's service providers include operations "in the United States", with Stripe named as payment processor. We found no public information on data residency for customer conversations or on a published subprocessor list. 5 3 6
The Skeptic
The contracting entity is named — PolyAI Limited, terms governed by the laws of England and Wales — and two subprocessors appear: AWS for hosting with no server region given, and Stripe for payments. We found no public information on where customer data is hosted, no published subprocessor list, and the privacy policy's supporting text mentions service providers with operations "including in the United States"; running an in-house model avoids an external model provider, but inference location is undocumented. 3 5 6
Pricing transparency
Show reasoningHide reasoning
How this is scored
Whether a company can compute the real annual cost from public pages alone: price per seat, per agent, per conversation or per usage unit, what is included in each tier, usage limits and overage, which models, connectors or agents cost extra, minimum seats and term.
0 — No public prices at all; every customer is a sales call.
3 — A starting price or "from" figure exists, but usage limits, included models and what scales the cost are unclear.
5 — Prices per tier are public with the main inclusions, but usage limits, overage, agent runs or enterprise features (SSO, audit logs, EU processing) are unpriced or "on request".
8 — Every tier priced publicly with seat or usage units, usage limits and overage rates, add-ons listed, minimum seats and term stated.
10 — Complete price computability: annual cost derivable for a given number of seats, agents or conversations, with every usage rate, model surcharge, add-on and support tier and the cancellation and data-return terms published.
The Head of IT
The pricing page publishes no figures and routes every buyer to "Talk to our team" or a demo. It does disclose the metering model — per-minute pricing billed per second of connected time, with no minimums and no rounding up to the minute, including maintenance and 24/7 support — and $100 in free credit for new Dialog-RSN-1 API accounts, but no rate, tier structure or enterprise add-on prices are public, so I cannot compute an annual cost. 4 8
The Data Protection Officer
The captured pages describe the pricing model — "priced on a per-minute basis" and "Billed on connected time, with no minimums and no rounding up to the minute" — plus "$100 in free credit for a limited time" for new API accounts, but no public price figures; the pricing page shows "Flexible pricing" and "Talk to our team". A company cannot compute an annual cost from these pages. 4
The Automation Lead
No public prices at all — the pricing page says "Talk to our team" and points to a demo form, so every customer is a sales call. The mechanics are described honestly — per-minute usage-based pricing, "billed on connected time, with no minimums and no rounding up to the minute", $100 in free API credit for new accounts for a limited time, with testing, model upgrades and 24/7 support included — but a buyer cannot compute an annual cost from any of it. 4 8
The Customer-Service Director
The pricing page shows no figures — "Flexible pricing... Talk to our team" — so my CFO cannot forecast an annual cost at all. The shape is public and genuinely useful: use is "priced on a per-minute basis", "Billed on connected time, with no minimums and no rounding up to the minute", including support, maintenance and model upgrades, and new API accounts get "$100 in free credit" — but no per-minute rate itself is published. 4 8
The Works Council Member
The billing model is public — the voice agent is "priced on a per-minute basis" including support, maintenance and model upgrades, billed per second of connected time with no minimums and no rounding up to the minute — but the captured pricing page shows no public prices and routes buyers to "Talk to our team". The only public figure is the time-limited "$100 in free credit" for new API accounts, so a real annual cost cannot be computed from public pages. 4 8
The Skeptic
The pricing page publishes no figures — "Flexible pricing… Talk to our team" — and every route to a number is a demo or a sales contact; what is public is the shape of the bill: per-minute usage pricing billed per second of connected time, with no minimums and no rounding up to the minute, testing and 24/7 support included, plus $100 in free API credit for new accounts. We found no public information on per-minute rates, tiers, usage limits or minimum commitments. 4 2
European sovereignty — proven facts
1 of 4 dimensions provenBuilt only from facts shown on the vendor's own pages. A dimension we could not prove is left open, not scored as zero.
| Legal entity | Not determined ⚠ unverified | — | uncited Report an error |
|---|---|---|---|
| Ownership | Not determined | — | uncited Report an error |
| Data residency | Not determined ⚠ unverified | — | uncited Report an error |
| Subprocessors | US CLOUD Act reach ⚠ unverified | 0/2 pts | 3 Report an error |
Questions about PolyAI
Is PolyAI European?
Yes, though not from the EU. whats-best.ai classifies PolyAI as UK / wider Europe.
Where does PolyAI store data?
Not stated on the vendor’s pages as of 26 Oct 2026.
Does PolyAI have a free plan?
Not stated on the vendor’s pages as of 26 Oct 2026.
What are the best alternatives to PolyAI?
As of 28 Oct 2026, the panel rates Langdock (4/5), Dust (3.5/5) and Unique AI (3.5/5) highest among the other 15 AI Agents & Assistants products evaluated. Ratings that look equal are rounded to half stars; the order uses the unrounded rating, then the name. The highest-rated European alternative is Langdock (EU-Made). Ratings are the AI panel’s opinion, formed from public evidence: how scores are made. The full list, with European products first: PolyAI alternatives.
Where this could be wrong
- Evidence ages. The oldest capture behind this page is from 26 Oct 2026. Vendors change pricing and policies without notice; every fact reflects its source as of the capture date shown in the registry.
- Weak sourcing — Legal entity, Data residency. Not confirmed on the vendor’s own pages as captured.
- Weak sourcing — Subprocessors. Only the hosting subprocessor is named, with no server region given and no published subprocessor list; the privacy policy's supporting mention of service providers with operations "including in the United States" sits in a section whose country list is skipped in the excerpt.
- AI can misread a source. Extraction and judgement are automated; a citation guarantees traceability, not infallibility. If something here is wrong, say so — no account needed, every report is decided within 5 business days, and accepted corrections are published.
What we left out
A claim that does not survive our checks costs us the claim, not the page. This is what was taken off this one.
- 18 product facts could not be confirmed on the vendor’s page as captured and were left out of this page and of the panel’s material. Know more? Tell us
- 4 compliance facts could not be confirmed on the vendor’s page as captured and were left out of this page and of the panel’s material. Know more? Tell us
- 4 legal facts could not be confirmed on the vendor’s page as captured and were left out of this page and of the panel’s material. Know more? Tell us
- 2 data facts could not be confirmed on the vendor’s page as captured and were left out of this page and of the panel’s material. Know more? Tell us
- 2 hosting facts could not be confirmed on the vendor’s page as captured and were left out of this page and of the panel’s material. Know more? Tell us
- 2 integrations facts could not be confirmed on the vendor’s page as captured and were left out of this page and of the panel’s material. Know more? Tell us
- 2 sovereignty dimensions could not be confirmed on the vendor’s own pages and are shown as unknown. Know more? Tell us
Sources (10)
The pages every claim on this page was read from — each one checked, dated, and kept verifiable.
- 1 Vendor homepage poly.ai Checked 26 Oct 2026 Details →
- 2 Features page (agent studio) poly.ai Checked 26 Oct 2026 Details →
- 3 Security page poly.ai Checked 26 Oct 2026 Details →
- 4 Pricing page poly.ai Checked 26 Oct 2026 Details →
- 5 Terms naming the legal entity poly.ai Checked 26 Oct 2026 Details →
- 6 Privacy policy poly.ai Checked 26 Oct 2026 Details →
- 7 Models & processing location — found from sitemap poly.ai Checked 26 Oct 2026 Details →
- 8 Models & processing location — found from sitemap poly.ai Checked 26 Oct 2026 Details →
- 9 Integrations, connectors & knowledge — found from sitemap poly.ai Checked 26 Oct 2026 Details →
- 10 Agent capabilities & guardrails — found from sitemap poly.ai Checked 26 Oct 2026 Details →