PolyAI or Replicant for inbound calls: telephony, actions, handoff, oversight

Telephony: PolyAI lists its connectors, Replicant leaves the question open

If you already run a PBX or a contact-center platform, start here. PolyAI's integration documentation groups its telephony options under call routing, transfers and live-agent handoffs: a generic SIP trunk, forwarding an existing number to a new PolyAI number, Five9 through a multi-tenant SIP trunk, NICE CXone through its Signal API, Twilio, Amazon Connect, Genesys Cloud BYOC, Zendesk Talk, Dialpad, and Zoom Phone and Zoom Contact Center. The docs say each of these is available from Integrations in Agent Studio.

Replicant's homepage presents phone, SMS and chat automation but does not name compatible telephony platforms. The directory record lists Genesys, Five9, NICE inContact, Cisco, Talkdesk and Ujet among its integrations, without saying which of them carry live calls. That is unknown, not absent. NJ TRANSIT's Access Link paratransit service offers one data point: the vendor case study says it added Replicant to complement its phone system rather than redo it, but it does not name the connection method.

12456

Business actions: what each agent does inside your systems

PolyAI's docs say the agent can look up reservations, create tickets and retrieve knowledge articles during calls. Connectors cover Salesforce, Zendesk, HubSpot, Dynamics 365, OpenTable and Epic, and anything exposing an MCP server can be added. For other platforms you define HTTP endpoints and call them from Python functions. Two vendor case studies show the shape: Hawksmoor connected the agent to its OpenTable availability so callers see the same table inventory as a reservationist, and Carter's agent looks up available discounts for price adjustments and sends refunds to the original payment method.

Replicant describes agents built from existing conversation data that combine language models with deterministic reasoning for business decisions. In its case studies, ECSI's agents authenticate callers, capture intent, and resolve or route calls for tax forms, refunds and loan servicing. NJ TRANSIT automates ride-status checks and cancellations. Evive Brands uses agents for scheduling, lead qualification, FAQs and after-hours calls. These stories describe tasks, not the connectors behind them, so ask which of your back-office systems each task reads and writes.

12678910

Handoff to an advisor: documented platforms versus documented outcomes

PolyAI documents handoff at two levels: live-agent transfer inside the telephony integrations, and chat or SMS handoff to Salesforce, Zendesk, NICE CXone, Amazon Connect, Genesys, Webex or Five9. In Audibel's vendor case study, the agent collects name, phone number, zip code and intent, then passes the call to a live agent with context. Carter's reports that its chat agent, not its voice agent, made more than 3,000 live-agent transfers with zero queuing over Black Friday weekend. On voice, Carter's added callback scheduling for callers who prefer a human later.

For Replicant, the evidence is outcome-level. ECSI's case study says escalated calls reach agents with a summary so they do not repeat steps, and Evive Brands describes complex conversations passing to human agents. Which platforms receive the transfer, and whether it is warm or cold, is not stated in those pages. PolyAI's integration page names platforms but does not list which context fields reach the agent desktop either, so both vendors owe you the same demonstration.

1891011

Oversight and data terms: where the two sources differ

Replicant's security page is the more detailed on day-to-day controls. It describes automatic redaction of personal and payment data across AI interactions, transcripts, analytics logs and QA workflows, granular permissions with audit trails, role-based access with multi-factor authentication, and analysis of every conversation for traceability. It lists SOC 2 Type 2, PCI DSS and HIPAA. The directory records training use as off by default, while the page wording is stronger, so get the exact term in the contract.

PolyAI's homepage lists SOC 2+ Type 2, ISO 27001, PCI DSS and GDPR, encrypted SIP, SSO and role-based access, data residency in the US, UK, EU or APAC, PII redaction and human-in-the-loop controls. Its security page adds AWS hosting and third-party penetration testing, and a data processing agreement is published. One tension to resolve rather than assume: the homepage says data is never trained on without opt-in, while the DPA grants PolyAI a licence to use communications data, such as recordings and transcripts, to create analytics data and improve its services. Ask how those two statements fit together.

3121314

What to test before a pilot (editorial suggestion)

These are our recommendations, not vendor features. First, route a real call from the agent into your own queue and look at what the human sees. Second, ask each vendor to show a transcript where sensitive passages were removed, and who can open recordings and logs. Third, request written wording on training and analytics use of your calls. Customer stories, such as Hawksmoor or ECSI, illustrate workflows; they do not prove that a function will work on your stack.

1314