Program report · filed November 14, 2026

CRM & Sales TechMeasured report

Retell AI API Integration: What G2 Reviews and HubSpot Workflows Reveal

G2's 13-month review analysis finds Retell AI's APIs praised for usability, with a HubSpot workflow revealing hard limits: 10-second webhook windows, three retries, US-only calling.

By Marcus Bennett4 min read761 words

Program notes

  1. G2 analyzed reviews submitted October 1, 2025 – October 31, 2026 and verified technical claims against vendor docs in November 2026
  2. HubSpot disabled new legacy private-app creation for existing accounts on October 26, 2026; Service Keys are the recommended replacement
  3. Retell expects a 2xx webhook response within 10 seconds and retries failed deliveries up to three times
  4. Retell-purchased phone numbers are currently limited to US destinations
  5. An optional idempotency_key prevents duplicate calls when retrying the same request within one hour
Retell AI API Integration: How to Connect It to Your Apps
PlateRetell AI API Integration: How to Connect It to Your Apps — StockSnap / pixabay

G2 analysts reviewed user feedback submitted between October 1, 2025 and October 31, 2026 to map how Retell AI, a voice-agent platform, actually connects to business systems — and where the integration path gets rough. The verdict: developers praise the API's accessibility, but several reviewers want deeper advanced documentation and more native connectors.

Retell provides APIs for building voice agents, managing calls, and retrieving conversation data that other applications can consume. It's listed in G2's AI Voice Agent Platform Software category and supports voice automation across sales, customer service, and business operations. The methodology behind the review is explicit: G2 checked technical claims against official Retell and HubSpot documentation in November 2026, and the review examples are qualitative, not survey data.

What changes for sales teams?

The core value proposition is timing. Retell's integration stack splits into two distinct mechanisms: custom functions let an agent query external services during a live conversation (checking appointment availability, pulling authorized customer data), while webhooks push call outcomes to connected systems afterward (updating a CRM contact with lead-qualification results). Getting that distinction right determines what your workflow can do.

The documented use cases span five areas:

  • CRM and lead qualification — initiate calls and write qualification results back to lead records
  • Appointment scheduling — check availability and book during the conversation
  • Customer support — look up order status and create tickets mid-call
  • Call analytics — retrieve transcripts, summaries, and structured analysis
  • Workflow automation — trigger downstream actions on call events like a callback request

What do G2 reviewers actually report?

The measured signal is consistent. One reviewer describes Retell's API as easy to integrate and well-documented for developers. Another highlights how easily it slots into an existing product. A reviewer building customer-support agents describes using custom tools to query databases, create tickets, and trigger workflows during live conversations — an agent interacting with business systems rather than relying only on its initial instructions.

On the outbound side, a reviewer describes extracting booking status, objections, and other structured results and passing them to a CRM through webhooks.

The asserted criticisms are narrower. Reviewers ask for clearer documentation on custom analysis, tool calling, and debugging. One who experimented with post-call webhooks and custom analysis schemas wanted better troubleshooting information for failed calls. Another wanted more extensive tool-calling docs. A sales-automation reviewer wanted a direct GoHighLevel integration — which Retell has since documented, so that gap appears closed.

G2 flags these as individual experiences during the research period, not confirmed current product limitations.

What does the HubSpot workflow expose?

The walkthrough — Node.js code checked against both vendors' docs but not tested with live credentials — surfaces hard operational constraints that matter more than the code itself:

  • HubSpot disabled new legacy private-app creation for existing accounts on October 26, 2026; the custom integration uses Service Keys with a crm.objects.contacts.write scope
  • Retell expects a 2xx response within 10 seconds and retries failed webhook deliveries up to three times
  • Retell-purchased phone numbers are currently limited to US destinations
  • An optional idempotency_key prevents duplicate calls when retrying within one hour

Two design decisions stand out. The example handler accepts only calls ending in agent_hangup or user_hangup, excludes voicemail unless Retell explicitly returns in_voicemail: false, and skips dial_failed-type outcomes — because AI-generated qualification values shouldn't auto-label a dial tone as an unqualified lead. And the in-memory event deduplication is explicitly labeled demo-only; production needs durable storage plus a separate worker, since simultaneous webhook deliveries can both pass a Set-based check.

A data-privacy note is easy to miss: Retell documents that transcripts and recordings may remain accessible through webhook delivery even when post-call storage is disabled, so payloads need access and retention controls.

What should buyers take from this?

For teams with standard CRM workflows, Retell's native HubSpot connector may be enough — though G2 notes the native setup docs still describe a legacy private-app token, so confirm the supported auth method before configuring. Custom API routes buy control over call initiation, event processing, and CRM writeback, at the cost of building queue infrastructure, retry handling, and signature verification yourself.

API and webhook access is included in Retell's pay-as-you-go tier, but voice agents aren't free — charges depend on usage, models, and telephony. The practical takeaway, grounded in the G2 analysis: define the workflow first, pick native versus custom based on the data your agents need mid-call versus post-call, and validate the full pipeline with test records before production.

via learn.g2.com (Source)

Filed under

Share this article:

More from Marcus Bennett

Marcus Bennett

Show full bio

Staff writer covering people, launches and day-to-day practice in Sales technique and methodology, sales enablement at Quota Brief.

120 articles

Also rated

  1. Monday.com Ships Harmony, a Voice AI Agent Platform for Sales
  2. Salesforce’s Cresta Acquisition Puts Voice AI Coherence to Test
  3. Reevo Ships Mobile AI Sales App for Selling Beyond the Desk
  4. AppDirect Buys Soul Machines to Put AI Avatars in B2B Sales
  5. Monday.com launches Harmony, a voice AI platform for sales calls

« Previous articleNext article »