Skip to content

Relevance AI

Platform for building teams of AI agents for sales, research and back-office work, now sold to enterprises by quote.

Reviewed by Andrew Miller · Updated

Our rating

3.6/5

A capable multi-agent platform for sales and operations teams with a procurement process, sold in a way that shuts out anyone who wants to test it first.

What it is

Relevance AI lets companies build and run groups of AI agents, which it calls an AI Workforce. Agents use tools and integrations for work such as prospect research, outbound sales, data enrichment and call or meeting handling, and start from CRM, email and calendar triggers, webhooks or schedules. Teams test agents with evaluations and A/B tests and can require human approval above set thresholds.

In 2026 the company moved upmarket. Its pricing page now shows a single Enterprise plan sold through sales, with unlimited agents, users and projects, custom Actions and vendor credits, SSO, role-based access and audit logs. The free, $19 Pro and $234 Team plans offered in April 2026 were gone by June.

The company, based in San Francisco and Sydney, raised a $24 million Series B led by Bessemer Venture Partners in May 2025. Billing is based on Actions (each tool run by an agent) plus vendor credits for model costs.

Our review

Relevance AI is now an enterprise purchase, and its security paperwork is in better order than many agent builders'. Its docs claim SOC 2 Type II and GDPR compliance, keep data in a region chosen at sign-up (US, London or Sydney), and say customer data is not used to train models without a specific partnership agreement. The enterprise page adds PII masking and a 99.9% uptime SLA, though the SLA's terms are not published.

The product is built for agents that act rather than just answer. Agents start from CRM, email and calendar triggers, webhooks or cron, can hand work to each other, and can be held for human sign-off, for example on a refund above a $500 limit. Evaluations, A/B tests, full tracing and OpenTelemetry export are the controls a team needs before letting agents write to Salesforce.

Cost is the weak point. There is no public price, and independent reviews say the two-meter model, Actions for every tool call plus vendor credits for model usage, made bills hard to forecast even when prices were public. The pricing page now lists both meters as custom, so the forecast depends entirely on the contract.

The official pages do not always agree with each other: the pricing page claims more than 2,000 integrations, the enterprise page more than 1,000. The data region is also effectively permanent; the docs say it cannot be changed after the organisation is created without going through support, which is worth settling before the first agent goes live.

Where it earns its place

  • A 40-rep outbound sales team on Salesforce

    Research accounts and draft first-touch emails before reps call

    A BDR-style agent pulls prospect data, enriches records and drafts outreach, with approval gates before anything is sent. Every tool call consumes Actions on top of model credits, so negotiate both meters and an overage cap in the contract.

  • A RevOps engineer wiring agents into the sales stack

    Trigger agents from CRM events and monitor them in production

    Use native CRM triggers or webhooks, connect custom tools through MCP, and export traces over OpenTelemetry into the existing observability stack. Choose the data region carefully at sign-up, because changing it later needs support.

  • A security reviewer at a European fintech

    Approve an agent vendor that will handle customer personal data

    Ask for the SOC 2 Type II report under NDA, choose the London region, and confirm PII masking and the no-training clause in the contract. The 99.9% SLA's credits and exclusions are not published, so get them in writing.

Before you commit

Setup
Sales-led: book a demo, agree a contract and implementation timeline, then choose a data region (US, London or Sydney) when the organisation is created.
Integrations
CRMs including Salesforce and HubSpot, Slack, Gmail, Snowflake, Databricks, Zendesk and GitHub, plus MCP and custom connectors; official pages claim 1,000+ or 2,000+ integrations.
Data and privacy
SOC 2 Type II and GDPR per the docs; TLS 1.2+ in transit and AES-256 at rest; no training on customer data without a partnership agreement; SSO, MFA and RBAC on Enterprise.
Cost at scale
Not published: Actions and vendor credits are both custom on the Enterprise plan, so a production workload's cost depends on negotiated volumes.
Limits
The data region cannot be changed after the organisation is created without contacting support, and there is no free tier to trial agents before buying.

What's good

  • SOC 2 Type II, GDPR and US, EU or Australian data residency
  • Human approval gates and agent evaluations
  • Ready-made sales agents such as the BDR agent
  • Triggers from CRMs, email, webhooks and cron

What's not

  • No public pricing or free plan since mid-2026
  • Two billing meters, both custom-priced
  • Data region fixed at sign-up
  • Conflicting integration counts on official pages

Best for

Mid-size and large sales or operations teams that want a supported multi-agent rollout.

Alternatives

Relevance AI: common questions

Is Relevance AI worth it?
A capable multi-agent platform for sales and operations teams with a procurement process, sold in a way that shuts out anyone who wants to test it first. We rate it 3.6 out of 5.
How much does Relevance AI cost?
Enterprise only on the public pricing page, priced by sales with custom Actions and vendor credits; the free, Pro ($19/month) and Team ($234/month, yearly) plans were removed in mid-2026.
What are the downsides of Relevance AI?
No public pricing or free plan since mid-2026; two billing meters, both custom-priced; data region fixed at sign-up; conflicting integration counts on official pages.
Who is Relevance AI best for?
Mid-size and large sales or operations teams that want a supported multi-agent rollout.
How much does Relevance AI cost for a team or in production?
Not published: Actions and vendor credits are both custom on the Enterprise plan, so a production workload's cost depends on negotiated volumes.
Is Relevance AI safe for company data?
SOC 2 Type II and GDPR per the docs; TLS 1.2+ in transit and AES-256 at rest; no training on customer data without a partnership agreement; SSO, MFA and RBAC on Enterprise.
What are the best Relevance AI alternatives?
The closest alternatives we have reviewed are Gumloop, Lindy and n8n.

Answers checked Oct 22, 2026.