Relevance AI: our review
Relevance AI has specialized agents built — one for prospecting, one for support, one for meeting prep — then has them work together as a team. A single plan, entirely quote-based, published by an Australian company.
What we take away from it
What counts in its favor
- Agents designed for a single, well-defined task rather than a general-purpose assistant: easier to evaluate and fix when an agent gets something wrong.
- Built-in orchestration (routing, queuing, evaluation, traceability) inside the same tool, instead of assembling several separate pieces yourself.
- More than 2,000 integrations announced, enough to plug an agent into most of the business software already in place.
What counts against it
- A single plan, Enterprise, with no public price list: impossible to compare a price without going through a sales meeting.
- No clear commitment found on training models from submitted data: neither the privacy policy nor the security page settles the question.
- A trust portal is advertised (trust.relevanceai.com) but its content cannot be read without creating access to it: impossible to check its certifications from the public documentation.
- No mention of French-language support, either on the site or in the pages reviewed.
Useful for a company that wants to deploy several coordinated business agents and negotiate a custom contract. Skip it if you want a listed price before starting the conversation, or a written answer on model training before connecting your prospecting or support data.
The facts
What happens to your data
- Is your data used to train their models?
- Subprocessors
- Neither Relevance AI's privacy policy nor its data security page explicitly confirms or rules out training models on data submitted by customers: the question remains unanswered in writing in the public pages reviewed. The publisher, OnSearch Pty Ltd (trading as Relevance AI), is an Australian company; its sub-processors are located in the United Kingdom, the European Economic Area, the United States and Australia. A trust portal exists (trust.relevanceai.com) but its content loads via JavaScript and could not be verified by an automated reader.
- Retention
- Kept under an "internal retention policy", with no published numerical duration, except where the law requires otherwise.
Relevance AI in detail
Relevance AI builds agents dedicated to a specific task — finding leads, preparing a meeting, chasing a contract — then has them work together like a team: one agent enriches a record, another schedules the appointment, a third drafts the write-up.
What it actually does
Rather than a generalist assistant asked to do everything, Relevance AI has you build agents each confined to a narrow task: a research and enrichment agent, a scheduling agent, an agent that acts after a call. These agents then coordinate within a "team" or a "workforce," with task routing, a queue, quality evaluation of responses and traceability of actions — all managed from the same platform rather than assembled from separate tools.
A typical use case: a prospecting agent identifies a company matching a target profile, an enrichment agent completes its record from public sources, and a third agent drafts a message before a salesperson approves it.
Why it interests us
Breaking a business problem down into narrow agents, each evaluated on a specific task, matches how we design our own AI agent deployments: one agent that does one thing well beats one agent that promises to do everything, badly. But the question that matters before signing remains the same as for any agent touching prospecting or customer support: where do these agents run, what data do they receive, and what does the publisher commit — or not — to not doing with it. On this specific point, Relevance AI's public documentation stays vague.
Where it falls short
First point, and the most restrictive in practice: no price is published. Only one plan exists, Enterprise, entirely on quote — impossible to know before a sales call whether the budget matches the project's scale.
Second point: the question of training models on submitted data remains unanswered in writing. Neither the privacy policy nor the data security page addresses it explicitly, one way or the other. A trust portal exists at trust.relevanceai.com, but its content loads via JavaScript and could not be verified through an automated read — which proves nothing in itself, but leaves the question open until it has been put directly to the sales team.
Third point: the publisher's subcontractors, an Australian company, are spread across the United Kingdom, the European Economic Area, the United States and Australia, with no guarantee of a single default region.
Our recommendation
Relevance AI makes sense for a business that wants to deploy several coordinated business agents and has the capacity to negotiate a bespoke contract. For an SME that wants to compare a published price before starting the conversation, or that needs a written answer on model training before connecting prospecting or support data, it's best to put the question in writing before any trial.
Who it's for, and who it isn't
Suited for
- A company with several agent use cases (prospecting, support, operations) that wants to coordinate them rather than multiply them in silos
- An organization ready to negotiate a contract rather than self-serve a subscription
- A team that has already mapped its repetitive tasks and wants one agent per task rather than a single assistant
Not a fit
- An SMB that wants a listed price before talking to a salesperson
- An organization that needs a written, verifiable answer on model training before any commitment
- Anyone looking for a French-language tool: no French-language support was found
Frequently asked questions
- Do you need to know how to code to build an agent on Relevance AI?
- Not for the basics: the platform presents itself as a no-code agent builder, with ready-to-adapt agents (prospect research, meeting prep, contract follow-up). API access and custom integrations exist for advanced cases, but aren't required to get started.
- Is my data used to train an AI model?
- The public documentation doesn't say clearly. Neither the privacy policy nor the data security page explicitly mentions training models from data submitted by customers. With no written answer available, ask the sales team directly before connecting sensitive data.
- How much does Relevance AI cost for a team?
- Only one plan exists, Enterprise, with no public amount: the price is negotiated case by case based on the agents deployed and usage volume. Impossible to compare a figure without a sales contact, unlike tools that display a first tier.
- Where is my data processed?
- Relevance AI shares data with sub-processors located in the United Kingdom, the European Economic Area, the United States and Australia, according to its privacy policy. No single region is guaranteed by default; the detail per customer would depend on the negotiated Enterprise contract.
- What's the difference with Gumloop or AutoGPT?
- Gumloop and AutoGPT offer a canvas where you build an agent's logic yourself, with a first self-service tier. Relevance AI starts from already-packaged business agents (prospecting, support, operations) that you assemble into a team, with no public price: the trade-off is less visual tinkering, but no price before talking to a salesperson.
Alternatives to Relevance AI
AI agents · Automation
Gumloop
Tool for building AI agents that automate tasks by connecting to Slack, Gmail and other business software.
AI agents
AutoGPT
Platform for building and running autonomous agents that can run task chains with no constant supervision.
AI agents · Automation
Lindy
AI assistant that carries out office tasks (emails, calendar, customer support) via no-code configurable agents.
Project management · Productivity
Taskade
Combines to-do lists, mind maps, and autonomous AI agents in a single workspace.
Our sources
- Pricing (Enterprise plan, quote-based)accessed on 30 July 2026
- Privacy policy (sub-processors, publisher identity)accessed on 30 July 2026
- Data security policyaccessed on 30 July 2026
Prices and hosting terms change often. We recheck them periodically, but the publisher's page remains the only up-to-date reference.