Relevance AI

Relevance AI: a no-code builder with a door left open for engineers

Relevance AI sits close to the no-code end of the build spectrum, and its documentation agrees, describing a low/no-code product for single agents and whole multi-agent teams¹. On the product page the promise grows to building, running and managing agents in large numbers², while the home page headline speaks of a home for the AI Workforce³. For a Head of Digital deciding which build style suits the company, Relevance AI is the clearest example on this site of a tool designed for the people closest to the work. It is also one of the few no-code products that gives engineers a route into the same platform from their own coding tools.

This page reviews the product from that angle: who builds on it, how the build moves from idea to production, what controls a central team gets, and what the published price list means for a department that wants to start small. It is ranked second on this site. The reasons for that placement, and the reasons it sits behind Nexus, appear in the verdict.

Who the product is for

Day-to-day agent management, in the vendor's telling, belongs to operations and revenue staff, since they sit nearest the work⁴. The agents page pictures groups of specialist agents that a customer's own staff write in everyday language, without code⁵. The plans follow the same audience. Pro is for solo GTM operators⁶. Team targets groups that build in volume⁷. Enterprise is pitched at firms that want growth without a matching rise in headcount, using an AI Workforce⁸.

Home page use cases span seven functions, Human Resources among them, next to Sales, Marketing, Operations, Research, Customer Success and Customer Support⁹. The vendor credits itself with thousands of AI workforce launches¹⁰ and shows a 4.5-star G2 score¹¹. Those are its own figures.

How an agent gets built

The documentation lays out five steps, and they map well onto how a business team would work through a first project.

The first step is the agent itself. A team can build with Invent, clone an agent from the Marketplace or start from scratch¹². Invent stands out among the three. A user writes a plain description, and Invent produces the agent with its prompt, tools and evals¹³. The home page adds that Invent helps your people identify what to automate¹⁴, which suggests the vendor sees discovery as part of the build.

The second step is tools. Agents get capabilities from a no-code tool builder¹⁵. For teams that need more than blocks, tools can run sandboxed Python and JavaScript¹⁶, which is where the low-code side of the product shows.

The third step is knowledge. Teams can upload files directly or keep content in sync from Notion, SharePoint, Google Drive or a website¹⁷.

The fourth step joins agents into workforces. Several agents can be linked on a visual canvas¹⁸, and the product page describes the orchestration as node-and-edge graphs¹⁹.

The fifth step sets guardrails through escalations, approval workflows and triggers²⁰.

Quality control is built into publishing. No change goes live until the agent's evals have run again²¹, and a sample of live runs keeps being scored afterwards so that drift shows up²². Agents carry draft, active and pinned versions²³. For a Head of Digital who has to explain to a risk committee how a business-built agent is tested, an eval gate on every change is a concrete answer.

The engineer's route

Most no-code builders stop at the canvas. Relevance AI adds a second door. Developers, the product page explains, can operate the same platform from Cursor, Codex or Claude Code²⁴; the MCP setup is a single command, mcp add relevance-ai²⁵. In build-style terms, this means a company can let business users build on the canvas while its developers work on the same agents from their editors. A buyer comparing Relevance AI with a code-first framework should test how far that route goes on a real use case, because the pages we opened describe the entry point and stop there.

Time to first agent

The vendor's speed claims are short. The documentation says agents can be deployed in minutes using the no-code builder or pre-built templates²⁶. For certain use cases, the agents page puts setup at just a few hours²⁷. Both describe the time to a working agent on the platform. Neither describes the time to an agent wired into a company's systems and approved for production, which is the date a Head of Digital will be asked for.

Delivery support

Relevance AI offers people as well as software. According to the agents page, a deployment team works inside the customer until launch and then coaches the customer's staff to build fast without help²⁸; the vendor also assists with mapping workflows and choosing the strongest use cases²⁹. The Enterprise tier includes a dedicated account manager³⁰.

This is a real strength among no-code tools, and it is the closest match on this site to the delivery model Nexus runs. The difference lies in what the buyer ends up doing. Relevance AI's stated end state is a customer team that builds on its own²⁸. In the Nexus model, consultants keep building with the customer, production begins at week 4 of a 3-month proof of concept, and the charge is per role automated (owner decision³¹).

Support outside that team has stated limits. Only Enterprise accounts can book calls with dedicated staff³². Help is available Monday to Friday from 9am to 5pm on Sydney time, and weekends are not covered³³. SLAs are not offered to free customers³⁴. Bug reports go to engineering, but the support page says the vendor does not respond to them³⁵. Harley, the AI support agent, is still a beta and may make mistakes³⁶.

Integrations and triggers

The integration story is broad, and the figures differ by page. The integrations page headline offers 2,000+ integrations, from Slack to Salesforce³⁷, and the directory itself shows 2,728 in total³⁸. The home page says Relevance connects to 1,000+ apps³⁹. On the product page, agents have over 1,000 native connectors, MCP support and room for connectors a customer writes itself⁴⁰. The pricing page lists 2,000+ integrations on the Enterprise tier⁴¹. With several counts in circulation, a buyer should check the specific systems by name.

Triggers start an agent from a schedule, an event in your apps, a webhook or a custom signal⁴². That range is what a business team needs to move from agents someone chats with to agents that run on their own.

Security and certifications

Relevance AI publishes a clear security baseline. SOC 2 Type II appears both as a badge on the enterprise page⁴³ and as a compliance statement in the documentation FAQ⁴⁴. GDPR appears as a badge⁴⁵. The page states a 99.9% uptime SLA⁴⁶, encryption at rest⁴⁷ and no training on your data⁴⁸. Per the documentation, Project Information stays confidential, with Relevance AI itself kept out⁴⁹. The Pro plan lists SOC 2 and GDPR compliance among its features⁵⁰. The changelog adds a feature to mask or block PII such as email addresses and phone numbers⁵¹.

For a no-code tool that business users will adopt, a stated uptime SLA and a SOC 2 Type II report are useful inputs for a security review. The badges Relevance AI shows are SOC 2 Type II and GDPR^43,45; Nexus adds ISO 27001 and ISO 42001 to those two (owner decision³¹).

Governance and access

Central control is described in some detail. Enterprise customers get role-based access control⁵² and SSO / SAML⁵³. High-stakes actions pause for human sign-off, and the customer sets the thresholds⁵⁴. Approval rules can be set for each tool⁵⁵ or made conditional⁵⁶, escalations can go out by Slack or email⁵⁷, and one console covers access, permissions and data security⁵⁸. Each agent, the enterprise page says, gets quality standards, guardrails and monitoring⁵⁹.

Enterprise customers also get audit logging⁶⁰. Logs can leave the platform too. The changelog covers streaming audit records and execution traces into storage the buyer controls⁶¹, with OpenTelemetry as one format⁶², and the enterprise page names OTEL and Delta Share export⁶³. For a Head of Digital who expects many business teams to build, this is the kind of control set that lets a central team watch without blocking.

Hosting and data residency

Data residency is one of the enterprise options named⁶⁴. Running the platform on-premises, in a customer VPC or as a self-hosted install is not publicly stated on the pages we opened. A company that requires its agent builder to run inside its own cloud account will need a different answer, and Dify⁶⁵, n8n⁶⁶ and CrewAI⁶⁷ each document such options on their own pages.

Pricing

Relevance AI publishes its small-team prices. Pro starts from $19 per month⁶⁸: $19 per month on annual billing, or $29 billed monthly⁶⁹. Team is $234 per month annually or $349 monthly⁷⁰. Annual billing saves 33% over the monthly rate⁷¹. Enterprise terms are quoted case by case⁷², and each account gets its own account manager³⁰.

The unit of consumption is the Action. A Pro subscription carries 30,000 Actions per year, equal to 2,500 monthly⁷³. Additional Actions are sold in blocks of 1,000⁷⁴, each block priced at $80⁷⁵. Model and tool usage is billed through Vendor Credits at wholesale cost without markup⁷⁶; another 10,000 credits cost $20⁷⁷. On Pro, customers may supply their own model API keys⁷⁸. Calling agents and meeting agents arrive with Team⁷⁹.

Plan limits are worth reading closely. Pro is limited to a pair of build users working in one project⁸⁰. Top-ups require an active paid subscription⁸¹. Subscriptions cannot move from one Organization to another⁸². Rollover rules are written down for Pro and Team, while Enterprise buyers are told to check theirs with an account manager⁸³. No free plan remains for new or current accounts⁸⁴, and the old Business plan was closed, its customers moved onto Team⁸⁵.

The consequence for budgeting is that cost follows activity. A department that scales from one agent to a workforce will see its Action consumption grow with usage, which is easy to model for a steady workload and harder for a new one. Nexus prices per role automated (owner decision³¹), which ties the bill to the scope of work automated.

Customer evidence

The home page carries several customer results, all published by the vendor. Send Payments saves 40 hours weekly by going AI-first across global operations⁸⁶. Zembl reports 30% more customer conversions with 24/7 AI sales coverage⁸⁷. Qualified is cited for $7M pipeline from more than 35 agents across the organization⁸⁸. A testimonial comes from a senior marketing manager at Autodesk⁸⁹. The pattern is revenue and operations work, which matches the audience the vendor names⁴.

Strengths

Relevance AI gives business teams a genuine no-code path, from a plain-language description to a tested agent¹³, with eval suites that re-run on every change²¹. It adds an engineer's route through MCP and coding assistants²⁴. Its security baseline includes SOC 2 Type II and a 99.9% uptime SLA^44,46. Its approval rules are granular^55,56. Its price list is public⁶⁹, and an embedded deployment team is available²⁸.

Limits

The integration figures differ between pages^37,39. Self-hosting is not publicly stated on the pages we opened. Support runs in Sydney weekday hours with no weekend cover³³, and dedicated calls are for Enterprise customers³². Pro is capped at two build users and one project⁸⁰, and there is no free plan⁸⁴. Speed claims describe build time on the platform²⁶, which is a different measure from time to a governed production release.

Verdict

Relevance AI is ranked second. For a Head of Digital who wants business users to build and a central team to govern, it is the strongest pure no-code option among the seven, with a documented eval gate, granular approvals and an engineer's route into the same platform. It ranks behind Nexus on delivery and commercial model. Nexus consultants take the customer to a production agent at week 4 of a 3-month proof of concept, the price follows the role automated, and the Nexus certifications add ISO 27001 and ISO 42001 (owner decision³¹). Nexus also publishes the larger integration count⁹⁰.