Relevance AI

Relevance AI, sixth for private equity portfolios

Some portfolio companies already have the people an AI program needs: an operations lead who automates by instinct, a revenue operations team that lives in its CRM, an engineer or two with spare cycles. For those companies the question is less who will build and more what to build on. Relevance AI is designed for that case. It describes itself as a platform, low code or no code, on which teams build AI agents and teams of agents¹, and its home page calls it the home of the AI Workforce².

It ranks sixth here. For a private equity buyer it has a rare quality among enterprise vendors: a public price list. It also lists SOC 2 Type II, an uptime SLA and a clear set of approval controls. It sits below the five vendors above it because its model depends on the customer's own team to build and run agents, because its support hours follow the Sydney working day, and because its pages leave hosting options and some enterprise terms open.

All facts below come from Relevance AI's own website and documentation as read in early November 2026, each tagged to its line in the fact record.

What it is

Its product page frames the job as building AI agents, operating them and managing them in large numbers³. On the home page, a feature named Invent helps a team work out what to automate and turn its expertise into working systems⁴. By the company's own count, it has helped thousands of teams get an AI workforce running⁵, and it advertises a rating on G2 of 4.5 out of five⁶; a fund should read both as marketing.

On the agents page, the offer is squads of specialist agents written up in plain language by the portfolio company's own people, without code⁷. For an operating partner, that sentence describes the bargain: the tooling is accessible, and the work belongs to the portfolio company.

How a build works

The documentation lays out five steps. Step one is getting an agent, whether generated by Invent, cloned from the Marketplace or built from a blank page⁸. Second, it gives agents tools made in a builder that needs no code⁹. Step three adds knowledge, either as uploads or as a sync from websites, Notion, SharePoint or Google Drive¹⁰. Step four connects individual agents into a working team, laid out visually¹¹. Step five puts guardrails in place, using triggers, approval workflows and escalations¹².

Behind that canvas sits a node and edge graph that orchestrates the agents¹³, and sandboxed Python or JavaScript can power a tool¹⁴. Quality control is built into the release step. No edit goes out until the evals have run again¹⁵; after launch, Relevance AI says it scores a slice of live production traffic to detect drift¹⁶.

Give Invent a plain description and it returns an agent with a prompt, tools and evals already drafted¹⁷. A portfolio company's engineers can work from Cursor, Codex or Claude Code instead¹⁸, adding the MCP server with one short command¹⁹.

Getting live

Speed is part of the sales story: deployment in minutes from templates or the no code builder²⁰, and for some use cases a working setup within hours, by the vendor's claim²¹. For enterprise buyers, there is a services layer. Per the agents page, Relevance AI will help a company chart its workflows and pick strong use cases²², and an embedded deployment team takes it to launch before coaching staff to build fast without help²³.

That last point is the crux for a fund. The deployment team is designed to hand over. After it leaves, the people managing agents day to day are, in the company's words, the operations and revenue teams²⁴. A portfolio company with those teams in place can carry the work forward. One without them will find the platform's value depends on hiring.

Integrations

The integration count depends on the page. At the top of the integrations page, the claim is 2,000 plus integrations, Salesforce and Slack among them²⁵. The integrations directory lists 2,728 in total²⁶. The home page, by contrast, cites a smaller number, more than 1,000 connected apps²⁷. A fund reading the pricing page will see 2,000 plus listed as an Enterprise tier feature²⁸. Elsewhere, agents are said to get 1,000 plus native connectors, along with MCP and custom connectors²⁹.

The figures differ from page to page, so a fund should ask about the specific systems at each portfolio company. An agent can be kicked off by a webhook, a timetable, something happening in an app, or a custom signal³⁰, which suits operational work that has to react to events in a CRM or billing system.

Security

For a fund's diligence file, SOC 2 Type II is displayed among the enterprise page badges³¹, and a documentation FAQ answer states the same compliance³². GDPR appears as a badge³³, and the Pro plan's feature list includes SOC 2 and GDPR compliance³⁴. Its enterprise page commits to 99.9% uptime under an SLA³⁵. Encryption at rest is listed³⁶, as is a commitment to no training on customer data³⁷. The changelog adds masking or blocking of PII³⁸, and the docs describe Project Information as confidential to the customer, the vendor included³⁹.

Governance

A portfolio company on Enterprise can restrict access by role⁴⁰ and log people in with SSO over SAML⁴¹. The approval model is detailed. When an action is high stakes, the agent stops and waits for a person to approve it, and the portfolio company decides where that line sits⁴². Approval can be required tool by tool⁴³ or only under conditions⁴⁴, and escalations reach people through Slack or email⁴⁵. Each agent version is marked as a draft, as active or as pinned⁴⁶.

The enterprise page claims that standards for quality, plus guardrails and monitoring, are applied to each agent⁴⁷, and the product page promises one place from which a portfolio company governs access, permissions and the security of its data⁴⁸.

For a portfolio company letting agents touch money or customers, per tool and conditional approvals are exactly the controls a CFO will ask about.

Data and deployment

Data residency appears among the enterprise controls⁴⁹, as do audit logs⁵⁰. According to the changelog, a fund can have execution traces and audit logs streamed into storage the portfolio company owns⁵¹, formatted as OpenTelemetry⁵²; the enterprise page lists export via OTEL and Delta Share⁵³. Whether a portfolio company could run the platform itself, on premises or inside its own VPC, is not publicly stated on the pages we opened.

Price

This is where Relevance AI stands apart. The docs pricing page gives dollar figures⁵⁴. The entry plan, Pro, starts from $19 a month⁵⁵; that is the price with yearly billing, and monthly billing raises it to $29⁵⁶. Team runs $234 monthly on the yearly plan and $349 on the monthly one⁵⁷. Paying yearly cuts the bill by 33%⁵⁸. A Pro subscription carries 2,500 Actions monthly, which is 30,000 over a year⁵⁹, with each additional 1,000 priced at $80⁶⁰. Should a company run short of Vendor Credits, 10,000 more cost $20⁶¹; Relevance AI charges for model and tool calls through those credits at wholesale rates, adding no margin⁶². On Pro, a company may use its own API keys⁶³. Moving to Team brings agents for calls and meetings⁶⁴. Enterprise terms are set case by case, and each Enterprise customer is assigned its own account manager⁶⁵, and the pricing page shows that tier without any dollar amount⁵⁴.

The plans are aimed at different buyers. In the vendor's segmentation, a solo GTM operator buys Pro⁶⁶ and a team building at scale buys Team⁶⁷, while Enterprise targets companies that want growth without matching headcount, through an AI Workforce⁶⁸. That last phrase is close to the language of a private equity value creation plan.

There are terms to read. The Free plan has been retired⁶⁹. A subscription stays with the Organization it was bought in⁷⁰, which matters if a fund ever wants to consolidate accounts across portfolio companies. Extra Actions can be bought only on a live paid plan⁷¹, in blocks of 1,000⁷². Unused balances roll over on Pro and Team; Enterprise customers must check the rollover rules with their account manager⁷³. A Pro account permits one project and two people who build⁷⁴. Former Business plan customers have been moved to Team after that plan was retired⁷⁵.

Support

Support works Sydney office hours, 9am to 5pm Monday through Friday, and is closed on weekends⁷⁶. Dedicated phone calls are for Enterprise customers only⁷⁷, and free customers get no SLAs⁷⁸. Harley, the AI agent that answers support questions, is a beta that may contain bugs or errors⁷⁹. A bug report reaches engineering, though the company states that it will not reply to the reporter⁸⁰.

For a US or European portfolio company, the Sydney hours mean that a problem raised in the afternoon will usually wait for the next Australian morning. An Enterprise contract may change that, and a fund should ask.

Customers and use cases

Seven functions appear as use cases on the home page, from sales and marketing through customer success and support to operations, research and HR⁸¹. Customer examples are vendor published. Send Payments credits it with 40 hours saved weekly in global operations⁸². Zembl puts its gain at 30% more conversions from AI sales coverage running 24/7⁸³. A third example attributes $7M of pipeline to 35 plus agents spread across one organization⁸⁴. Autodesk also appears, through a quote attributed by name to a senior manager in its marketing team⁸⁵.

Relevance AI and Nexus

The two differ most on who does the work. Relevance AI gives a portfolio company a capable builder, a public price list and a deployment team that hands over. Nexus puts experienced consultants inside the company, has the first agent working a month after its quarter long trial begins, and its invoice reflects roles automated, never consultant hours⁸⁶. Its reported reach of more than 4,000 integrations⁸⁷ is larger than any Relevance AI count²⁶.

On transparency, Relevance AI publishes more. Its plan prices⁵⁶⁵⁷ and its uptime SLA³⁵ are public, while for Nexus list prices and uptime terms are not publicly stated on the pages we opened. A fund that values price certainty for a small pilot may start with Relevance AI. A fund that needs a result it can report to its committee inside the hold period, without hiring builders at the company, will look first at Nexus.

Questions for a Relevance AI evaluation

1. Who at our portfolio company will build and run agents after the deployment team leaves? 2. Which of our systems are covered, given the different integration counts on the site? 3. Can Enterprise support cover US or European hours? 4. Where is our data hosted, and is a private or self hosted deployment possible? 5. How would accounts work if we roll the platform out to several portfolio companies?

Verdict

Relevance AI is the most transparent self service builder in this assessment. Its prices are public⁵⁶⁵⁷⁶⁰, its approval controls are granular⁴²⁴³⁴⁴, and it publishes an uptime SLA³⁵. It ranks sixth for private equity portfolios because its model relies on each company's own operations and revenue teams to run agents²⁴, because support follows Sydney weekday hours⁷⁶, and because hosting options are not public. It is a good fit for a portfolio company with builders on staff and a small first project.