Ask two different teams "does your company use AI at work" and one answers from inside Jira, the other answers from inside Slack. Atlassian built Rovo to be the first answer: generative AI woven into the tools a team already opens to track a sprint or write a page. Sanaf is built to be the second: one AI employee hired into the channel a team already talks in, holding a job description and working it over days and weeks. Both are real answers to "AI at work." They are not the same question.
This is a fact-checked comparison. Atlassian's own Rovo pricing, product and connector pages were read the same week this was written, because Rovo's credit-based pricing and its connector list have both changed since launch, and a comparison built on old numbers would misdescribe what a team actually gets today.
What Atlassian Rovo actually is
Rovo is Atlassian's generative AI layer, built to "unlock organizational knowledge with GenAI" across the company's own tools. It ships in four pieces. Rovo Search and Rovo Chat answer natural-language questions against a company's connected content. Rovo Agents are automations that carry a piece of work forward, drafting a next step or flagging something stalled, inside Jira and Confluence workflows. Rovo Studio is where an admin or a partner builds a custom agent for the organization's own process.
All four sit on top of Atlassian's Teamwork Graph, the index that connects "teams, work, and goals" across a company's tools. Rovo ships with pre-built connectors for 50-plus apps: Google Drive, Slack, Microsoft Teams, GitHub, Figma, SharePoint, Bitbucket, Loom, Salesforce and Notion among them, with Rovo Search able to query across Jira, Confluence, Jira Service Management and dozens of connected third-party sources from one search bar. There is a Rovo app in the Slack Marketplace, so a question can be typed from Slack and answered by an index that spans a company's other tools.
Rovo is not sold as its own subscription. It is included in Standard, Premium and Enterprise Cloud plans for Jira, Confluence, and the Teamwork and Service Collections, at monthly AI credit allowances that scale with the plan: 25, 70 or 150 credits per user per month on standalone Jira or Confluence, and 250, 700 or 1,500 credits per user per month on the bundled collections. A basic AI interaction runs about 10 credits; a premium interaction or an enriched API call costs more. Credits pool at the organization level, reset each billing date, and extra usage beyond the monthly allowance bills at roughly a cent per credit, with admin alerts at 80 and 100 percent of the pool and an option to cap or disable overage spending outright. It is a serious, well-integrated product for the job it names for itself: making the knowledge already sitting in Atlassian's tools, and the tools connected to them, easier to find and act on without leaving that ecosystem.
What Sanaf is instead
Sanaf is not a layer added on top of a wiki. It is one AI employee, hired into a Slack or Microsoft Teams workspace or into its own chat in the console, holding a written job description across thirty-five roles on the roster today, or a role a business defines itself, and working from that description as requests arrive over days and weeks. Ask it to draft the weekly update this morning and chase an overdue invoice tomorrow, and both requests land against standing jobs the whole team can see get done, not a search result surfaced once and forgotten.
That is the real shape of this comparison. Rovo starts from the Teamwork Graph, a company's own content and work items, and reaches outward through connectors to search and summarize what is already there. Sanaf starts from a job description and reaches outward through app connections, browser sessions and integrations to get new work done: drafting a reply, updating a record, filing a report, booking something, chasing something. One is built to help a team find and understand its own work faster. The other is built to be handed the work itself.
Side by side
| Dimension | Sanaf | Atlassian Rovo |
|---|---|---|
| What it is | One AI employee with a standing job description | An AI layer built into Jira, Confluence and the Atlassian suite |
| Where it lives | Slack, Microsoft Teams, or its own console chat | Inside Atlassian products, with a Slack app for search |
| Unit of work | A job: a standing description of something it owns | A search query, a chat answer, or an in-workflow agent step |
| Reach | Drives a browser or an app connection as part of a job | Indexes 50+ connectors through the Teamwork Graph |
| Getting started | 35 written jobs on the roster, or a 36th you define | Comes bundled with an eligible Jira, Confluence or collection plan |
| Approvals | Per kind of action: do it, ask, or never, with conditions | Admin-configured connector and content access controls |
| Trial | Shadow mode: drafts for review, promoted a level at a time | Available on Standard, Premium or Enterprise trial plans |
| Pricing model | Free to start with $120 of credit and no card, then credits from $20 that never expire | No standalone price; credits bundled into the Jira/Confluence plan, extra usage at roughly $0.01 per credit |
| Best fit | A business function a team shares and can see get done | A team already living in Jira and Confluence that wants faster answers from its own content |
The pricing comparison, done honestly
Rovo's pricing is easy to underestimate because there is no separate invoice for it: the credits arrive bundled with a Jira or Confluence seat a company was likely already paying for, scaled to the plan tier. That is a genuinely good deal for a team already committed to Atlassian's ecosystem at Premium or Enterprise, since the AI layer rides along with a purchase decision the company already made for reasons that have nothing to do with AI. The tradeoff is that the credit pool, the 10-credits-per-interaction cost, and the roughly one-cent-per-credit overage rate are all denominated in a currency a business only sees clearly if it is already deep in Atlassian's admin console tracking seat counts and plan tiers.
Sanaf's pricing has nothing to do with which tools a company already pays for. It starts with $120 of free credit and no card, then moves to packs from $20 that never expire, spent by whoever in the workspace asks for the work regardless of which plan tier they sit on. A team that has never touched Jira or Confluence can start today; a team that has could add Sanaf without touching its Atlassian contract at all. The two are not really priced against each other so much as priced on two different premises: Rovo's premise is "you already pay Atlassian, here is AI on top," and Sanaf's premise is "here is a hire, priced on the work it does."
Where the real difference sits: an index of what exists, or a hire that does something new
Rovo's strength is depth inside a specific kind of company: one that runs its engineering and knowledge work through Jira and Confluence, with enough other tools connected through the Teamwork Graph that a single search bar answering "what's the status of this" or "where did we decide this" saves real time. For that company, Rovo Search and Rovo Chat turn a company's own history, scattered across dozens of connected apps, into one answerable index, and Rovo Agents can carry a Jira ticket or a Confluence page's next step forward without a person opening the tool.
Sanaf's strength is that it does not require a company to already run its work through any particular tool. A job like invoicing, lead follow-up, or a support inbox is not a search against existing content, it is new work landing continuously, and Sanaf sits in the Slack or Teams channel where that work already gets talked about, doing it rather than indexing it. The cost of that difference is that Sanaf is not trying to be the smartest way to search a company's existing Confluence space; it is a coworker for the tasks that were never written down in the first place.
Where Rovo wins, honestly
Choose Rovo if the company already runs on Jira and Confluence at a plan tier where Rovo ships included, and the actual pain is that answers to "what's the status" or "where's the decision" are scattered across tickets, pages and a dozen connected tools nobody has time to search by hand. Choose it if a Rovo Agent automating the next step in an existing Jira or Confluence workflow is genuinely the highest-leverage AI investment available, since nothing on Sanaf's roster is built to live inside a sprint board the way Rovo Agents do.
Where Sanaf wins, honestly
Choose Sanaf if the work in question is not sitting in Jira or Confluence at all: an invoice, a lead, a support ticket, a weekly report, a booking. Choose it if the business is not deep enough into the Atlassian ecosystem for a bundled credit pool to be the natural way to buy AI, or if the team lives in Slack or Microsoft Teams and wants the hire where the conversation already happens rather than folded into a wiki plan. Choose it if the job is standing and recurring, something a person would describe as "part of my role" rather than a one-off question to search for an answer to.
Can you use both?
There is real room for both, and less overlap than "AI for work" suggests at first glance. A company running Jira and Confluence at Premium can keep Rovo answering "what's the state of this project" from inside that ecosystem while Sanaf runs collections, lead follow-up and support from the same team's Slack or Teams workspace. The two answer different questions well enough that stacking them does not double-pay for the same job; the confusion only starts if a team expects Rovo's connectors to do work Sanaf was built for, or expects Sanaf to become the search index for a decade of Confluence pages.
The bottom line
Rovo answers "what does my company already know, and how do I find it fast," from inside the tools that already hold that knowledge, priced as a credit pool bundled into a Jira or Confluence plan. Sanaf answers a different question: "who owns this job for the business," hired into the channel the team already talks in, with a job description, per-action approval rules, and a credit balance the whole team draws from rather than a pool tied to a seat count. Neither is wrong for what it is built to do. The honest way to choose is to ask whether the problem is finding what already exists, or getting something new done, because that answer decides which of these two is actually the fit.
Explore every job Sanaf can hold, see what starting free actually costs, or get in touch to talk through where the line falls in your own business.