Atlassian is rolling out new generative AI capabilities that will be embedded in the company’s entire portfolio of workforce management cloud products and are designed to help both service-based and project-based work teams be more efficient.
The technology, dubbed Atlassian Intelligence, was announced Wednesday and is built on in-house AI models gained through the company’s acquisition of Perceptive.ai in 2022, and a collaboration with Microsoft-back OpenAI — the creator of ChatGPT, whose launch last year sparked a virtual AI arms race among top tech companies including Google and Meta.
Through the use of large language models, the foundation of generative AI, Atlassian Intelligence constructs custom teamwork graphs showing the types of work being done and the relationship among them, while the open approach of the Atlassian platform provides additional context and data from the third-party apps teams use, according to the company.
“This is really about accelerating teamwork across the different types of products that we have,” said Sherif Mansour, product manager at Atlassian.
“Atlassian intelligence is a set of capabilities that exists within all our cloud products to address different use cases and scenarios, whilst leveraging the power of artificial intelligence and large language models,” he said.
Using generative AI to support teams
Atlassian has outlined multiple ways for Atlassian Intelligence to help support teams, including accelerating work, getting instant help, and building a shared understanding of projects. Atlassian Intelligence can also translate natural language queries into Jira Query Language to find issues across all Jira Cloud products.
To help accelerate work, Atlassian Intelligence uses generative AI technology from OpenAI to create, summarize and extract information from content. For example, decisions and action items can be quickly summarized from meeting minutes, and tweets can be drafted based on information held in Confluence.
While ChatGPT’s technology will be used to help Atlassian Intelligence process natural language requests, Sherif stresses that the company has an agreement that no customer data will be stored by the software or OpenAI.
The technology can also provide employees and customers with immediate responses and fast resolutions via the new Jira Service Management virtual agent, which can automate support interactions from within Slack and Microsoft Teams. Atlassian Intelligence also takes on repetitive requests on behalf of support teams, allowing them to focus on more important work.
Virtual agents to resolve work requests
The virtual agent is designed to resolve help requests instantly based on its understanding of knowledge-base articles and can also ask follow-up questions to take any necessary actions. It can also summarize activity on help requests to get assigned agents up to speed, and help craft responses to customers which can also be edited for tone.
In addition, Atlassian Intelligence can surface previously resolved or related issues and recommend relevant articles and related pages to resolve requests and incidents faster.
Atlassian Intelligence also helps to build a shared internal library of knowledge and terms used to describe projects, acronyms, and internal systems. It can provide shared context with an on-demand dictionary specific to your company, your teams, and their work. You can highlight a term and ask Atlassian Intelligence to explain it with the definition, source of information, internal subject matter experts, and how it connects to related work based on the teamwork graph, Atlassian said.
Natural language instead of SQL queries
Atlassian Intelligence’s understanding of natural language questions also means that users can ask questions and have the technology generate insights using data from multiple sources in Atlassian Analytics — a hub to analyze and visualize data across Atlassian products and connected third-party tools — without needing to know how to write SQL.
Since launching to the public, large language models like ChatGPT have faced criticism for providing incorrect or sometimes even defamatory information in response to requests from users, an issue Mansour is sympathetic with but not one that he thinks will be a major issue for businesses looking to leverage the technology.
“In the enterprise, it’s less of a problem because there’s far less fake news,” he said. Referencing the internal library offering from Atlassian Intelligence, Mansour acknowledged that while it’s technically possible for someone to attribute false definitions to terms, it’s a highly unlikely scenario as there’s nothing to be gained from doing so.
Despite this, a key principle for with Atlassian Intelligence is that the user remains in control. Mansour explained that while the technology might help an agent craft a response, that message will only be delivered if the user manually presses the send button.
Furthermore, Mansour said that although Atlassian Intelligence largely adheres to the company’s existing privacy policy, Atlassian users will also need to opt into the technology.
“The user always has to be in control,” he said. “The technology should never do anything without the user first confirming it.”
Atlassian customers can sign up to Atlassian Intelligence’s early access waiting list from April 19, although it cannot guarantee specific timing for when users will gain access.
