Atlassian demos ChatGPT tie-ins for Confluence, Jira Cloud
Atlassian teased early previews of cloud platform AI that supports natural language queries and automated links between Jira, Confluence and third-party products.
Atlassian unveiled AI-driven features on the way for its cloud platform that enhance automated workflows among Jira Cloud, Confluence Cloud and third-party tools.
Atlassian Intelligence, shown by Atlassian officials in a set of demos at the Atlassian Team '23 conference this week, combines intellectual property from Atlassian's acquisition of Percept.AI last year and OpenAI's GPT-4 large language model (LLM). These features are built on Atlassian's data lake, which also underpins the Atlassian Analytics service made generally available this week. Other cloud platform updates this week included expanded data residency support and a bring-your-own-key (BYOK) security program.
Both Atlassian Analytics and Atlassian Intelligence draw on work Atlassian has done over the last three years to integrate services on its cloud platform, adding automation links between Jira Cloud, Confluence Cloud and third-party services such as Slack, said Sherif Mansour, head of product for Atlassian Intelligence at Atlassian.
"What is unique from our lens is the data that we have, based on the specific jobs that customers do on our platform -- project and service work that's unique to Atlassian," he said. "Most of the tools that we have revolve around those two types of teamwork, and together we effectively have a teamwork graph [that tracks] how teams find these services and how they run projects together."
Atlassian Analytics, released in beta a year ago, is now available to all cloud customers. It includes a set of pre-built dashboards as well as a query engine that searches and reports on data across all products. Atlassian Analytics uses the Jira Query Language (JQL) as well as low-code/no-code visual querying and chart templates.
Next, natural language query support for Analytics, as well as individual Atlassian cloud tools, will be added to the cloud platform through Atlassian Intelligence, according to Mansour.
"If you can learn [JQL] syntax, you can ask it anything, and millions of users use JQL on a weekly basis," he said. "But with AI, we can democratize this power to even more users. ... Now, you can just ask it a simple sentence."
From Analytics to Intelligence
This is where ChatGPT, licensed through a partnership with OpenAI, comes in. Atlassian uses ChatGPT to help extract intent from natural language prompts, which the LLM can then translate into JQL and other coded inputs to retrieve data, Mansour said. Atlassian Intelligence features are being developed for all Jira Cloud services, including Jira Work Management, which already caters to nontechnical users.
Procurement teams that use Jira Work Management are a good example of how Atlassian Intelligence can improve queries in Jira Cloud, according to Mansour.
"Depending on how you model the workflow, you could say, 'Show me all contracts that are pending legal review that are over $1 million in value that are due next week,'" he said.
Atlassian Intelligence will come with the first chatbot assistant for Jira Service Management, based on Percept.AI. Company officials this week demonstrated the virtual assistant, Charlie, answering user questions via Slack.
Another AI feature for Jira Service Management will open a help desk ticket when a user indicates a problem that the virtual assistant can't solve, automatically attaching a summary of the issue and suggesting potential responses from a human agent. ChatGPT features will enable that person to automatically change the tone of their response to be more empathetic, writing it automatically based on the context of the ticket.
Will McKeon-WhiteAnalyst, Forrester Research
For Confluence Cloud, Atlassian Intelligence was demonstrated summarizing meeting notes in a Confluence document and adding context to potentially unfamiliar terms. For Jira Software Cloud, demos showed Atlassian Intelligence adding a software test matrix automatically to an issue. Both demos included pulling in data and suggestions from third parties, such as Figma, as well as from Atlassian's Atlas and Compass tools.
All these features, once delivered, would bolster Atlassian's claim as a competitor in the enterprise service management (ESM) market, where it has previously challenged incumbents such as ServiceNow with its ITSM tools, said Will McKeon-White, an analyst at Forrester Research.
"ServiceNow is likely to be moving with their own generative systems, but these enhancements firmly disrupt the entire ESM space," McKeon-White said.
As with any disruption, however, generative AI will also require these vendors to rethink some aspects of existing products, he said.
"Suddenly, we're starting to see things like workflow [management] becoming semi-arbitrary," McKeon-White said. "We're anticipating the ESM space itself expanding and beginning to compete with other markets even outside the traditional ITSM and ESM players."
Security, reliability concerns linger for Atlassian and AI
McKeon-White also sounded a note of caution given the emerging nature of ChatGPT and generative AI, which don't always generate accurate or reliable results.
"[Atlassian] has an excellent setup for a generative system, but there are still opportunities for things to go wrong," he said. "Everyone should always be on the lookout for hallucinations and validating the outputs. Today, we do not recommend nonintermediated usage of any generative systems."
It remains to be proven that Atlassian Intelligence will generate reliable results -- an early access program waiting list has opened, but so far, users aren't able to test the tools themselves. Meanwhile, the specter of a difficult year for Atlassian in 2022 -- including a high-profile cloud outage and ongoing security concerns -- could hamper users' confidence in embracing the new features, McKeon-White said.
To assuage worries about AI, Atlassian Intelligence will be opt-in only, Mansour said, initially available only for Atlassian Cloud Enterprise edition users. Admins will be able to view and control all use of AI within their cloud platform services through a central management console. Atlassian's agreement with OpenAI includes opting out all user data from being used to train the GPT model, he said.
This week, Atlassian also began to offer users control over the encryption of their data on the Atlassian cloud platform with a BYOK feature. That update, as well as a threat detection service released in beta this week, could quell fears about ongoing security vulnerabilities in Atlassian products, McKeon-White said.
"It provides an insurance policy for customers," he said. "When you can do something like BYOK, you can ensure that even if there is a provider breach, the data remains encrypted, and if a provider's keys are cracked, your data will still remain safe."
Beth Pariseau, senior news writer at TechTarget Editorial, is an award-winning veteran of IT journalism. She can be reached at [email protected] or on Twitter @PariseauTT.