IT Service Management (ITSM) practices are highly beneficial for companies as they aid in streamlining and optimizing IT service delivery, improving customer satisfaction, and enhancing overall business efficiency. By implementing ITSM practices, companies can effectively manage their IT services, align them with business goals, and ensure a seamless experience for both internal and external customers.
This article highlights a notable AI-powered feature within Atlassian's ITSM solution Jira Service Management – virtual agents. We'll aim to maintain a high-level overview and delve into the specifics in a second part of this series.
Jira Service Management for ITSM and ESM
Jira Service Management (JSM) is a powerful ITSM solution that can help with Enterprise Service Management (ESM) as well. By leveraging JSM, companies can extend their ITSM practices to other areas of the organization, beyond just IT. This means that JSM can be used to streamline and optimize service delivery across various departments, such as HR, Facilities, Legal, and more. JSM offers an intuitive, robust platform which can be expanded with functionalities via apps from the Marketplace.
A short introduction to Atlassian Intelligence
Atlassian partnered with OpenAI to incorporate generative artificial intelligence in the Atlassian Cloud platform. Atlassian Intelligence features are available to Premium and Enterprise customers.
Atlassian Intelligence is designed to accelerate any type of work, so almost every product you can think of offers a bunch of features leveraging OpenAI’s groundbreaking large language model ChatGPT. The benefits are obvious: save time, guide through complexity, improve work quality, prepare for the future.
- How do we save time? Use the predefined prompts to improve your writing or translate a text you just wrote into a different language. (Jira, Confluence).
- How do we guide through complexity? Quickly summarize the content of a page or an issue to get the gist about what is going on (Jira, Confluence).
- How do we improve work quality? Use AI to help you improve your code (Bitbucket).
- How do we prepare for the future? Use AI as a teacher. Ask what you missed in your concept, or how you can learn more about the topic at hand (Confluence).
You can find an overview of the currently available features in the Atlassian documentation.
A word on data protection
You may think I don’t want to train future language models with my organisation's data and you are absolutely right. Atlassian's got you covered on that, though: Atlassian Intelligence adds the information from the issue you are working on (Jira) or the page you are writing (Confluence) to the context of the request sent to ChatGPT. It’s like a prompt you don’t have to enrich yourself – you just say what you want to do and the AI will handle it. No data will be used to train the model. More information can be found in the FAQ regarding AI.
Getting started with virtual agents
Every process needs to be reviewed at regular intervals to ensure its accuracy and suitability for the problem it once tried to solve. In a long list of checks (one of which should always be "Does the problem still exist?"), we also have to check whether there are any parts in the process that can be automated. Any automation frees up valuable resources in the workforce – resources your business can use to support other areas.
A big step forward in automating service management requests (ITSM or ESM) is the virtual agent feature in Jira Service Management (Premium or Enterprise Cloud). Virtual agents essentially are machine learning enhanced and AI capable agents that start your conversation with the customer. The goal is to offload repetitive tasks, information retrieval and answering easy questions with articles from the knowledge base.
Transforming Slack into your Service Desk
Slack is an enterprise chat solution that has gained importance during the pandemic. It allows teams to communicate, exchange files, and collaborate in real time. With many workflows shifting to remote work, the need for efficient and flexible communication tools like Slack has become even more apparent.
Virtual agents integrate your Jira Service Management solution into Slack, eliminating an additional step and tool to provide the support your customers need.
Atlassian acquired Halp in 2020 and is integrating features from Halp into JSM. Virtual agents are available for Slack, the integration for Teams is currently being worked on (https://jira.atlassian.com/browse/JSDCLOUD-12279) with an estimated time frame until the end of 2024.
The integration of virtual agents with Slack is simple yet ingenious and extremely easy to set up. Via your service project, they are added to your organization's workspace as a Slack app.
The Slack bot Assist typically operates in three types of channels:
- an agent channel where agents assign, edit and discuss requests
- a test channel where you can test the virtual agent
- a request channel where your customers will raise requests
Introducing intents
The next step is to determine the intents to which your virtual agents should respond. An intent is basically a set of phrases the virtual agent is trained with (by machine learning) so that it can answer a request with a certain conversational flow. There are a bunch of easy to setup and preconfigured intents, but you can also teach you virtual agents new phrases and create new intents.
Conversations with the customer
How does the virtual agent offer help? Once an intent is matched (as in: you type something in Slack, the virtual agent tries to match the request against his machine learned dictionary), it answers with the associated conversational flow. This flow can be configured by adding steps of different types – in an easy to use no-code editor similar to the workflow editor we know from Jira. It's as simple as that. Once again, Atlassian remains true to their ethos by providing a product that is easily configurable for teams and individuals without specialized knowledge.
The workflow can become quite complex, including asking the customer for input and potentially sending web requests to gather additional information. It will always result in a solution or an escalation, creating an issue for a human agent to address.
Integrating Atlassian Intelligence into your service desk experience
There very well might be situations where your carefully crafted intents are not matching the customer needs. In this case and when enabled, the virtual agent can use Atlassian Intelligence to lookup articles from the knowledge base. These articles might be presented to your customers in the form of a link.
At the moment, this solution is not comprehensive and does not fully reflect the potential of generative AI for us. Certainly, as a human agent, you can utilize all the features that Jira has to offer in relation to Atlassian Intelligence. However, the virtual agent is restricted to this particular scenario. We are excited to see what Atlassian has on their roadmap for this feature in combination with AI this year.
Designing and automating conversational flows
I think everyone has had a bad experience with a first level support at one point. The goal of intents and conversational flows is to enhance the process and reduce the burden on human agents, while ensuring that it does not negatively impact our internal or external customers. I recommend endeavouring to create conversation processes. Well phrased and designed, they are the guarantee of success for the acceptance of the virtual help desk and ultimately achieving the goal of automating the service experience.
In theory, you could create an entire process, for example troubleshooting your WiFi issues. Step types might involve providing choices, sending information as a message, requesting information, and even retrieving information via web request. However, overengineering is a concern here, and you wouldn't want to create a flow with numerous steps. Your customers are aware that they are interacting with a bot and will eventually end the conversation.
Keep it simple and lightweight, and leave the in-depth analysis jobs to human agents.
I like decision matrices a lot and maybe this here can help. Whenever you want to create a virtual agent decide whether:
- Is it easy to identify the problem (range goes from “very easy to pin down” to “very hard to pin down”)?
- Is it possible to create an automation here (everything is possible, but some activities require approval processes and human intervention)?
- Green: Low hanging fruit – The intent is specific and the automation is easy to create.
- Yellow: Depends on the benefit – Only move forward if the intent is specific enough. Keep cost vs. benefit in mind with regards to an automation.
- Red: Don't touch – You can’t win this one.
Conclusion and further outlook
In conclusion, the introduction of airtual agents in Jira Service Management represents a significant step forward in automating service management requests. By leveraging machine learning and AI capabilities, virtual agents aim to offload repetitive tasks, retrieve information, and answer common questions, ultimately freeing up valuable resources for businesses. Their integration with communication tools like Slack further enhances the support provided to customers, with the potential for future integrations. The ability to design and automate conversational flows offers a promising outlook for enhancing the service experience. By carefully crafting conversational flows, businesses can reduce the burden on human agents while ensuring a positive customer experience.
Would you like to find out more about Atlassian Intelligence and the use of virtual agents? Make an appointment with our AI specialist Sascha Masanneck or contact us.
