Hello,
Generative AI/ LLM is a revolution that started a few months back with GPT-3 and now making its way to concrete implementation and use cases.
In fact, according to Gartner, generative AI is projected to represent a staggering 10% of all data generated by 2025, a significant increase from its current share of less than 1%.
I am starting this discussion if you are interested in Generative AI for Atlassian. We have been testing Generative AI with JSM, Confluence, and more. I want to share some feedback and ideas.
What are the key capabilities of Generative AI?
Greg Brockman, the president, Chairman, and cofounder of OpenAI, recently said in an interview, “Everything is language in an enterprise.” You can think of the information in various platforms such as Jira Service Management tickets, Confluence knowledge base, SharePoint, website content, and conversations in Microsoft Teams or Slack.
Generative AI models such as GPT-4 are powerful language models that can perform a wide range of natural language processing (NLP) tasks with high accuracy. Here are some specific functions that Generative AI can do:
- Text generation: Generative AI can generate natural-sounding text in various contexts, such as essays, stories, poems, and computer code.
- Language translation: Generative AI can accurately translate text from one language to another.
- Question-answering: Generative AI can answer questions based on contexts, such as a paragraph or a document.
- Summarization: Generative AI can summarize long pieces of text into shorter, more concise summaries.
- Sentiment analysis: Generative AI can analyze the sentiment of a given piece of text, determining whether it is positive, negative, or neutral.
- Chatbot: Generative AI can be used as a chatbot to simulate human-like conversations with users, answering questions and providing assistance.
- Text completion: GPT-3 can predict the next word or phrase in a sentence or paragraph, making it useful for tasks such as autocomplete and autocorrect.
- Text classification: GPT-3 can classify text into different categories based on content, such as news articles, product reviews, or social media posts.
- Language modeling: GPT-3 can be fine-tuned to generate text in a specific style or domain, such as technical writing, academic research, or creative writing.
What are the use cases where Generative AI adds value on top of Atlassian?
Generative AI becomes truly valuable when it creates meaningful outcomes for specific personas, such as employees, contact center agents, and customers, within their unique work contexts (reviewing a ticket in Jira Service Management, for instance) and utilizing relevant enterprise knowledge (Confluence for instance). Let's explore some use cases where Generative AI can deliver significant value.
Agent Assists
Operating in a contact center can be demanding, as employees are tasked with delivering exceptional customer support and service, often in the face of challenging situations or interactions with frustrated customers. Furthermore, they must efficiently manage a high volume of tickets or calls while adhering to the expectations set by management. To excel in customer service, contact center employees need comprehensive training and the ability to navigate diverse scenarios. Equally important, they must demonstrate patience and maintain a courteous demeanor.
In such a working environment, Agent Assist becomes an invaluable tool to lighten the workload and make being a call center agent more enjoyable.
Agent Assist employs Generative AI to offer real-time support to contact center agents during customer interactions, utilizing several methods to achieve this. For instance, Agent Assist can examine incoming tickets in Jira Service Management and generate a draft response by drawing from knowledge articles in Confluence. After the Generative AI completes these tasks, Agent Assist updates the ticket status in Jira Service Management. Contact center agents can then review and modify the content before sending the ticket to their customers, saving up to 70% of their time. This increased efficiency enables the workforce to meet service-level agreements (SLAs) and boost customer satisfaction. Improved first contact resolution (FCR) also encourages customers to return in the future, fostering long-term loyalty.
Chatbots
Chatbots were once touted as the future of customer support. Around 2016-2017, chatbot hype peaked, with numerous businesses adopting them to enhance customer service. Chatbots were expected to provide prompt, efficient, and personalized responses to customers' queries. However, despite the high expectations and positive intentions, chatbots for customer support failed to meet the expectations.
With Generative AI, Chatbots are back, and they offer a range of unique features:
- Large-Scale Pre-Training: Generative AI has been pre-trained on a massive corpus of text data, allowing it to understand and generate responses to various topics and conversation styles.
- Training Data: Creating a fine-tuned model (custom) with your training data is easy.
- Generate responses: Generative AI leverages enterprise content such as Confluence to generate personalized responses.
- Multilingual Capabilities: Generative AI can understand and generate responses from Confluence in multiple languages, making it a more versatile conversational agent.
Now, Intelligent Chatbot allows organizations to sidestep the 24/7 tsunami of internal and external correspondence by proactively responding and assisting customers or employees. In doing so, chatbot deflects approximately 40% of service management interactions. From IT & HR to Finance & Customer Services, Chatbots respond across all departments and becomes an invaluable solution by quickly learning from user interactions via machine learning.
What is required to make this work?
Here, I provide feedback on generative AI models and tools required to securely combine the best of Generative AI with enterprise data, knowledge, and Atlassian apps.
GDPR-compliant Generative AI models.
GDPR, security, and ethics are critical in Enterprise. Selecting the best Generative AI models designed with compliance, privacy, and safety in mind is essential. For instance:
- Dedicated Training Data: Make sure that the data provided for training doesn't contribute to the training or improvement of any other generative models.
- Rigorous Content Filtering: All data submitted undergoes thorough content filtering and processing. No prompts are stored or used to train, retrain, or improve any models.
- Proactive Abuse Prevention: The content is filtered, and retention of prompts and completions are for up to several days only. This vigilance will ensure that your data is used responsibly and ethically.
Prompt engineering
Prompt engineering is a process of designing and optimizing prompts to generate desired outputs from a language model such as GPT-3. It involves crafting specific prompts to guide the language model towards producing high-quality and relevant text based on the user's input or query.
Prompt engineering aims to improve the efficiency and effectiveness of natural language processing tasks by using carefully designed prompts that provide relevant context and constraints to the language model. This can help the model produce more accurate, coherent, and appropriate responses to user requests.
Prompt engineering involves several steps, including selecting the right input and output formats, creating training data, fine-tuning the model, and evaluating the results. It often requires domain expertise and knowledge of the target use case, as well as an understanding of the strengths and weaknesses of the language model being used.
Overall, prompt engineering is a crucial step in building robust and effective natural language processing systems that can support a wide range of applications, from chatbots and virtual assistants to language translation and content generation.
Knowledge retrieval for Confluence
Generative AI has been trained on large amounts of text data, enabling it to generate human-like responses to a wide range of natural language inputs. However, to provide more contextual and accurate answers for your customers or employees, Generative AI needs access to enterprise data and knowledge.
You need a comprehensive knowledge retrieval solution that seamlessly integrates with Confluence and extracts pertinent information. Knowledge retrieval involves using techniques such as keyword matching, semantic analysis, and machine learning algorithms to retrieve relevant information from Confluence. Knowledge retrieval can significantly enhance the performance and accuracy of Generative AI, enabling it to provide more informative and personalized responses to user queries.
Integration with Jira Service Management
For most use cases (agent assist, chatbot, ...), you also need a plug-in in JSM to leverage enterprise data (the text contained in a ticket, for instance) with generative AI. For example, you need this plug-in if you want generative AI to analyze a customer ticket and prepare a response for your contact center agent leveraging Confluence knowledge.
Conversational User Interface
A conversational user interface (CUI) is a computer interface that allows users to interact with a software system or device using natural language, such as spoken or written dialogue, rather than traditional graphical user interfaces (GUIs).
A CUI aims to mimic human-to-human conversation by using techniques from natural language processing (NLP), machine learning, and artificial intelligence (AI) to understand user input and generate appropriate responses. This can be achieved through various means, such as chatbots, voice assistants, and messaging apps.
One of the key advantages of CUIs is that they offer a more intuitive and user-friendly experience than traditional GUIs. Users can interact with the system in a way that feels more natural and conversational, without needing to learn a specific interface or menu system.
CUIs are increasingly being used in a variety of applications, such as customer service, e-commerce, healthcare, and education, to provide personalized and efficient interactions with users.
Overall, conversational user interfaces are an exciting and rapidly evolving technology that has the potential to transform the way we interact with computers and devices.
The Benefits of using Generative AI
Generative AI is a groundbreaking technology. Its integration with Jira Service Management (JSM), Confluence, and more will empower support teams to provide instant assistance, alleviating the strain of repetitive requests.
It is now possible to tailor solutions to specific use cases. This competitive advantage will enable every enterprise to elevate and simplify user experiences, optimize workflows, and deliver superior results to their clients - all while freeing up the support team to focus on more complex issues.
I am looking forward to this discussion.
