Asana’s Work Graph: The Key to Unlocking the Potential of AI Teammates
The rapid advancement of large language models (LLMs) has given rise to a new era of artificial intelligence. These powerful models have captured the public's imagination through chatbots and AI assistants, which have become the most recognizable form for LLMs. However, at Asana, we believe that the true potential of LLMs lies not in standalone chatbots, but in the creation of AI teammates – intelligent agents that work alongside humans to advise on work, take action, and adapt to each organization.
From Chatbots to AI Teammates: Shifting the Paradigm
While chatbots and AI assistants have showcased the impressive capabilities of these models, they often operate as isolated systems, disconnected from the context and workflows of organizations. In contrast, AI teammates in Asana are designed to be deeply integrated into the fabric of work, leveraging LLMs in the context of teams and their workflows, and giving them access to the necessary tools to become valuable collaborators in the workplace.
Models + Tool Use + Scaffolding = The Foundation of Effective AI Teammates
LLMs enable sophisticated natural language understanding, fluent language generation, and reasoning capabilities that allow AI teammates to communicate effectively and tackle complex problems. We’re independent of any particular model provider, so can offer customers flexibility in which underlying model they want to leverage, starting with those from our key partners OpenAI and Anthropic.
But to be truly effective, AI teammates need to go beyond just understanding and reasoning; they need to be able to take action and interact with the work around them. This means giving them the ability to use tools, like searching for information in another system using APIs, or creating saved items like tasks and projects in Asana. By integrating with various systems and services, Asana AI teammates can access real-time data, analyze information, assist with process management, and take on tasks like discovery and drafting, making them valuable collaborators in the workplace.
With AI teammates that are built off powerful models and equipped with a versatile toolbox, the last step is defining a set of instructions that guides their process and dictates how they interact with people to achieve outcomes. This is the scaffolding part of the system – the guidance in the form of prompts, sequenced steps, and specialized UIs to handle parts of the process. That could mean predefined flows like requesting clarification from users in smart chat or filling out a column of custom fields in your project using smart fields. While chatbots dominate the imagination when people think of AI these days, many are also familiar with more integrated kinds of AI, like Github Copilot or even AI-powered reviews on Amazon.
By constraining intent by the context of what customers are trying to achieve and clearly specifying the format of the outputs, scaffolds can generate consistent, accurate, and useful results. Importantly, this can happen without everyone becoming experts at prompt engineering and spending effort for each engagement to set up the correct inputs and execute each step. For Asana, this means managing exactly how models interact with the Work Graph in the context of specific, relevant workflows. Asana AI teammates include both ones we design and ones built by our customers.
From Human Teamwork to Human-AI Collaboration
At Asana, we have been investing in the Work Graph since our founding, with the primary goal of enabling seamless collaboration and coordination among human teams. The Work Graph is a comprehensive map of an organization's work that connects work and workflows to higher level goals, and understands the relationship between them. It provides a clear structure and context for work, allowing teams to stay aligned and achieve their objectives more effectively.
Initially designed to make work effortless for human teams, and to give clarity and drive accountability on who’s doing what by when, how, and why, the Work Graph turns out to be the perfect scaffolding for AI teammates. By providing AI teammates with access to the rich, structured data of the Work Graph, we enable them to understand the context of work, give customers more consistent results, and make meaningful contributions to the team's success.
Because of the relationships in the Work Graph, like which portfolios are associated with what OKRs, the dependencies between tasks, and how all the people are involved, we know which context to look at, and we don’t try to look at ALL possible data which is how you easily end up with errors. Language models confabulate when they try to give you an answer based on what is in the training data, but they’re vastly more accurate when asked to give you an answer based on what is in the context window. Our strategic advantage is being able to identify the most important context well because it’s explicitly identified by the relationships between the tasks, projects, portfolios, goals, and people in the Work Graph.
It also facilitates human-AI teamwork. The Work Graph acts as a shared language between humans and AI teammates, enabling seamless communication and collaboration. As team members use Asana to manage their work, they provide valuable feedback and input that helps refine the AI teammate's understanding and performance.
Asana AI Teammates in Action
Let’s illustrate the power of the Work Graph as “scaffolding” in a common use case for our customers: product launches.
Product launches involve complex coordination across multiple teams, including product, marketing, and sales. In Asana, the product launch can be managed as a portfolio with each team’s roles and responsibilities clearly laid out in projects and tasks that are assigned to each team.
The Work Graph serves as a central hub for all product launch-related information, including requirements, timelines, dependencies, and stakeholder responsibilities. Asana AI teammates can leverage this structured data to:
Get the launch planning process started faster by identifying the right roles and responsibilities, helping craft a brief, and scheduling team kick-offs
Identify potential risks and bottlenecks in the launch timeline and suggest mitigation strategies
Assist in creating and maintaining a comprehensive launch checklist, ensuring that no critical tasks are overlooked
Provide real-time updates to stakeholders on the progress of the launch, identify common themes, and keep everyone aligned and informed
Analyze customer feedback and market trends to provide data-driven insights and recommendations for product improvements
Take action on specific tasks like generating drafts of product requirement docs (PRDs), crafting messaging and positioning ideas, preparing sales enablement materials, and writing marketing copy for a company website or blog
By utilizing the Work Graph as a scaffolding, Asana AI teammates can orchestrate the complex product launch process, ensuring that teams stay coordinated, deadlines are met, and the product is successfully brought to market.
The Future of Work with Asana AI Teammates and the Work Graph
At Asana, we have always been committed to helping teams work more effectively together. With the rise of AI teammates, we see a tremendous opportunity to extend this mission and revolutionize the way we work. By leveraging the Work Graph as a scaffolding for AI teammates, these intelligent collaborators become deeply integrated into the fabric of our organizations, working seamlessly alongside their human counterparts.
As we continue to explore the potential of AI teammates and the Work Graph, we are excited to be at the forefront of this transformative journey. The future of work is not about replacing humans with machines, but rather about harnessing the power of human-AI collaboration to achieve new levels of productivity, creativity, and success. With Asana and the Work Graph, we are paving the way for a new era of teamwork – one where humans and AI work hand in hand to shape the future of work.
If you’d like to join our Asana AI teammates beta, please sign-up for early access here.
