Super Work: Pioneering Collaboration for the AGI Era
Fusing Technology and Methodology to Reshape the Future of Work
July 12, 2024
Navigating the Intersection of Apps and AI: Interface & Methodology Perspectives
Earlier this year, with the launch of OpenAI's GPT Store, ChatGPT integrated third-party apps as backend plugins, placing AI at the forefront of the interface. This move was a completely different approach from the idea that AI would be embedded in existing software, becoming the backbone of all technology and products. Now that we've entered the era of Agents beyond AI chatbots, we present two perspectives to speculate on how the "interaction between AI and apps" will evolve to change the way we work.
There is a significant difference between the UI/UX of apps and AI. Apps typically have standardized interfaces designed to meet market demands. This means that whenever customers adopt a new application, they need to learn a new UI/UX. Although the work processes remain the same, users must adapt to the app's interface, often leading to challenges in cultural shifts and change management. This difficulty is especially pronounced in non-tech-savvy sectors where the desire to adopt new technology is high, but the ability to do so is low.
Conversely, AI systems tend to offer personalized interfaces, making them relatively easy for anyone to use regardless of an organization’s technical readiness. However, the productivity of AI outcomes varies greatly depending on an individual's AI literacy. For instance, unless a user is a developer skilled in prompt engineering, AI might seem like an unintelligent personal chatbot that provides irrelevant answers.
In summary, while apps present a high initial barrier but can elevate everyone to a certain level of productivity, AI systems have a lower initial barrier in terms of individual productivity but require continuous learning for effective use by teams and organizations.
Unlike AI, apps are often tightly linked to the specific market categories and methodologies they are designed for. For example, within the goal management software category, some apps are specialized for KPIs (Key Performance Indicators), while others are tailored for OKRs (Objectives and Key Results). Similarly, in the project management category, certain apps are built for Agile methodologies, while others are designed for Design Thinking or WBS (Work Breakdown Structure). Each methodology aims to improve processes and outcomes, yet they have distinct focuses, principles, frameworks, and applications.
This dependency on specific methodologies significantly increases the complexity of digital transformation. The more apps an organization uses, the more users must become proficient not only with different UI/UX but also with managing conflicts, interference, redundancy, and confusion arising from disparate sets of work principles. As any framework does not function outside their designated frame, friction among employees can increase, reducing their autonomy. This creates a vicious cycle that hampers tooling and digital transformation efforts.
Optimizing Collaboration: Integrating Apps and AI for Enhanced Productivity
If the advantages of apps and AI are combined, the dichotomy of “Digital vs. AI” can shift to a cooperative relationship of “Digital & AI”. How should we mix these advantages?
To optimize personalization, AI should lead the user interface, orchestrating apps automatically in the background. This approach enhances productivity by eliminating the need for users to learn complex app UIs and UXs. Instead of navigating through intricate menus, users can simply ask an AI agent to perform tasks, such as tracking projects or generating reports. The functions of various apps can still be accessed through natural language via AI agent function calls, replacing traditional rule-based methods. This seamless integration is the fastest way to locate what you're looking for, removing the need for multiple tab switching.
However, relying on third-party apps, like OpenAI’s GPTs, poses challenges in workplace productivity environments, although it works for simple consumer app functions. Each app developer would need to create custom GPTs tailored to each company's workflow, which is impractical. Moreover, sufficient context input is necessary for correct plugin orchestration from the GPT Store, often making it easier to use a separate browser. Customers would also need to create a supervisor to route agents from various apps, a task manageable only by AI companies. Even if all these hurdles are overcome, the latency from API endpoint requests to third parties can extend runtime to tens or hundreds of seconds, though this might improve with more efficient models and GPUs in the future. Additionally, context switching is not smooth, complicating the achievement of desired outcomes in one go. Therefore, AI should be embedded within the app-based architecture.
The flowchart below illustrates the complexity of integrating custom GPTs, such as goals, projects, and channels, from the GPT Store into ChatGPT as third-party apps. It highlights the challenges users face in understanding the structure of each third-party app and the subsequent development hurdles they must overcome to implement them. Additionally, it demonstrates the limitations in usability that persist even after implementation.
The rapidly evolving landscape of AI necessitates a new methodology capable of accommodating its pervasive influence across all apps. Presently, apps are constructed upon methodologies tailored to address specific challenges, with their user interfaces reflecting these underlying principles and best practices. However, with the UI shift towards an AI-first approach, there arises a demand for a category-neutral AI methodology akin to the agile framework for the AI era. This new methodology would standardize the integration of AI within and beyond various apps, fostering seamless interaction and enhanced efficiency across diverse tools and platforms.
Initially, we considered adapting Agile methodologies for this purpose. However, Agile, originally conceived for software development, presents challenges when applied to non-technical teams. Transforming its core values like “Scrum (Backlog, Sprints)”, “Extreme Programming (Test-Driven Development, Continuous Integration), or “Dynamic Systems Development Method (DSDM)” to non-technical teams proved difficult and unnecessarily overwhelming.
Additionally, many Agile frameworks are visually interface-centric, such as Kanban boards, Burn-down Charts, Release Planning Boards, or Value Stream Mapping, which contradicts the AI-first approach. Recognizing these limitations, we've devised a novel work methodology tailored for cross-functional teams in the age of AI, introducing what we call "Super Work."
Super Work Methodologies: Blueprinting Productivity in the AGI Era
Super Work refers to a new set of principles and methodologies aimed at enhancing workplace productivity through Human-AI collaboration. As we approach the era of AGI, the term “Super” has been adopted to denote the transition from AI agents as personal assistants to superintelligent AI with superhuman capabilities. Super Work envisions AI-first frameworks essential for a future where AI extends beyond app boundaries to assist with human tasks, perform tasks humans used to do, and collaborate among AIs, aiming for Human-Human, Human-AI, and AI-AI collaboration. The core components of Super Work methodologies are outlined as follows:
- Collaborative AI Workspaces: SynergiSphere
- Description: Integrated digital environments where humans and AI systems can work together on tasks, share information, and track progress.
- Purpose: To provide a unified platform for collaboration, enhancing productivity and communication.
- Real-Time Collaboration Tools: FlashTeam
- Description: Platforms and tools that enable synchronous collaboration between humans and AI, including shared goals, real-time communication, and collaborative projects.
- Purpose: To facilitate immediate and dynamic interaction, enhancing the speed and quality of collaborative efforts.
- Hybrid Task Management: TaskFusion
- Description: Systems that coordinate and assign tasks between humans and AI agents based on their respective strengths and capabilities.
- Purpose: To optimize task allocation, ensuring that both humans and AI agents contribute effectively to the collaboration.
- Natural Language Interfaces: ChatFlow
- Description: User interfaces that allow humans to interact with AI systems using natural language, both spoken and written.
- Purpose: To make AI systems accessible and user-friendly, reducing the learning curve and enabling more intuitive interactions.
- Role-Based Access and Permissions: SecureRole
- Description: Systems that define and manage access controls based on user roles, ensuring that both humans and AI agents have appropriate permissions.
- Purpose: To maintain security and privacy, ensuring that sensitive data is only accessible to authorized entities.
- Adaptive Learning and Feedback Loops: IntelliLoop
- Description: Mechanisms for continuous learning where AI systems adapt based on user feedback and evolving requirements.
- Purpose: To ensure that AI systems remain relevant and effective, improving over time based on real-world usage.
- Multi-Modal Interaction Capabilities: OmniInteract
- Description: Support for various forms of interaction, including voice, text, and visual inputs.
- Purpose: To provide flexibility in how humans interact with AI, catering to different preferences and contexts.
- Contextual Awareness: ContextIQ
- Description: AI systems that understand and consider the context in which they are used, including user intent, environmental factors, and task-specific details.
- Purpose: To enhance the relevance and accuracy of AI outputs, making interactions more meaningful and effective.
- AI Collaboration Analytics: CollaboMetrics
- Description: Tools that analyze human-AI collaboration patterns, measure effectiveness, and identify areas for improvement.
- Purpose: To provide insights into the collaboration process, helping organizations optimize their human-AI interactions.
By leveraging these frameworks, Super Work methodologies aim to facilitate effective Human-AI collaboration and prepare organizations for the AGI era. Emphasizing transparency, adaptability, and user-friendliness, they foster a productive and harmonious environment where humans and AI systems work together towards common goals. Given that AI is an evolving technology, the Super Work frameworks mentioned above should be continuously developed to ensure optimal performance.
How Swit is Implementing Super Work
The Swit team is transitioning essential elements of work productivity, such as goals, projects, and channels, from a heavy, hard-to-use, multi-app ecosystem to an intuitive AI-driven interface. This shift aims to implement the Super Work methodologies, transforming Swit into a Super Workspace with the following components:
- Agent Studio: A space for creating and managing custom AI agents based on roles, including their unique knowledge, skills, and tools.
- Multi-agent Architecture: A hierarchical structure where agents collaborate as a team, with a supervising agent orchestrating and routing agents as needed.
- Permission-aware Agentic RAG: Each agent must have distinct functions and roles, including data and human access configurations. For example, a CFO agent must have real-time access to data that the CFO can access and must not provide this information to unauthorized individuals.
- Agent Tools for Collaboration Essentials: The essential components for workplace collaboration, such as Goals, Projects, and Channels, should be implemented as powerful apps capable of addressing business needs, with each function connected through agent function calls.
As the product advanced, the organization had to adapt as well. Transitioning from a software company that created tools solely for people to an AI company focused on collaboration between AI agents and humans required a shift from interface-driven design and rule-based development methodologies to an AI-defined approach. This transition posed a significant internal cultural challenge. However, our team is now working with more passion and confidence than ever, believing that we are on the right trajectory to prepare for the future of work, where traditional boundaries between teams, apps, and even methodologies no longer exist.
Josh Lee, Co-Founder and CEO of Swit
