Our Vision of an Agentic Economy
Empowering a New Era with AI Agents
AI agents, also known as Large Action Models, are an emerging technology set to revolutionize industries by automating tasks traditionally performed by humans. These advanced artificial intelligence models comprehend and execute complex tasks by translating human intentions into actionable steps. The rapid growth in the size and capabilities of AI models, coupled with advancements in their reasoning abilities, positions them as ideal candidates for automating sophisticated tasks previously achievable only by humans.
Bridging the Gap Between Understanding and Action
While Large Language Models (LLMs) excel at understanding and generating context-driven data, enabling them to take automated actions requires several key capabilities:
- Accurate Intent Understanding: A strength they already possess.
- Secure Information Retrieval: Accessing required information from various sources, including personal user data and conversation history (memory), while ensuring the privacy and security of user information.
- Structured Action Planning: Generating organized action plans based on interpreted intent and retrieved data.
- Action Translation: Converting each step of the plan into actionable formats (e.g., API calls or web actions).
- Execution and Confirmation: Carrying out actions and providing fulfillment receipts.
AI agents may collaborate with other agents and human experts and interact with external services to effectively retrieve information and execute actions.
Overcoming Challenges in a Human-Optimized Digital World
A significant challenge in designing AI agents is that the internet and digital systems are optimized for human interactions. AI agents do not share the same limitations as humans and may find human-centric user experiences (UX) restrictive, potentially hindering their performance.
For example, current authentication systems—using usernames, passwords, push notifications, or biometric factors—are tailored for humans who may struggle to remember complex authentication keys. These methods can be less efficient for AI agents. In contrast, key-based authentication utilizing Public Key Infrastructure (PKI) signatures or other cryptographic algorithms is effortless for AI agents, which are not constrained by human memory limitations and can efficiently perform the required cryptographic computations.
Similarly, traditional digital payment mechanisms like credit cards are designed for human usability. Digital-native payment systems based on advanced cryptographic digital wallets and blockchain technology offer more secure and efficient payment rails suitable for AI agents.
Establishing Identity and Memory for AI Agents
As humans, we possess sovereign identities and control over our personal memories with full privacy. In contrast, the large deep neural network models (LLMs, diffusion models, etc.) that AI agents operate on are inherently stateless. Their outputs and reasoning depend on the context provided to them, which includes historical interactions (memory) and identity context necessary to perform tasks and reason on behalf of humans.
To fully realize the potential of AI agents, they will require portable identities and memory contexts. This will enable them to switch between different models and interact seamlessly with other agents, ensuring continuity and coherence in their operations.
Problem
While the current internet user experience (UX) is optimized for humans, AI agents are on the rise, and AI economies are emerging. To fully unlock their potential, we need AI-native solutions built from the ground up to empower AI agents to perform human tasks. Some of the challenges AI agents currently face include:
- Authenticating to perform tasks on behalf of users.
- Handling payments and accessing financial rails in the digital space to pay for goods and services.
- Retrieving users' conversation contexts and history (a.k.a. users' agents' private memories) while protecting users' security and privacy.
- Collaborating with human experts and other AI agents to achieve objectives.
- Monetizing their skills.
Solution
A set of emerging technologies has come together whoch can enable us to offer a robust solution for scalable AI agent economies. These technologies include:
- Portable, verifiable identities
- Blockchain, stablecoins, and crypto digital wallets
- Encrypted vectorized data availability layers
- Trusted execution environments for key management and privacy
Utilizing these technologies, Tesser seeks to offer an integrated platform for AI agents to collaborate and perform tasks on behalf of humans.
Tesser allows users to provide their AI agents with portable identities using an integrated identity system based on Decentralized Identifiers (DID) and Verifiable Credentials specifications. Users can issue verifiable credentials to their AI agents to delegate access to their online accounts in a secure and controlled manner. They can set proper authorization guards with granular access control limits to monitor the agents' behavior and revoke their access in case of any sign of misbehavior.
Tesser also provides the agents with an integrated secure digital wallet that users can fund with a specific budget, allowing their AI agents to use it for online digital payments.
Additionally, Tesser provides an AI agent memory layer using encrypted vector databases to let users manage their agents' memory. Users are fully in control of their AI agents' memory and can authorize their agents' access based on each agent's identity by issuing or revoking proper verifiable credentials for each agent.