2024 is the year when AI agents are the talk of the town. AI agents have the ability to execute a multitude of tasks for you. The last couple of weeks were intense in this space as major tech companies have demoed their upcoming agentic products.
Microsoft gave a demo on their Copilot Studio, Google showcased Jarvis, and Salesforce launched their own low-code agent building tool called Agentforce. Major AI architectures based on agentic workflows are emerging.
Many people can build an agentic framework or multi-agent AI workflow highly specific to their use case today. OpenAI’s Swarm library is open source on GitHub for anyone to experiment with.
Artifact AI's vision is to become the global standard in accounting software. They've developed a customizable, agent-driven workflow that covers everything from bookkeeping to tax filing and everything in between.
Multi-agentic AI systems are essentially LLM-based agents that operate almost like an LLM in a loop, but with memory and states. This system is what we call a multi-agent AI system.
Ariel Harmoko, the CEO of Artifact AI has spoken with 300-400 founders. He believes that most founders are not building their companies bottom-up with AI agents because this space is still new and requires deep technical knowledge of LLMs to stay current with the latest advancements.
For those looking to build AI-native companies, it is recommended to build a tailored evaluation framework for different LLMs based on their use case. You don’t need to be a technical founder to start building such workflows.
The next big leap in AI may be on the hardware front. Already, we are seeing this in the push towards AR/VR and the Tesla Bot gaining traction. The fusion of multimodal AI with hardware, enabling interaction beyond chat to include vision and more, is going to be thrilling to watch.