Hercules AI unveils ‘assembly line’ to help companies quickly deploy AI agents

Hercules AI, a generative AI company innovating around the future of work, has developed a new methodology for quickly deploying virtual AI workers in the enterprise. In addition, the company formerly known as Zero Systems has introduced RosettaStoneLLM, a model helping companies in regulated industries automate complex workflows that require cognitive decision-making.

Dubbed an “assembly line” process, organizations can choose prefabricated components to develop and deploy virtual AI workers. Everything is premade, tested, and pre-configured in advance—no custom ordering. Hercules AI says doing this will produce AI agents that are “high quality, cost-efficient, and easily scalable.” All that would need to be done is fine-tune the model so the bot follows the specified workflow.

Most companies may prefer a bot that caters to their needs and specifications. However, doing so in a regulated field requires time and money, which can harm customers. However, those in finance, insurance, and legal services may be more accepting because each component will have been vetted by a regulating body to follow the law and ensure data security. If it works for others in the space, why not follow suit? The only customization allowed would be with the large language model (LLM).

Another offering Hercules AI has is the RosettaStoneLLM. It’s built off Mistral-7B and WizardCoder-13B and has 7 billion parameters. Companies using this can convert structured data from within spreadsheets so it can be mapped and transformed by the AI. Imagine the volume of database exports and spreadsheets a regulated business might possess. Converting these files into useful systemwide data can be costly and time-consuming. This LLM is designed to turn large volumes of structured data into something that aligns with your internal workflows.

The company claims that early results show RosettaStoneLLM can do tasks like entity mapping and code generation better by up to 30 percent than general models of GPT-4.

As a company spokesperson told VentureBeat: “In the corporate world, particularly in sectors like insurance and finance, structured data is crucial for operations. Consider an insurance firm preparing a healthcare insurance quote for a client. The client submits their HR database information, which often comes in various formats and structures, with unique naming conventions for columns, rows, and abbreviations. Traditionally, converting this data into a format compatible with the insurance company’s internal systems for analysis is time-consuming. However, RosettaStone can achieve this transformation in mere seconds, a task that would otherwise take human hours to complete.“

Based in Campbell, California, Hercules AI has raised $12.1 million in venture funding, and while it didn’t disclose customer numbers, it did say it’s used by Fortune 1000 companies and 30 percent of the top law firms in the U.S.