• by Laurie Sullivan , Staff Writer @lauriesullivan, April 22, 2024

Firsthand is preparing to run a pilot of its AI Agent this spring, says Jon Heller, co-founder of the company.

The purpose of the pilot is to understand how consumers react and learn how the systems can improve. From Heller's description, the pilot will be kept to one or two publishers and brands.

Firsthand is an AI agent platform built for brands and publishers. An AI Agent is software-based technology that can understand language, learn from data, and take actions to achieve goals automatically. It is based on no code programming.

Heller, along with Michael Rubenstein, founder and GM of DoubleClick Ad Exchange and Wei Wei, cofounder and CTO of Firsthand and a Chinese entrepreneur, are concerned about the impact that generative artificial intelligence will have on publishing and media.

With the advent of large language models (LLM) training on data with little to no attribution, acknowledgement, or compensation give to publishers, the companies see control slipping away.

Heller said AI applications can give back control of the user experience, data and content for publishers -- as well as the ability to monetize by leveraging AI technology. The key is finding a way to better personalize ads and accelerate the process of content marketing.  

These AI agents, as Heller calls them, can take multiple forms. The result is a collection of AI capabilities that can present content, navigate someone to the correct location and construct landing pages on the fly.

When someone sets up an agent, is is a process "similar to a brain that can understand context,” he says. “It’s wrapped in a campaign, giving a business the ability to answer questions in a chatbot. Let’s say someone has questions about a BMW, so they go to the dealer’s website to get answers.

The AI chatbot is operated by BMW using BMW’s data to answer the questions.”

It also can work inside news articles on a publisher website, on a retailer’s website, in a search engine, or a brand's website.

One agent might answer questions on a financial page and other on an entertainment page. The AI agent adapts to the circumstances and questions.

The agent feeds off what is available by the publisher. Heller calls this “retrieval augmented generation” -- meaning that it does not make new content. It only provides information that is known to be true, he said.

“We think the consumer should be able to ask the manufacturer of the content or product,” Heller says. “This is a way to do that without having to dig through the website.”