Artificial intelligence-native search startup Objective Inc. is tapping AI to improve the quality of its search results.

Today it announced Anton, a new AI agent that can help its search service customers to boost the relevance of their search results. It does this by evaluating them much like a human user would.

Founded by machine learning experts who previously worked at firms such as Apple Inc., Amazon.com Inc., Google LLC, Meta Platforms Inc. and Twilio Inc., Objective is the creator of a low-code platform for building “multisearch” tools that can be embedded into any website or application.

But Objective’s search experience is quite different from traditional services such as Google Search, Microsoft Corp.’s Bing and DuckDuckGo. It’s a multimodal search engine that’s able to handle not just text, but also images, video and audio, and its outputs are delivered in the same four formats. The company says it generates contextually relevant results and allows people to search with their natural voice, or even combine their voice with images, text queries and so on.

The startup, which raised $13 million in funding last October, sells its multimodal search platform to various types of companies, including retailers that need to integrate a product search function, and news websites that want to make it easier for users to search their archives.

Objective’s customers include Lessin Media Co., which operates the popular technology news website The Information, and Brewster Home Fashions, one of the world’s largest suppliers of wallpaper and home decor.

The Information’s search capability is powered by Objective, enabling readers to search its website for archived articles using their voice, so they can say complete and meaningful sentences to obtain the desired results. Meanwhile, Brewster Home Fashions uses Objective to power its product searches, enabling customers to upload a specific color they like, for example in order to find something that matches it.

While The Information is focused on using search to boost engagement with its readers, Brewster Home Fashions simply wants to sell more products, and Objective’s search experience seems to have done the trick. Since it started using Objective, it claims to have seen an 11% improvement in its checkout rate and a 9% boost in conversion rates.

Turbocharged search engineering

Those customers will now be able to refine their search experiences even more, thanks to the arrival of Anton, which can be thought of as a generative AI search engineer.

When customers sign up to use Objective’s search tools, they generally need to do quite a bit of work to optimize its out-of-the-box search experience and ensure it delivers more relevant results. This process is known as “search engineering,” and it involves writing code and using lots of data to train machine learning models that aid in search refinement.

As Objective explains, search engineering is a tricky business that involves understanding both the user’s intent and the content on the site or app. It also involves knowing how to comb through that content and retrieve and rank the results in the most optimal way.

Traditionally, search engineering has been a fairly mundane task because there’s no easy way to evaluate a search engine’s outputs, other than to sit down and look at them. This burden falls on humans, who need to sit and work their way through various queries and observe the results, then tinker with the algorithms under the hood to try and improve their relevance. That’s why rapid iteration has been extremely challenging, as teams are limited by the pace at which humans can work.

Anton changes that. Although Objective notes that humans are “unequivocally the most precise judges of search results,” it believes most of its customers can tolerate slightly less precise results in order to accelerate their search engineering efforts.

Objective said Anton can be used for tasks such as the ad-hoc evaluation of search results, building real-time search observability systems and more. It’s being made available to the startup’s customers starting today, the first of several AI agents it intends to release in the coming months. Ultimately, it wants to give its customers the ability to evaluate thousands or even millions of search results per month programmatically.

Objective co-founder and Chief Executive Pablo Mendes said his team possesses a wealth of experience in search engineering, as it has been involved in building some of the largest search engines ever made.

“In each of those teams, at each of those companies, we all needed a tool like Anton, almost every day,” he said. “This would have absolutely changed the way my team at Apple worked.”

Image: Objective