You.com Launches ARI Enterprise, an AI Research Agent for Financial Analysts and Consultants

In February, the enterprise AI platform You.com introduced its Advanced Research and Insights (ARI) agent, designed to perform deep research and act as a personal AI consultant. Now, just three months later, the company is rolling out ARI Enterprise, a new version designed specifically for consultants, financial analysts, and researchers. And according to You.com, it outperforms OpenAI’s Deep Research in head-to-head testing.

“The best AI analysts and researchers connect company internal knowledge with the best information on the web. Making both useful is critical for getting the right answer,” Richard Socher, You.com’s chief executive and co-founder, says in a statement. “ARI Enterprise represents a paradigm shift from periodic, expensive research projects to continuous, trusted strategic intelligence.”

A Deep Research Platform Built Around Accuracy

ARI Enterprise piggybacks off the original ARI to analyze all critical data sources, from internal documents and data to that found on the web and premium databases. This business agent provides insights using customizable and visually rich reports.

In an interview, Socher tells me that ARI’s enterprise focus was because “that’s where we saw the most traction, customers having real needs. Already, we have massive consulting firms and hedge funds use You.com a lot, and so we’re leaning into that and building something that’s custom for them with their specific premium datasets in their own company internal datasets, so they can be more productive.”

You.com contends that ARI Enterprise delivers four key capabilities. First, it can analyze more than 400 sources at once from across the web, private documents, and premium databases, giving users confidence that no critical insight is overlooked. Second, it features a proprietary, model-agnostic reasoning layer that filters out noise and surface connections and insights, which it says are often missed by other deep research agents. Third, the user remains in the loop at every step. And lastly, ARI Enterprise supports continuous research and monitoring without usage limits. This allows for an always-on strategy that most competitors may not be able to provide.

When compared against similar offerings from OpenAI, Perplexity, and other competitors, ARI Enterprise is said to surpass them all. When benchmarking complex consultant/investment research questions, it bested OpenAI’s Deep Research “three out of four times, with a 76 percent overall win rate.” And in a FRAMES benchmark study customized for deep research, You.com boasted ARI Enterprise scored an 80 percent on accuracy, “the best known performance of any AI model in this study.”

Socher and his team emphasize ARI’s high accuracy, and it’s evident why it’s crucial. If you’re in the financial space or conducting scientific research, ARI must return correct information. To do so otherwise means monetary consequences or worse.

This isn’t You.com’s first rodeo when it comes to deep research. Socher shares that his company was among the first to launch a deep research bot—even before the term “agent” became mainstream. It has previously released deep research compute capable of programming and running complex data analysis on a user’s behalf, and also developed a creation agent that can generate images for you.

A New Standard for Business AI

Along with ARI Enterprise’s release, You.com is debuting a new open-source model benchmark that promises to show developers “more complex kinds of queries that you’d actually see in real-world business problems,” along with how they’re evaluated. Socher states that this benchmark used OpenAI’s most advanced model to compare how its deep research agent answered complex knowledge questions versus ARI Enterprise. The results indicated that You.com’s solution outperformed OpenAI’s.

“That is a non-trivial feat,” Socher declares. “It’s something we’re very proud of and something that will change the lives of many people and companies that work with You.com.”

The company isn’t a stranger to establishing benchmarking. However, seeking to avoid garnering public distrust, it’s fully open-sourcing it so anyone can run their models against it and see for themselves. Socher states that it’s going to run like an automated evaluation, meaning “people can actually evaluate automatically and compare their own models and offers with ours, and so that is something we think is important for transparency so that people can continue to trust us. Obviously, our customers that are sophisticated enough to run their own benchmarks already choose You.com, but most companies aren’t able to run their own real benchmarks, and then might be stuck with inferior solutions to You.com, and so we want to make that easier.”

“We think that a very important part of AI is to have transparent benchmarks,” he continues. “And there is no really good benchmark for complex, real-world business questions.” He asserts that there are many benchmarks for “quick, simple factual things” such as “who is the president of France?” However, when asked to generate a 20-page report, “and then have someone go through the 20 pages is actually non-trivial. It’s a lot of work. So you have to help people with those kinds of evaluations.”

Featured Image: AI-generated image of human surrounded by layers of data streams. Credit: Adobe Firefly