Agentic research startup Brightwave Inc. is looking to move beyond the financial realm with the launch of its new “research agents” product, which enables users to control an entire fleet of specialized agents to create comprehensive reports on almost any subject.

Brightwave said its new offering takes inspiration from the system and user interface designs of artificial intelligence-powered coding tools such as Cursor, and makes it simple for anyone to create “long-running” autonomous agents that can perform open-ended tasks without supervision.

The platform uses conversational AI and natural language understanding to infer a research plan based on the user’s inputted request, and from there will define a sequence of activities it needs to perform to accomplish that task. For instance, it will understand that it’s being asked to create a long-form deep research report, and then work out that to do this, it needs to clone a previously authored document, refine and update existing documents, construct information-dense charts or evidence-linked tables, in order to complete what’s requested of it.

Once the research plan is ready, Brightwave’s asynchronous background agent will then unleash multiple search, reasoning, planning, presentation and fact-checking agents to compile the required report. After the background agent believes it’s finished the job, its work is submitted to a large language model judge, which is an independent platform that uses powerful reasoning to decide if the outputted research fulfills the user’s original request.

Similar to how an AI code generator uses unit tests to determine if a generated code sample meets certain standards, Brightwave’s agents rely on “satisfaction criteria” to ensure they only produce high quality and well-grounded, factual research reports.

Brightwave co-founder and Chief Executive Mike Conover (pictured, right, alongside co-founder and Chief Technology Officer Brandon Kotara) said agentic AI systems need to be laser focused on deliverables, in the same way that most AI coding tools already are. “AI’s impact on software engineering is a great example [of what needs to be done],” he said. “Lightweight, asynchronous feedback on complex work products wins against close management every time.”

The startup’s agentic AI research roots can be traced back to the financial services industry. The company came to attention when it raised $15 million in funding last October with a platform that was originally focused on generating financial reports and forecasts, aiming to help clients make smarter investment decisions. It generates financial research reports by synthesizing insights from numerous documents, including earnings reports, conference call transcripts, financial statements, Securities and Exchange Commission filings, news stories and more.

Having mastered financial reports, it makes sense for Brightwave to bring its agentic AI research skills to other industries. The startup said it can, for instance, generate reports on which real estate and energy companies are likely to benefit most from U.S. President Donald Trump’s AI policies, using data from sources such as Whitehouse.gov, public filings, regulatory disclosures, blog posts and news reports to create a comprehensive report with rich visuals. After reading the report, users can ask clarifying questions about anything that gets their attention, or refine the substance of the report.

Conover said Brightwave has been focused on realizing this vision since the beginning.

“From day one, it was clear to our team that this type of autonomous research is where the market would go and every engineering and product decision we’ve made is a reflection of that belief,” he said.

Photo: Brightwave