Quick summary:
- Benzinga pays its editors to learn AI, and one senior editor now does nothing but experiment with it
- Brokerage clients still demand a human checking every piece of financial news before it reaches investors, and reporters who can call sources and land exclusives are becoming the hardest to replace
- The open internet is heading toward pay-to-play distribution, and Olesen expects consolidation across financial media as independent publishers get locked out
Brad Olesen's internal reports at Benzinga, one of the largest financial news providers to North American brokerages, now run 85 pages, every one of them generated by AI. He doesn't have time to read them. So, he feeds them into another AI and reads the summary instead. The newsroom jokes about it.
Olesen, the company's chief content officer, has spent two decades in financial journalism, starting at Bloomberg and Seeking Alpha before joining Benzinga a little over four years ago. He has watched the industry automate itself wave after wave, and AI is moving faster than anything before it.
On "The Disruption Is Now," Olesen told host Greg Matusky how he restructured his newsroom around AI, why he converted editors into full-time AI experimenters, and why he believes the open internet is closing.
Watch now:
Key takeaways
Reporters who pick up the phone are gaining value
Twenty years of commoditization in financial media have turned structured data into something anyone can buy. Earnings estimates, stock movements, and analyst ratings all flow through machine-readable feeds, and AI can pull together a Wall Street bank's view on a stock from a data warehouse, compile analyst opinions, and format everything for publication.
However, "it can't pick up a phone and make a phone call to that analyst on a sell side research desk or it can't talk to a new source or get an exclusive that only you can get through our relationships that we've formed over many years," said Olesen.
Even Benzinga's most AI-native brokerage clients still want a human performing a final sanity check on any piece of news before it reaches investors. The comfort level varies by client, but the demand hasn't budged.
Fake author scandals turned transparency into a competitive advantage
Over the past two years, several financial publishers were caught running AI-generated bylines attributed to people who didn't exist.
Olesen watched those scandals unfold and decided Benzinga would go the opposite direction. The company now discloses AI use at the bottom of every article. If a story carries a human byline, a human editor has reviewed and published it. Automated content produced for certain clients is labeled explicitly.
"I wanted to be very, very careful that we were disclosing what we were doing and that we were using AI to help expedite the process of creating content, but not fully writing the content," Olesen said.
AI handles the “legwork,” according to Olesen. It pulls data from Benzinga's internal warehouse, compiles analyst views, and assembles research. The reporter then applies judgment about what to keep and which angle to pursue.
Content exhaustion is already setting in
Social media was supposed to be the guaranteed path to reaching audiences, but companies that bet everything on it are watching engagement decline as more people try to disconnect. Some publishers have even abandoned digital content altogether and chosen to rebuild their revenue around events and in-person networking.
Olesen thinks the pendulum has swung too far to one side, as it typically does, and will correct. Readers are getting familiar with AI's writing, too. Last year, certain styles AI tended to use were easy to identify. Now, the "not this but that" construction is the new giveaway.
Benzinga allows AI-assisted content as long as it follows the newsroom's traditional writing and passes the smell test. "You need to have that proper lede and you need a nut graph and you need to have that conclusion," Olesen said. AI copy is easy to identify because it's verbose to the “nth degree.”
Editors who know the business make better AI experimenters
Over the last couple of years, Benzinga built a large data and business intelligence team, with people it elevated and hired who work in machine learning and AI. They help guide the tools Benzinga builds in-house.
Olesen also moved editorially focused people into AI-focused roles. He calls them forward deployed engineers, but in reverse. These editors already knew what happens on the front end, and they went back to learn the engineering side and AI experimentation. One senior editor, who was responsible for a lot of the content that went out the door, is now solely tasked with experimenting.
"I basically told a couple of our team members and some of my managers that I'm gonna pay you to learn," Olesen said. "Take some time, build your own programs. What would you do to help your time? And I will pay you to do that."
Those editors now bring new capabilities to the table as each frontier model arrives. They also look for ways to layer in Benzinga's exclusive information so its content stands apart from the rest of the internet.
The open internet has an expiration date
AI platforms and search algorithms already determine which content reaches readers. As these systems grow more targeted, the cost of reaching anyone goes up. Independent blogs and upstart media companies will find the path forward nearly impassable.
"If you're a blog, if you're an independent upstart, it's gonna be very, very hard for you to pierce that veil right now that is being largely gate kept by a lot of these AI machines," Olesen said.
He expects consolidation across financial media as a result because companies will need the scale to afford distribution in a pay-to-play environment. A mature product across every customer surface won't be optional, and paid media will consume a much larger share of every publisher's budget.
Strategies that depend on organic reach will lose ground as AI gatekeepers tighten control over distribution channels.
Key moments
- Why Benzinga now serves "the professionals, the retail, and also the machines" (2:47)
- The value of the human reporter keeps rising as financial data gets commoditized (3:56)
- Disclosing AI use after publishers were caught with fake authors (6:00)
- Where Olesen draws the line between AI-assisted and AI-generated content (7:00)
- How AI copy fails a newsroom's smell test (11:20)
- Why Olesen reassigned editors to full-time AI experimentation roles (14:29)
- The reason people in the trenches belong in product meetings and engineering syncs (19:34)
- State-backed aggregators and the case for a diverse media diet (25:23)
- Why Olesen expects the open internet to become pay-to-play (29:30)
Q&A with Brad Olesen, chief content officer at Benzinga
Q: What should communicators understand about how news decisions get made?
A: Many people don't know how news is made in the first place. They think news just “is.” But news is created.
There's not necessarily a bias associated with it, but there's a process. There are news makers, people who say things, and companies that say things and have to disclose things. It's ultimately up to an individual who has learned what is newsworthy and what news judgment is. For instance, is there size? Is there scope? Is there historical precedent? Does it move stocks? Does it move other equity values?
In financial media, we have purposeful blinders. We focus on what the market is telling us and where the money flows. We don't have to conjure up crises or sensationalize things that might be more mundane than they actually are.
Q: How do you keep your media diet from getting too narrow?
A: If you don't go out and actively find new sources and force yourself to take in alternative information, you're going to struggle. Your social media feeds will show you the same things you already love. But that's not what the world is.
There are companies that have been funded where their entire job is to curate a world on a very local basis for many people, and those strings are being pulled by state actors. You need to be very careful about who backed your sources, who supports them, and what their motives are. If a business seemingly came out of thin air and has a backstory that doesn't quite match, you should wonder what they're feeding you.
Q: How do you handle the process of building AI tools?
A: There's frustration with the speed of development across every organization. You feel like, why don't we just have it now? So, it's important to me that we include a lot of my management in our product meetings and some of our engineering syncs. They see that there are professional processes that need to be adhered to. There are sprint cycles, and you just have to get familiar with them and prioritize what works for you.
