The proliferation of AI tools in finance is fundamentally reshaping how Wall Street analysts operate, with proprietary vertical data and near-zero tolerance for errors emerging as the critical moat for generating outsized returns in financial markets.

In a recent episode of the investment podcast How I Invest, Chris Ackerson, Senior Vice President of Product at market intelligence platform AlphaSense—which has surpassed $500 million in revenue—shared deep insights on how AI is transforming analyst roles, the commercial value of proprietary data, and the future applications of AI agents in finance.

As generative AI sweeps across industries, markets remain fixated on fears of labor displacement and the real monetization potential of AI tools. Ackerson offered a clear market observation: AI is not eliminating financial professionals but dramatically reallocating their time and resources.

"As our systems become more powerful, we can take on a lot of the tedious work—the manual tasks analysts used to stay up late doing," Ackerson noted. "Now they can focus on higher-value activities like meeting with clients and engaging with company developments."

These efficiency gains translate directly into measurable business results. Ackerson revealed that one major investment bank, after carving out a portion of its business, ran several months of testing that proved AlphaSense could increase average coverage by 15%, significantly boosting banker productivity. This means financial institutions can expand their operational footprint without adding headcount, underpinning the sustained growth logic for AI tools.

Why horizontal LLMs stumble in finance and vertical AI breaks through

Despite the rapid advancements of broad-based models like OpenAI's offerings, their commercial adoption in finance—a domain with exceptionally low tolerance for error—faces substantial headwinds. This is precisely why investors are scrutinizing whether vertical AI players can establish defensible technical moats.

Ackerson pointed directly to the weakness of general-purpose models: "Just last week, a leading AI lab released data showing error rates between 5% and 9% when using leading market data. That's completely unacceptable for any serious investment professional. Any error rate prevents true AI adoption."

To mitigate hallucinations and reduce error margins, vertical AI's solution lies in extreme data control. By aggregating data, indexing information, using AI to read every line of every document to understand cross-source relationships, and then integrating these into agent systems for precise extraction, vertical platforms minimize mistakes. Ackerson provided a compelling statistic: "By using frontier AI models within the AlphaSense platform combined with our proprietary data and tools, we can deliver assessments three times higher quality at lower costs. This is the true advantage of vertical integration."

Proprietary data as the ultimate moat: "Most of the world's knowledge isn't written down"

As AI labs shift competitive focus from pre-training to post-training, proprietary data—which directly impacts model quality—has become the scarcest and most valuable asset in the industry.

"All AI companies will eventually transform into data companies. We see this trend intensifying as more players emerge," Ackerson stated.

In financial due diligence, channel checks and expert interviews serve as vital sources for generating investment returns. One of AlphaSense's primary moats is its vast archive of interviews with top global investors and industry experts. The repository currently holds over 300,000 interview transcripts, expanding by more than 25,000 each quarter.

"Most of the world's knowledge doesn't exist in written form—it lives in people's minds," Ackerson emphasized. Building on this proprietary data, AlphaSense developed its AI Interviewer system, unlocking substantial commercial potential. Whereas analysts might previously have conducted five expert calls, they can now leverage AI agents to perform twenty or even fifty deep-dive interviews across different sectors, extracting quantitative data and predicting earnings guidance before reports are released. The system's interview quality "has already reached or even exceeded the level of our best analysts."

The future battleground: slashing costs by 40x and the premium on human judgment

If every Wall Street institution deploys the same AI tools and datasets, where will hedge funds find their edge? This is the ultimate question concerning the long-term logic of financial AI investments.

In the research phase—which constitutes approximately 80% of total financial research costs—blindly using frontier LLMs to search the web is not only expensive but also introduces irrelevant information, causing what Ackerson calls "context contamination." By training proprietary search models that precisely call upon the right tools, he revealed that AlphaSense can "deliver higher quality than frontier models while reducing costs by up to 40 times."

Looking five years ahead, the deliverables of the financial industry will shift from static documents like slides and memos to real-time updated custom software and agent ecosystems. A "Super Analyst" will work around the clock like a tireless researcher.

In this environment, the depth of human-machine collaboration and human judgment will ultimately determine investment outcomes. "How you direct the system, the questions you ask, the choices you make when interacting with AI—these are the differences created by human factors," Ackerson concluded, citing his CEO's framework for defining future enterprise value: "A company's value is the sum of all its decisions. Applying judgment on top of AI systems' decisions will always be critical."

Fireside conversation highlights

Throughout the interview, Ackerson detailed how AlphaSense addresses AI skepticism through rigorous evaluation frameworks. "We've built evaluation systems that comprehensively test all AI systems against detailed criteria for every task users perform," he said. The platform combines automated and human review, employing financial analysts to sample and validate outputs.

Addressing the technical architecture, Ackerson explained the evolving selection of models: "We're seeing meaningful differences between models—even at the frontier. Much of the differentiation now comes from post-training, where companies specialize models for specific tasks. For example, Claude Opus excels at generating high-quality slides but doesn't perform as well as a context retrieval or search system."

On partnerships, he noted, "We work with Cerebras, an excellent low-latency inference provider, to deploy top-tier open-source models. Low latency is crucial for building conversational, real-time systems." The company is also developing its own models, leveraging a decade of proprietary search history to fine-tune performance.

Regarding market volatility concerns driven by AI symmetry, Ackerson reframed the role of technology: "Our mission is to help markets allocate capital more efficiently. By integrating high-quality data into decision processes—one of AlphaSense's core value propositions is eliminating blind spots—users gain confidence that their decisions are more likely to benefit their business."

He also addressed the future of software delivery, confirming that "the market is moving toward agents collaborating via protocols like MCP." AlphaSense's Super Analyst already operates within PowerPoint and Excel, with plans for broader integration across systems and workflows.

On the biggest challenges, Ackerson cited maintaining trust and verification at scale: "We're intensely focused on delivering decision-grade answers. Our priority is ensuring users can close their laptops on Friday night confident about the reliability of data sources—without worrying about fabricated metrics or financial figures."

Reflecting on industry lessons, he offered advice for aspiring professionals: "Everyone in AI has underestimated how quickly models improve. We've learned to be unafraid of removing scaffolding and moving boldly toward the future."

"Maintaining a beginner's mindset is essential," Ackerson added. "You have to wake up every day like a novice because everything is changing around you. Our investments in content datasets, search data, and millions of search queries will continue to power the company for its entire lifecycle."