Bloomberg’s Chief Technology Officer there is, by his own admission, a job that is half about building the future and half about stopping the future’s more awkward impulses from getting anywhere near production. “Half my job is to keep the crap out of the company,” Shawn Edwards says the Observer, and he’s not entirely kidding. The nonsense he’s referring to is the tidal wave of generative AI noise that has swept every boardroom with a Bloomberg The terminal. His job is to find the narrow section between what his engineers can dream up and what the people who actually trade bonds, screen loans and prepare for earnings calls are desperately trying to achieve – not what they’re doing today, but what they’re actually trying to achieve. In Edwards’ worldview, this distinction is the organizing principle behind ASKB, the conversational, agent-based AI system that Bloomberg has built directly into Terminal.
Ask Edward for a use case and he describes the old way of doing things. “Prior to ASKB, a user would have to go to different places in the Bloomberg Terminal to look at company fundamentals, to see what the street ratings are, to look at various performance measures, company KPIs, and another place to look at the printout, peer analysis and alternative data for that quarter’s performance — reading a lot of company news and constantly reading research papers and a lot of the company.” Edwards pauses. “And then they would have to synthesize all this information.”
Since FebruaryBloomberg users can ask the system (ASKB) to synthesize information. “He knows where to go and he knows where to find all this information and he gives you a detailed enough analysis so that you can be prepared.”
The machine does not replace the analyst, and Edwards is careful, almost insistent, on this point. “ASKB doesn’t do all the work for the analyst, but it does 80 percent of the work of gathering and synthesizing that information. And it frees them up to do the value-added thinking—to really decide where to really dig.” Workflows vary by desk. For equity analysts, it is event preparation and thesis monitoring. For credit analysts, it is liquidity analysis and bond review. Common initiation is a research problem.
Faith is not a feature
If there’s a single obsession that runs through Edwards’ narrative for the past several years, it’s the plausibility of an engineering discipline tied to a technology that was never built for finance in the first place.
“A big focus, basically a focus for the last two years on my team, along with our core engineering team, has been how to build AI that is reliable for our customers to make mission-critical decisions,” he says. Among the principles that went into the development of ASKB was the refusal to let the model speak for itself. “We never want her to generate an answer from her world knowledge,” Edwards explains. Rather, ASKB is guided by and based on Bloomberg’s decades of proprietary data, risk analysis and price generators — what Edwards calls “sources of truth.”
Getting there means building validators into every step of the process – some fact-checking in real-time (“you didn’t make up a fact, I can check it… you summarized it, and I can look at all the facts and compare the two”), others catch more subtle failures, like an inverted sense of being read. Layered on top of this is a continuous evaluation framework, “autonomous and manual evaluations to verify that we are actually moving forward, nothing has moved and nothing is going wrong with our system.”
The last layer is transparency. ASKB tells the user about the source material – the paragraphs, from millions of documents, that provided knowledge. It shows users the question that asked them. He shares the analytics call he made. Edwards is open about how underrated this is from the outside. “They’re underestimating this wholeness of these different layers that have to work together to get something that’s reliable,” Edwards says, of the handful of clients who have tried to replicate what Bloomberg does with off-the-shelf AI models. “It’s not easy. It’s actually very difficult to direct the AI to do that. It wants to be very helpful and sometimes it’s not helpful.”
Part of the difficulty lies beneath the model, in the unglamorous work of getting the data sources to talk to each other. “How do you connect data? How do you rationalize all these different sources of information and rationalize a data model that you can bring together across different data sources?” More importantly, Edwards argues that the people leading that work can’t just be AI engineers. “Domain experts are increasingly building our systems. They’re the ones who help guide AI—who say, ‘No, you’re wrong.’
Even the CTO has a learning curve
For all the engineering rigor, Edwards acknowledges something strangely human. The tools are more difficult to use than anyone initially expected – including yourself. “There was a lot of expectation at first that it’s so natural to use, to have a conversation,” he says. “But I think we all learned that, in fact, there is a learning curve. You have to put in the energy and effort to learn how to be a good user of these tools, and the more you put in, the more you get out. That was maybe a little surprising from the beginning.”
Bloomberg is building toward more personalization to smooth that curve, allowing users to tell the system their preferences, their coverage universe, their habits, so that “over time the system gets smarter about how the system will react to your questions.” But Edwards is honest that this is early days, and that it raises a really unresolved question in the AI industry about how much memory is too much. “There’s a lot of research and a lot of tools coming out about memory—how to compress memory, how to use memory, how much of past interactions do you want to use? Right now, we’re very strict about NO withholding certain information, so it’s kind of a balancing act.”
Bloomberg’s current CEO, Vlad Kliatchkostarted the same year Edwards did, and the two rose through the ranks of engineering together—a detail Edwards offers as evidence of a company you either “really take and stay for a long time, or leave very quickly.”
Culture traces to Michael Bloomberg‘s open, intentionally flat work environment, where “anyone can come up with a great idea.” Edwards describes his contribution as protecting and scaling that instinct, creating the conditions for cross-functional collision between a technologist, someone with an arts degree, a former portfolio manager and an operations person, all on the same board. The construction of ASKB required tearing up that organizational chart, at least temporarily. Bloomberg’s traditional products are built by discrete teams owning discrete functions, with UX partners designing individual screens. An AI system that spans every domain in Terminal doesn’t fit that model. “The surface of ASKB and the way it works is different from the way we build other systems,” Edwards says. “How you work together, how you build this, is completely different… applying our old structures to this new way of making the product just didn’t work. We had to change our thinking.”
Hiring for the ability to explain
Asked what he looks for in the people he hires, Edwards doesn’t mention credentials. It recruits for communication—the ability to take a complex idea and explain it on multiple levels, as a physicist might do to a child, a college student, or a PhD student. This requires the ability to listen and be flexible with your ideas, especially within cross-functional teams where the value is often not the new idea itself, but the discipline it enforces. “There is some research that says that cross-functional teams work better with a variety of backgrounds, not necessarily because there are just new ideas, but because it makes each player work harder at expressing their ideas, and therefore they think better about the problem.”
As for why talent wants to work at Bloomberg in the first place, Edwards points to the sheer breadth of data—commodity patterns that feed weather data, document analytics that power search, streaming prices in real time—because “finance is the world” and “many, many different aspects of the world affect finance.” Edwards also notes the speed of Bloomberg’s influence. “You can build a feature or a product or a capability and actually go see customers using it, talk to them, get feedback from them. That’s exciting.”
Pushed on what has shaped his thinking, Edwards cites a history of Bell Labs and its cross-functional glory days—unsurprising, given his obsession with collision-prone teams. Less expected is his next choice: Hermann HesseS ‘ Steppenwolfa novel that taught him that people pigeonhole themselves into a limited view of who they are, when in fact “we can take different forms and use different parts of our personality and mind to grow, and we can bring out different abilities at the same time.” His career has progressed almost exclusively through periods of self-imposed discomfort. “Unpleasant challenges”, he calls them.
What’s Next for AI at Bloomberg? Edwards responds with something almost surprising. “We’re just scratching the surface of what generative AI and our approach will do,” he says. “There’s so much more to our vision of what we can achieve. This technology has allowed us to dream bigger and tackle problems that we had dreams about but just couldn’t build. Now we’re able to build it.”
