Why AI Will Either Make or Break Your Credit Research Franchise
Credit research is rarely paid for directly, and AI is making that harder to ignore. What the sell-side can learn from data, exclusivity and, surprisingly, Netflix.

It is well understood that most of credit research is never paid for directly. It gets monetized through trades, or through mindshare that wins more business over time. Even under MiFID II, flows still tend to follow the dealers that bring the most value. As I wrote in Credit Research Can Drive More Revenue Than You Think, the connection between credit research and revenue is real, yet many firms still leave it to chance rather than proactively using their research to win business.
In the age of AI, there is real downside risk in leaving that monetization to chance. Last month I moderated Singletrack's Q3 2026 Benchmark panel and wrote about what AI means for the value of sell-side research. That conversation centered on equities, but below, I take each of its key takeaways and look at what it means for credit.
Credit's thorny client data problem
Firms need a single and accurate source of truth for every account and contact. Some of the challenges to accomplishing this are common to every asset class: individual accounts organized by pods, readership data scattered across platforms, duplicate contacts, and the list goes on. Credit adds its own layers of complication: linking research engagement to activity in individual bonds and loans is complex, for example, and over-the-counter bids and asks often go unrecorded.
But by far the biggest gap is Bloomberg chat. Banks capture these conversations for compliance, but Bloomberg is notoriously guarded about letting them use the data for anything else. With full access to every chat through its API costing as much as $1 million a year, according to The DESK, this is hardly surprising. Will there be pressure to push Bloomberg into making the data more accessible? If not, only the largest firms will be able to afford to use it, creating a new competitive divide.
Fixing these gaps matters. At scale, a model can learn from years of readership, chat, inquiry, and trading history what tends to happen before a client trades, spot the same pattern in accounts today, and alert the right salesperson or analyst as it happens. But as I argued in Building AI Without Fixing Your Client Data? Good Luck, this will not work and will not be trusted if the client data underpinning the analysis is unreliable.
Exclusivity as a strategy
The sell-side hears from every side of the market every day. Research analysts, salespeople and traders talk to dozens of clients about the same names every week. No AI tool or single buy-side firm can replicate this. Within appropriate limits, knowing how other investors are positioned, what is worrying them and where views are shifting is valuable to every client they speak to next. Few firms are deliberate about which clients get this information.
External access is a separate source of exclusivity. The sell-side can convene what investors can’t easily arrange themselves: a call with a restructuring lawyer, a dinner with a corporate treasurer, a roundtable of senior PMs. Should credit be building a more equity-like program for this type of access? It does happen today, but it tends to be very ad hoc rather than designed with intention as part of an overall strategy. A more deliberate program would put the parts of credit research that AI can't replicate front and center.
Automate the simple, invest in the complex
Credit trading has spent more than a decade moving simple flow onto electronic venues. Research has its own version of simple flow: earnings recaps, new issue summaries, relative value screens and the other routine output every credit research team produces. AI can now approximate this work quickly. It still matters and clients still use it, but it is getting harder to point to when asking a client for more business.
What a machine can’t approximate is judgment on questions that are complex and nuanced, where the answer requires interpretation. How will a covenant package hold up in a liability management exercise? Is a political or policy shift about to reprice an EM sovereign before the market sees it coming?
A model can process every public document and precedent, but it can’t predict what sponsors, management teams or governments will actually do. An experienced analyst who has spoken with the people involved, and is willing to put a view on the outcome and stand behind it, can. That is the work that gets a PM to pick up the phone.
AI can do for research what electronic trading did for simple flow: take on much of the routine output. This frees analysts to spend their time where judgment matters most: stressed credits, complex capital structures, and idiosyncratic names. A differentiated view on those names makes it much more likely that clients direct trading to the firm that gave them the analysis and market color.
Without attribution, research goes unrewarded
Research attribution used to be automatic: the client opened a report and saw the bank's name on the cover. Increasingly now, clients ask questions of AI tools that draw on many providers' reports at once and return a single answer. In these cases, firms lose credit for their insights. A client who doesn’t know your analysis was behind their view has no reason to give you business. Sell-side firms should expect any platform that ingests their research to be clear about how it will be attributed and what engagement data comes back to them.
The Netflix case for frictionless engagement
Imagine if Netflix treated its subscribers the way much of the sell-side treats its clients: the same homepage for everyone, no recommendations based on your viewing history, and a separate app and login to watch whatever you finally choose. The business model would fail. By Netflix’s own account, most of what its members watch is suggested to them rather than found through search. A PM choosing research isn’t choosing a show, but the principle holds. Make the right content easy to find and act on, and engagement follows.
Netflix no doubt draws on more than viewing history, but it brings everything together in one place and acts on it automatically. The sell-side sees its clients across dozens of systems and platforms, and still relies largely on memory and manual effort to connect them. Changing that requires the unified client data described earlier, and the AI that can analyze it at scale. Together, they can create a sensible customer journey and point each client to the right next step, whether that's a piece of research, a call with an analyst, a conversation with a trader, or an invitation to an event.
In credit, there is no question that price will always matter most. But when prices are close, the firm that is easiest to navigate will have a competitive edge.
Where do we go now
AI is changing where the value in credit research will come from. The firms that come out ahead will invest in the client data foundation that makes AI worth trusting. They will be deliberate about how they share market color and build access programs with intention rather than leaving them to chance. They will automate the simple work and focus their analysts on the complex credits where a differentiated view wins the trade, and insist on attribution wherever their research is consumed. And they will make themselves easier to do business with than anyone else.
As always, I'd love to hear what you think.
KTB



