AI is the trend everyone is thinking about heading into 2025. It’s everywhere: in consumer tech, on product roadmaps, mentioned during earnings calls, and of course, being discussed in capital markets. 

I’ve spoken about the impacts I see AI having on the industry: I recently wrote a piece titled AI in Capital Markets: the Map is Mostly Empty (and recently was on a panel at the Unbundling Conference in November to that effect), and to be clear, I think the impacts will be significant. It will impact everything from building models, to distributing event invites, to client coverage, and trading systems. On that last note, I’m still waiting to see which brokerage will launch the first “AI powered algo”. 

One of the interesting results is that, given the demand for content to train these systems, the sell side appears to be (rightly) very concerned about how their content and products are used by clients in AI models/applications. Don’t get me wrong, the sell side has always been focused on protecting their IP, but they seem increasingly concerned about how their content is being leveraged by new systems. There is a perception (whether right or wrong) that clients feeding sell side research, sales notes, and commentary into an AI model is breaking the commercial agreement. 

Additionally, we’re hearing from buy side clients that integrating external content into these models is extremely challenging. Most brokerages don’t have a structured way to ‘feed’ the content to clients (such as via an API), so they end up either having to send it to a generic email address, forward it on internally, or receive scheduled data dumps. 

Part of the reason the buy side is having to develop workarounds, and thus raising concerns from their broker counterparties, is that there is simply no other way to achieve their desired outcome. 

The sell side needs to accept reality: these models and systems aren’t going away. Their adoption is only going to increase, as investment professionals across the industry increasingly incorporate them into their workflows. Systems like chatGPT, but verticalized for the industry, are the future copilot of investment professionals. Firms will leverage industry content from both external sources (like the sell side) and internal sources (like investment memos) to further verticalize their model. 

So, if you’ve made it this far, here is the summary: the sell side has the content, but isn’t quite sure how they feel about firms using it. The buy side wants the content, but they can’t find an easy way to access it. 

Seems like an opportunity for everyone to win.

I’d like to argue a different path for the sell side: that the buy side adoption of AI tools could be a major revenue generator, and not only unlock demand for new products and data that the sell side historically hasn’t had demand for, but also do so at some of the highest margins they’ve ever seen. 

Let’s start with written content. For the most part, sell side content ages quickly. Desk commentary usually loses its value after a few days, or even a few hours. Most research loses its value relatively quickly too, with a few pieces retaining longer term value (such as initiations or thematic pieces). Regardless of the exact timeline, it’s fair to say that the value of content deteriorates quickly: which is part of the reason that sell side firms are constantly producing new content. They need to continually add value. 

Then there is data created off that content, one of the most obvious being interaction and engagement data. Today, clients receive their firms’ interaction data on a delayed basis; usually semi-annually or annually. Clients then use that data for use cases such as broker voting and commission allocation. The data is already being captured, structured, and delivered to the client. Funds could use that data in real time to feed into their model, giving it additional perspective on which content received is the most ‘valuable’.

In both of these cases, brokerages have the opportunity to take work they are already doing (creating content, and organizing client specific consumption data, for example) and provide it to clients in a new form to drive a deeper relationship, and additional revenue. 

The content produced ages quickly when it comes to making investment decisions, but it will have a much longer shelf life when it comes to training models. The sell side also doesn’t need to be in the business of offering the best model. Instead, they can provide the ‘picks and shovels’ – i.e. the inputs needed for each client to make their own model. 

The final push back I often hear from the sell side is “well, we want them to read our research report, not feed it into a model and have it get lost amongst a hundred other reports.” There is a concern that by allowing clients to feed your information into a model it either a) commoditizes it, or b) results in a lack of recognition. 

I understand the concern, but I would provide two counter points:

1) This is essentially what already happens in a client’s brain. They read your report, then ten others, then attend a management meeting, then read some internal analysis. They combine all those different information sources together to form an opinion.  The only difference is now your content is in the firms ‘brain’ – and will likely be more broadly leveraged.

2) It has the potential to drive a closer client relationship. Today, most clients go to aggregators because it’s easier. They can get everything in one place, vs. having to go to several locations. Many large clients have built their own internal aggregators. If all the content were fed into a model, the client could interact with that model, and get driven to the research report, then back to the author/sell side contact. Ultimately, the buy side wants to have a conversation about what they just read (if it’s valuable). Models also could highlight content and sources who’s relevance wasn’t immediately obvious if you were searching via an aggregator. It promotes intelligent and non-obvious discovery. Even if you distrust models, they are definitely the lesser of two evils between them and aggregators.

The same logic goes for a client’s readership data. You’re already providing it to them for their broker vote, why not give them the same data to plug into their emerging models? As an added benefit, it’s a unique data source that no other fund will have access to, giving them something unique to leverage in their internal model. 

As we go into the new year, we’re working with our clients to help them think about how they can support their clients in this new era of AI. This is an emerging use case for the industry, but it also presents a massive opportunity for brokerages to be on the leading edge, and support their clients’ adoption of AI tools (and generate some revenue while they’re at it).

As I said in my last article, the map is mostly empty. We’re in the early innings, but 2025 promises to be an exciting year, and likely the first year that AI becomes mainstream for the buy side in capital markets. The only question is: how can you help?

Your sales note from six months ago likely isn’t valuable to anyone but an AI model. Why not take a win-win opportunity? 

As always, if you have any questions, don’t hesitate to reach out. 

Onward and upward,

Blair Livingston

CEO

Sign up to get Blair’s thought pieces delivered straight to your inbox here.

Street Context’s recent webinar: Winning II Voting Season with Street Context

Blair’s recent webinar: The Evolving Role of Salespeople in the Capital Markets

Previous Thought Pieces:

How well do you actually know your clients? The “Roster” Problem

AI in Capital Markets: the Map is Mostly Empty

Secret email addresses, fax machines, and shared inboxes: the buy side’s solution to email

Ideas for a new Stock Exchange… in Texas

The changing relationship between corporates and investment banks

What will happen to Excel? Evolving for the industry or being replaced?

Does the Bloomberg killer end up being…economics?

When intelligence is artificial: focusing on the basics

Investment Banking: trends continuing into 2024

Bloomberg opens its ecosystem: chatbots to do your bidding

The future of salespeople: extinction or evolution?

When does an industry practice become “table stakes”?

Client coverage and relationship management in a post-COVID world

The end of MiFID driven unbundling, and the start of something new: a UK government backed aggregator?

Generative AI, ChatGPT, and the capital markets

The Cloud Wars have Officially Arrived in the Capital Markets

The SEC Officially Says “No” to Unbundling

A Blockchain Blow-Up You Missed Much Closer to Home (Spoiler: Not FTX)

People are getting fired for using Whatsapp, but no one gets fired for using Email