AI in Information Services: Why Trusted Data Matters More Than Ever

00:21, Chris

Hi everybody.

Welcome back to William Blair Thinking Presents.

Today is Friday, August 28th, and we're discussing one of the more interesting debates emerging from the AI economy, which is whether artificial intelligence poses a threat to information services companies or whether it could ultimately make their data assets even more valuable.

I'm joined by Andrew Nicholas. He is a William Blair Global Services analyst covering information services, consulting, and HR technology.

He actually recently just published a report titled AI in Information Services Revisited: Momentum is Building, Sentiment is Lagging, Time to Lean In that revisits a framework he first introduced in 2023 for understanding how artificial intelligence could impact information services businesses.

So, Andrew, thanks for joining us.

01:05, Andrew

Thank you, Chris. It's great to be here.

Obviously, one of the more, if not the most important debate across our coverage. So, I'm excited to talk about it and how our thinking has evolved on this topic in particular.

01:32, Chris

Absolutely.

So, one of the key conclusions from your report is that while investor sentiment toward the sector has weakened, the evidence increasingly suggests AI may be creating opportunities rather than disruption.

Starting at a high level, you know, for listeners who may not follow information services closely, what businesses are we talking about, and why has AI become such an important debate for the sector?

01:41, Andrew

Sure.

At the highest level, information services companies own, aggregate, organize, and analyze data that other businesses use to make decisions.

This is primarily a B2B business.

That includes credit and employment information from companies like Equifax and TransUnion.

Insurance data and analytics at Verisk, financial and ratings information at Moody's and S&P Global, investment research data at FactSet.

The list goes on and on.

Many of these companies also offer software, analytics, and workflow tools, but the common thread is that data and information are the foundation of these companies' value propositions. The AI debate is essentially whether making information easier and cheaper to source, to structure, and to analyze reduces the value of these companies.

I think the bear case says that clients can use general purpose models or alternative data sets, or even internally developed tools, to recreate some of what they have historically purchased from third parties.

I think the bear case is also that AI agents could also reduce the value of the traditional user interface, because the agent is retrieving the answer without the customer really ever having to open up a platform.

I think there probably is some truth to individual pieces of that bear case, particularly for products built primarily on easily accessible information. Or products that are mostly differentiated by the interface that they operate or sell through.

But we think the broader narrative overlooks what customers are actually paying for. We think they're not just paying for access to a number. They're paying for accuracy, structure, historical context, auditability, in many cases, and maybe most simplistically, they're paying for the confidence that the information that they're looking at or that they're digesting can support a decision, an important decision in a lot of cases.

And that really matters in investment banking, that really matters in asset management, and insurance and in major decisions within CPG.

And in a lot of those environments, there's a major distinction between pretty good or pretty accurate. And I think that a lot of those need it to be very close to perfect.

So, our conclusion is that the market has largely treated AI as a substitution problem or as a substitution event, and I think it's more likely to become a consumption event or a productivity event.

And I think AI should allow these companies to create more information, make it more accessible, and allow them to operate more efficiently.

And that's really what we kind of outlined in our original report in 2023 and kind of touch back on or revisited in our latest report.

04:34, Chris

So, you first laid out a detailed framework around AI and information services in 2023, as I mentioned earlier.

Three years later, what has played out as expected? And, you know, what would you say has surprised you?

04:45, Andrew

Yeah, I think, you know, even before getting into what surprised us or what's changed, I would emphasize that we are as confident in, kind of, the core thesis today as we've ever been.

I think in 2023, we argued that AI would benefit information services companies through three sector-wide tailwinds: the greater supply of data and content, stronger demand for information and data, and then improved operational efficiency.

Three years later, we see evidence across all three.

Companies are collecting and processing data more efficiently. They're introducing new AI-enabled products. They're seeing higher consumption from AI-oriented clients in particular, and they are seeing measurable productivity benefits.

I'm sure we'll — or I suspect we'll — hit on some of those examples later.

What has surprised us is less, kind of, the direction of the thesis, and more the timing and the way that the benefits are materializing.

I think the technology has been there for some time, but the scaled monetization has taken a little bit longer than what we initially expected. I think information services companies have needed time to productize the technology, even though they've been working with it for several years, much, much before the generative AI wave. And in public consciousness, they have been, kind of, bleeding edge when it comes to data science and machine learning.

But customers also have needed to be ready for it themselves. They’re addressing governance, privacy, integration. And I think those hurdles are particularly meaningful in regulated end markets where a lot of these companies operate. There's strict requirements there around governance, around security, around auditability.

And that has slowed some of the adoption or some of the demonetization over the last couple of years. I think that's the major thing.

There's also some other more nuanced changes to our thesis. Really quickly, I'll hit the four that we outlined in the report.

The first is proprietary data has become even more important to how investors think about these companies than we had previously appreciated. We always understood this to be a key differentiator, but in a lot of cases, it's the primary and only thing that investors seem to be reacting to at this point.

The second is that — and this is what I already hit on — is monetization has been slower. But we do think that incumbents are probably better positioned than we appreciated previously. Some of that is just the ability to have that trust with their clients. And that's a key part of initial AI spend in our view.

And then third, there's seat-based exposure across these organizations. That's a real risk. But it is uneven across the different models that that we cover. And so that's something that we've had to readjust our thesis and our understanding of these companies' opportunities or risks near term to address.

And then the fourth is just software, workflow, and analytics products. All of those products are embedded in a lot of these impost services names. I think they come with perceived disruption risk. And although we think that concern is frequently overstated, it is something that you also have to address.

So, the short answer here is that we have not changed our view on the ultimate outcome.

If anything, the evidence has increased our conviction, but we have refined the timeline and some of the competitive advantages that will matter most along the way.

08:25, Chris

One of the strongest themes in the report is that proprietary data may be actually more valuable in an AI-enabled world.

Walk us through why that might be?

08:34, Andrew

Yeah, I think proprietary data is valuable because AI cannot generate information that it cannot access. Models can make existing information easier to find, easier to combine, to analyze, but they cannot recreate an exclusive data set simply by becoming more and more intelligent. I think as models and interfaces become commoditized or increasingly commoditized, the ownership of that type of information, scarce information, becomes one of the few competitive advantages investors can underwrite with confidence.

It gives the provider control over access, it supports pricing power, and I think it ensures that customers or AI agents, in some cases, still need to come back to that provider for the underlying information.

That is particularly true for contributory kind of consortium-based data sets. Equifax's work number database is a great example. They have built that through employment and income contributor relationships.

Equifax and TransUnion maintain credit files populated by financial institutions.

Verisk aggregates industry-wide insurance information.

These are data sets as examples that have historical depth, big networks, regulatory relevance. And we think all of those things would be really difficult to reproduce or replicate. AI may improve how customers interact with them, but it does not eliminate the scarcity of that underlying information.

Having said that, and I think this is an important nuance, I do think that investors may overestimate, a little bit, just how important proprietary data is, if they treat that as the only thing that matters.

I think that publicly available information can still be differentiated if there is historical context, if there is linking to original sources, or perfect accuracy or near-perfect accuracy, or if that data has been embedded in customer workflows.

So, I would distinguish between, kind of, exclusive data, raw exclusive data, and the decision-grade data infrastructure needed to make that still valuable. I think proprietary data is still the strongest moat, but accuracy, richness, applicability, historical depth, all those things matter too. And ultimately, that conclusion is what makes us bullish on the sector as a whole.

11:09, Chris

At the center of the report, you know, you argue that AI benefits information services companies through greater supply, stronger demand, operational efficiencies. Where are you seeing the clearest evidence of those benefits today, would you say?

11:23, Andrew

Yeah. So, our report from July, I mean, it was really kind of focused on this, right?

I think, at least in our opinion, there is growing evidence across the sector of all three of those things happening. On the supply side, AI is helping companies create more content, collect more data, do it more efficiently, and develop products faster.

Gartner is using an AI-driven model to help expand its Active Insights Library. Equifax has all of its new models and scores, or at least all the new models and scores it introduced in 2025, were powered by AI. S&P Global used Kensho Link to connect more than 75% of the data sets it acquired through With Intelligence to its platform in less than a month.

So, there are examples where AI is increasing the amount of information available and shortening the time required to collect, integrate, and productize it. And so, on the supply side, we're already seeing it.

On the demand side, I think the evidence is maybe most tangible. A few examples there.

FactSet reported over 450 clients are already engaged with or trialing its MCP server. API call volumes are reaching 13 times the levels that they saw just last quarter. NIQ talked about 25% growth in client data consumption through the first half of 2026, and really nice growth amongst its AI-native clients and products.

Moody's and Verisk have also talked about distributing content through AI-enabled channels and integrating their intelligence into more workflows.

I think the pattern is that clients using AI tend to consume more data, they're consuming it more frequently, and they're consuming it across more parts of the organization. So, both the supply and the demand side of this market are moving in the right direction.

And then the last piece, which you asked about, and which we highlighted as the last vector of our thesis, is on operational efficiency.

Again, a lot of examples there. FactSet that has talked about benefits in software engineering, and data operations, and client onboarding.

Moody's has talked about its AI-enabled customer service assistant reducing staffing requirements in that part of their business.

NIQ has cited more than 40% improvement in customer support productivity as a result of AI.

Equifax described AI as a meaningful part of its $150 million synergy target over the next three years.

So, these savings can support margins, or margin expansion, I should say. They can also be reinvested into new products, more data coverage, faster innovation. And we think all of that is supportive of a better fundamental story, or accelerating fundamental story, for information services names in 2027 and beyond.

14:20, Chris

One of the biggest names in AI today is the rise of agents in model context protocols, which are better known as MCPs.

How would you say are those changing the way information services companies deliver the products and think about growth opportunities?

14:32, Andrew

So just to provide some context. So, an MCP, or model context protocol, is essentially a connector — a standardized connector between an AI application and an external data source or an external tool. So instead of having to build a custom connection between every single model and every single data set, an MCP gives an AI system a consistent way to retrieve that information, to use external tools, and to complete workflows.

For information services providers, that means their data can be available readily and cleanly inside environments such as Claude, such as ChatGPT, or even a customer's internal AI platform, without giving up control of the underlying information, which is obviously important to the extent that that is a major part of their competitive moat in IT and advantage.

That matters because I think agents consume data very differently from humans, right?

A human can log into a workstation, run a search that comes to mind, review the information sequentially.

An agent can do it all at the same time, can query multiple data sets, repeat the same analysis across hundreds, if not thousands of companies and customers, and continuously refresh that work.

The result from an info services perspective is that even if human user growth is relatively modest going forward, we expect machine-driven data consumption to grow much faster, and MCPs are one channel for that.

Essentially, this is reinforcing our increasing demand thesis, and I think MCP-enabled workflows are one component of that. We're already seeing MCPs introduced across the space. FactSet, S&P Global, Moody's, Verisk, and NIQ have all introduced MCP or comparable, kind of, AI-ready connections.

S&P Global and FactSet, which I alluded to earlier or noted earlier, have already reported very sharp growth in client engagement and API calls. NIQ the same for them.

So, we're seeing this as a new channel for data. And it goes back to the point that I made earlier about the value of that data, particularly proprietary data. If you have that, you can monetize it through a variety of different channels and agent consumption. And through MCPs, in a lot of cases, is one way for them to do that.

17:07, Chris

Despite improving fundamentals, many information services companies are continuing to trade well below historical valuation levels. Why do you think investors remain skeptical, and what are the key things you're watching over the next year or two?

17:21, Andrew

Yeah. So first off, kind of on the valuation piece.

I think it's important to acknowledge there are reasons besides AI that the sector's multiple has come down. There are company-specific dynamics at play. A number of these firms have exposure to interest rates, certain end markets — higher rates also have a disproportionate valuation impact on companies that investors traditionally view as kind of long-duration, competitively advantaged compounders.

But as a whole, you're right. There's clearly some skepticism in the market. Investors are, in our opinion, treating the sector as guilty until proven innocent as opposed to the other way around.

And I think investors kind of see the bear case first. It's that AI can reduce switching costs, or it can automate certain analytical functions, or pressure seat counts.

But at this point, because many of the benefits are only beginning to become visible, the multiple has contracted ahead of that.

And I think the encouraging part for us is not only that we're starting to see evidence of that growth, which we just hit on in the last question, but also that, from where valuations sit today, the group doesn't really need to return all the way back to its historical multiple for investors to generate an attractive return.

Even getting part way back would be meaningful.

I think some of that will depend on sentiment. Actually, a lot of it will probably depend on sentiment, but what ultimately changes sentiment is evidence.

Proof that organic growth is stable or accelerating. Quantifiable AI revenue through paid products and enterprise expansion. Measurable cost savings, increasing usage. Some of the path back is simply the passage of time without the bear case materializing.

But the sector also has an opportunity, in our opinion, to prove proactively that AI is a benefit through greater product innovation and data consumption. And that's the evidence that we'll continue to look for over the coming quarters.

19:29, Chris

So, as we wrap up, if investors remember just one thing from your report, what should it be?

19:34, Andrew

Yeah, I think the single most important takeaway is that AI is only as valuable as the information it can access. As models become more capable or even less expensive, I think the demand for trusted, accurate, decision-grade data will continue to accelerate.

That doesn't mean every info services company is equally protected, or that there won't be disruption in certain workflows and pricing models, but the companies that own differentiated data, understand their end markets, and can deliver intelligence through a variety of channels are really well positioned to benefit.

And I think there's a ton of evidence from all the companies in this sector that we cover that those things are starting to transpire. So, we remain very optimistic on the sector as a whole. And I think that that thesis will continue to prove out in the coming quarters and years.

20:29, Chris

Andrew, appreciate your time and insights. And for those interested in learning more, Andrew’s report is titled AI in Information Services Revisited: Momentum is Building, Sentiment is Lagging, Time to Lean In. Thanks for listening to William Blair Thinking Presents. We'll see you next time.