We held the first meeting of our AI Leaders in Information & Research Providers Network today, bringing together AI leaders from major financial terminal providers, global data and ratings agencies, boutique research businesses and premium economics publishers. The topic: what clients actually expect from MCP, and how they perceive the value of what's being delivered through it.
The sales enablement job nobody's finished yet
A lot of the discussion came back to sales teams and how much they need to understand right now. It's not just being able to explain what MCP is. It's understanding what this shift actually means for client workflows, being able to present the solution properly, and knowing how to spot and explain the risks, like when content shows up inside a client's own systems and it's no longer clear to the end user that it's your content, your brand, or your methodology behind it. That's a genuine literacy gap for a lot of sales teams right now, and closing it looks like a real upskilling programme, not a single briefing.
There is uncertainty about which clients need this
There's real money going into MCP-enabled delivery across the room, but the ROI isn't there yet for most. That doesn't mean the investment is wrong. For some clients, MCP is clearly going to be the optimal way to receive information going forward. And for clients focused on AI sovereignty, wanting to run their own models rather than someone else's, MCP is genuinely critical. It's how they get their own AI working with licensed data without handing over control of the model. For those clients, this isn't optional.
Privacy cuts the other way
MCP leaves a trail. For clients who guard their usage patterns closely, that's a real concern, not a hypothetical one. Public sector clients in particular were flagged as pushing back here. So the same feature that makes MCP powerful for sovereignty-minded clients makes it uncomfortable for privacy-conscious ones. Worth remembering these aren't always the same client.
What's next
We'll be running further sessions on a few things this first meeting only scratched the surface of:
Partnerships and licensing, and specifically where cooperation on AI systems tips into cannibalisation. What clients actually think premium sources are worth inside a partnership, and what end users perceive a product to be once it's aggregating several licensed sources together.
Commercial models and pricing, including how other industries are approaching this.
Protecting premium brands, particularly the risk that partnerships lead users to encounter your content serendipitously elsewhere rather than by coming to you directly, without ever registering it as premium.
The sheer volume of AI-generated content now in circulation, and how businesses are coping with it.
How sales teams need to upskill more broadly when clients are drowning in too much of the wrong information.
And a harder question underneath a lot of this: how do you defend the value of deep research and content work when the end product a client actually sees is one narrow answer, with none of the work behind it visible? If an AI engine only ever surfaces the single data point someone needed, how do you communicate everything that went into making that answer possible?
Finally, go-to-market. Is demand for MCP genuinely being pulled by clients, or pushed by sales teams who've found something new to talk about? Is interest in MCP specifically different from interest in APIs, from different kinds of users? And underneath all of it: when do we actually see ROI on this, or is it simply table stakes now, ROI or not?