MCPs and the finance moat reset

And why the role thought most exposed to AI may be the one that makes the whole stack work

MCPs everywhere

The Model Context Protocol (MCP) frenzy is alive and kicking in the world of finance and accounting platforms, it’s almost as if offering MCP access to your platform is table stakes in the industry. This applies from the industry incumbents of NetSuite and Intuit’s Quickbooks all the way through to tech-forward platforms such as Aleph, Numeric, Pigment, Rillet, and Tabs. This has all happened in the last few months, following on the heels of the Claude and agentic AI releases in late 2025.

Anyone writing about MCPs feels the need to include a definition—Anthropic created the standard so I’ll just reference their text, “It provides a universal, open standard for connecting AI systems with data sources, replacing fragmented integrations with a single protocol. The result is a simpler, more reliable way to give AI systems access to the data they need.” Your AI (powered currently by LLMs) can talk directly with external platforms, it can understand how the data is organized and what it means, and it can interact with the data using functions provided by the platform.

Might it be that the world up till now, where platforms delivered primarily a visual interface to interact with their data model and workflows is on its way out? It does seem that with the uptake of AI and agents that we are moving toward an empowered user experience, able to architect your own workflows and ultimately natural language interaction with data storage and compute environments.

I’ve even seen commentary on how command line interfaces (CLIs) are re-emerging as a work medium for non-software developers. Of course, chat interfaces, common in AI since the first ChatGPT release, are in a way light-CLI environments too.

New world of substrate access

Sivulka says in his article In Defense of Vertical Software “The value of enterprise software comes from understanding the process and the organization well enough to make the software do exactly the right thing.” and continues on to call this “last mile” the critical component, which I think is a great way to describe the value.

Vertical software capitalized on owning the information delivery and control mechanisms in products, enabling sticky relationships with users and pricing models to match. A different world is emerging, where custom interface and workflow design is almost entirely within reach of the user. Driving this new world are foundational AI models, constantly increasing in capability and usability, that have given users the ability to create one-off and repeatable solutions, do work more efficiently and agentic frameworks in which autonomous agents, provided with scope and direction, can work alone or together in teams to tackle much larger scale tasks.

The trend we’ve seen as a result of this is finance and accounting platforms providing access at the core substrate level, so that the underlying utility of the platform is exposed and more transparent. Vendors may not yet have dispensed with front-end UIs but they have provided the option to bypass UIs for those teams that are sufficiently skilled to use these new developer paradigms.

I feel this does lead to the question though: What characteristics will successful platforms in this new world embody?

Two positions that will survive the trend

I think Gamble in her article The Two Positions that Create a Durable Moat captures two key concepts that will help differentiate success from failure in the new platform wars; the mint position, where the platform owns the data or context at the time an event happens, and the accountability position, where the platform is responsible whenever something goes wrong.

With mint Gamble says it clearly as “the vault stores, the ledger records, the mint creates”—at the very instant a transaction occurs or knowledge or other value is built your platform captures the context. And with accountability it’s the party that underwrites the outcome and absorbs the downside when things fail. These I think will exist outside of any foundation model or environment in which a user customizes their experience and workflow.

I think the concurrent trends in AI right now are strengthening the case that these two moats or value positions will define which platforms secure longevity in the future. With the mint position, the external user or system interaction generates the ground-truth record—foundation models will interact with systems that store that record. The accountability position is closely related to this growing sense of outcomes-based commitment in products and services. You need the vertical integration to achieve a service-level guarantee that requires significant expertise and understanding of the task at hand but built upon a solid, dependable and auditable data substrate.

Accountability and the knowledge worker

The future operating paradigm could be a merger of data platforms that expose base-level constructs along with functions to operate on and transform that data (back to our starting topic of MCP), with general utility AI and with application expertise sufficient to provide a service-level guarantee. Such a guarantee may, for example in accounting, be auditability and/or accuracy.

Interestingly, the very resource thought to be at most risk of replacement with AI may instead be the most fundamentally valuable resource that could make the whole stack work: The knowledge-based worker, who has an intricate understanding of what a service-level guarantee represents and how to coordinate the technology to deliver. Of course, the guarantee is only as good as the person who understands what it promises. As Andrej Karpathy highlighted recently in the context of AI agents “you can outsource your thinking, but you cannot outsource your understanding.”