The Two Positions That Create a Durable Moat
Over the past few weeks, there has been a lot written around vertical AI defensibility. A bunch of great posts, such as this one, this one, and this one, have circled the same core idea from different angles: scar tissue from operating inside actual systems, process embodiment, deep integration, regulatory complexity, liability absorption.
Most of these arguments distill down to a question of where you sit in the value chain, what you control once the model itself is no longer the differentiator.
I think there are two positions that matter more than everything else for durable defensibility. First: owning the moment data is created, something I call the mint position. This means not just reading what’s already in the system of record, but being in the flow when an action turns into structured data. Second: owning the accountability position, which means being responsible if/when things go wrong. This forces you into the systems and relationships that create the record, the audit trail, and the outcome itself.
In a world where models get cheaper and better, these are the two places where moats still form.
Owning the Mint Position
Sitting on top of company data and/or being the system of record are important, but will not alone create a durable moat. You need to take it one step further and own what I call the mint position: being the system that stamps work into official records at the moment it happens. It’s the difference between being where the data gets logged versus actually defining how, when and where data is created. A vault stores value, a ledger records it, but a mint creates it.
Sometimes hardware matters because it makes it easier to capture the moment right where the work happens. Sometimes owning the transaction matters because it makes you the layer that records the actual state change: the thing that moves money, inventory, liability, or authorization through the system. Either way, the key is occupying the mint position so you capture the data as it’s born, not after the fact.
For example, a lot of warehouse software has the data because it sits on top of the Warehouse Management System (WMS) and pulls inventory records. But the inventory record is often wrong because the real truth gets created in the receiving aisle: partial shipments, swapped SKUs, damaged cartons, mislabeled lots. Owning the mint position means capturing the discrepancy the moment it happens via a photo, video, voice note, which then writes the state change into the WMS immediately. Quarantine this lot, split the receipt, adjust inventory, trigger a supplier chargeback or claim. You’re creating the inventory truth and the transaction trail the business runs on.
Once you own the mint position, three things happen:
You’re now generating the labeled ground truth data.
That ground truth data becomes the way teams make decisions and report on those decisions.
Switching costs become higher the more decisions and reports you enable.
Owning the Accountability Position
Owning the accountability position means that if something goes wrong, you are responsible (even if in a bounded way). This is similar to outcome-based pricing, but takes it a step further because you don’t just price on outcomes, you underwrite them. You stand behind a specific promise (accuracy, timeliness, compliance, recovery, reconciliation), and so if/when something breaks, you’re the party that has to fix it, pay for it, and prove what happened. That forces you into the transaction path and makes your system the authoritative record.
Regulatory environments are inherently defensible, and many of the recent essays are right to emphasize it. In this sense, it’s about becoming the compliance surface presented to institutions. It means your records are the ones regulators and auditors accept as the authoritative trail.
Freight forwarding and customs is a good example of this. The entries and supporting documentation you assemble and file are what enable (or delay) clearance: classification, valuation, origin, PGA requirements, and the full document packet. It’s much harder to replace the system that generates and maintains that audit-ready record than it is to build an assistant that drafts notes an actual broker still has to translate into a compliant filing.
But you don’t need a regulator in the loop for this to matter. Outside of regulated industries, you can create the same kind of defensibility by absorbing the downside on a defined slice. When you’re the party that guarantees resolution, you naturally then become the system that creates the record, the transaction, and the audit trail.
Position > Model
Foundation models will only keep improving. Workflows are visible, which means they can be reverse engineered. Deployed engineering is a service line that is sometimes necessary, but never sufficient to create durable moats.
If your product sits on top of someone else’s database (reading from it, writing back into it), you are, by definition, not the canonical layer. You can be valuable. You can even become sticky for a while. But you’re still displaceable the moment a better model or a cheaper competitor or incumbent offers “good enough” workflow assistance.
The companies I’m most excited about own where and how the data gets created, and they own the downside. They’re building the record. And over time, they’re shaping the ontology the industry uses to describe itself.
Author’s note: An LLM was used for light copy editing only (spelling, grammar, and clarity). Content, meaning, tone, and structure remain unchanged.