The strongest objection to a vertical strategy arrived in August, from the platforms themselves. Google Cloud shipped dedicated Financial Services and Legal editions of its enterprise AI on the same day, each with domain connectors and purpose-built skills, and said Healthcare and Life Sciences are on the horizon. If the platform is now building the verticals, connectors and compliance-aware skills and all, the natural question is whether the vertical moat a partner was counting on just got filled in from above.
Look at what shipped, and at what shipped alongside it. The platform released the vertical product. It did not release the accountability for a regulated outcome, and in the same announcement it named the implementation partners for each edition. That pairing is the whole argument of this final article. The first piece in this series held that platforms build platforms, not professions. The vertical editions do not refute that. They prove it. The platform will build the vertical product and hand the regulated result to a partner willing to stand behind it, because standing behind it is the one thing the platform structurally will not do.
The Platform Will Build the Vertical Product
Be honest about what compresses here, because the pattern is the same one every article in this series has traced. The connectors to the industry's systems of record, the library of domain skills, the vertical packaging: that is real work a partner used to do, and it is moving into the platform SKU. A partner whose differentiation was that it had built the financial-services integrations or the legal document skills is now on the same clock as everyone else, because the platform is shipping those to every partner at once. The technical layer of vertical depth commoditizes exactly the way model access, configuration, and cost governance did in the earlier articles. Defending it is defending a layer that is closing.
It Will Not Answer for the Regulated Outcome
What does not move into the platform SKU is the liability. In a regulated vertical, being wrong is not a bug ticket. It is a failed audit, a regulatory finding, a malpractice exposure, a patient harmed. Someone has to be accountable for the result in a way that has a regulator, a license, or a court attached to it, and that someone signs an attestation, carries the professional liability, and stands in front of the examiner. A horizontal platform will not take that on. Its economics run toward the average customer and the repeatable workload and away from jurisdiction-specific liability, which is precisely why it ships the product and names a partner to deploy it. The accountability is not a layer the platform forgot to build. It is a layer its business model will not let it own.
The evidence that this is where the value concentrates is that AI is landing first in the most regulated work, not the least. The audit itself, the most conservative and heavily governed product in professional services, is being rebuilt on AI. PwC US committed a billion dollars to generative AI in 2023, and PwC Australia is now piloting an AI-native audit platform, built with Microsoft, that examines the whole population of transactions rather than a sample. That investment goes to the regulated core first because that is where being accountable for the result is worth the most. And the customers on the other side are nowhere near ready to own that accountability themselves: in Grant Thornton's 2026 survey of manufacturers, only 12% were confident they could pass an independent AI-governance audit, and only 7% had a tested plan for when an AI system fails, the lowest of any sector. The gap between running AI and being answerable for it is the entire opening.
Regulated Is the Sharpest Case, Not the Only One
It would be a mistake to read this as a retreat into regulated industries and nothing else. Regulation is the sharpest form of the moat, not the whole of it. What defends a partner is being accountable for an outcome the platform will not stand behind, and regulation is simply the case where that accountability is most explicit: a license and a regulator make it unbridgeable. The same defense holds wherever a partner owns the result and keeps what produces it. A revenue number lifted against a control, a defect rate driven down on a specific line, a service level met inside one operation: none of these carries a regulator, yet each is defensible when the partner stands behind the measured outcome, keeps the accountability the platform will not take, and retains the workflow memory and the domain context that no edition ships. The test is not whether the industry is regulated. It is whether the platform has to answer for the result, and whether it can reproduce what the partner holds. Where the answer to both is no, the ground is durable, regulated or not. Regulation simply guarantees the answer stays no.
In Regulated Work, Stop Competing on Price
This is the one place in the series where the cost argument goes quiet, and it is worth saying why. When the deliverable is a regulated outcome, the price of a token is a rounding error against the cost of being wrong. What matters instead is whether the model can actually do the regulated reasoning, and whether the stack underneath it carries the data residency, the certifications, and the controls the vertical demands. That tilts the decision toward capable, compliance-aware, vertically-packaged platforms and away from whichever option is cheapest per token. It is not an argument for any one ecosystem. It is an argument for choosing the stack on capability and compliance rather than cost, on whichever cloud clears the regulatory bar for the work, and for never letting a regulated deal be won or lost on the price of inference. In the earlier articles, the cost curve was the thing to watch. Here, it is the thing to stop watching.
Why This Reaches the SMB-to-Corporate Market
A regional bank, a group of clinics, a mid-sized law or accounting firm carries the same regulatory exposure as a large enterprise and almost none of the internal apparatus to manage it. The large institution has a compliance department, in-house counsel, and increasingly an AI-governance function. The SMB-to-Corporate customer, from small business up through roughly a billion dollars in revenue, has a few people wearing all of those hats at once. It cannot self-serve a regulated outcome off a platform edition no matter how good the edition is, because the product does not come with anyone who will answer for it. That is the deepest version of the argument the whole series has built toward: the customer least able to own the accountability is the one that most needs a partner who will, and in a regulated vertical that need does not soften with the next model release. Regulated markets also move slowly and reward incumbency, so a partner that earns trust in one becomes very hard to replace.
The Common Thread
Eight articles in this series, one piece of ground. Every article came back to the same place from a different direction: the durable value is not the intelligence, it is the accountability for what the intelligence does inside a real business. Regulated verticals are where that is most true and least reversible, because there the accountability is not a figure of speech: it has a license number, a filing deadline, and a regulator who will ask who was responsible. But the moat is the accountability itself, not the regulation, and a partner can own it in any workstream where it stands behind a measured result and keeps the context that produces it. The platforms will keep getting cheaper, more capable, and more vertical, and in time they will ship an edition for every industry there is. They will not stand in the room and answer for the result. That room is the moat, regulated or not, and it is the one place in the whole value chain that was never theirs to take. Own the workflow consequence. Regulation is where it is hardest to take from you, but it was always the thing to own.
Part of a series
This post is one part of the Insights series, our post-by-post working through of the thesis the 2026 Channel Forecast sets out in full.
Sources
- Google Cloud's launch of dedicated Financial Services and Legal editions of Gemini Enterprise, with named implementation partners and a stated pipeline of Healthcare, Life Sciences and further professional-services editions: Google Cloud Press Corner and Google Cloud Blog, August 2026.
- PwC US's $1 billion generative-AI investment: PwC US newsroom, April 2023.
- PwC Australia's AI-native audit platform (built with Microsoft), moving from sampling to full-population analysis, as the clearest case of AI entering the most regulated professional work first: PwC Australia newsroom, October 2025.
- Mid-market AI-governance readiness, including the share of manufacturers confident they could pass an independent AI-governance audit (12%) and the share with a tested AI-incident-response plan (7%): Grant Thornton 2026 AI Impact Survey (manufacturing cut), treated as directional given its advisory-survey source.