The governance layer for enterprise Claude adoption.
The Firm
When a Claude deployment stops being an experiment and starts being infrastructure — when agents move money, sign external communications, and touch customer data — the question is no longer whether the model works. It is who approves what, audited how, and reversible by whom. Keystone Meridian is a consulting practice built for that question. We are not a generalist AI consultancy. Our role in an organization is narrow and explicit: we are the piece that holds it together — the keystone at the apex of the arch, the meridian line that serves as reference.
What We Do
Sequenced adoption plans for moving an organization from first-use to production Claude deployment. Each phase carries a defined gating criterion — what the business must observe, not just build, before the next expansion runs. The deliverable is a signed plan a CIO can defend to risk and audit, not a maturity-model diagram.
Approval-gated operating models for AI-driven work: risk tiering, human approval checkpoints, audit surfaces, and the separation of planning from action. Delivered as an operational document the internal teams will actually run against — not a slide deck that dies after the readout.
Independent review of live Claude-enabled systems: safety boundaries, observable behavior against intended behavior, audit-trail completeness, credential and data-flow review. Findings land as a written report with severity-tiered remediations routed to the owners who can act on them, not as an advisory note addressed to no one.
A written assessment of where Claude adoption will create durable advantage for a specific organization — and where it will not. The output is a capability map with honest limits, scoped to the business the organization actually runs. A decision document, not a sales artifact.
Hands-on enablement for the internal teams who will own AI-assisted work after we leave. Materials are written for the operators on the floor — the people running approvals, reading the audit record, and living with the system week to week — not for the executive who read the deck.
The Discipline
01
Agent actions that could bind an organization — money movement, external communications, protected operations — run through explicit approval gates. No binding action ships without a human-approved plan on file, named to the person who approved it. The cost is visible: engagements move at the speed of a human review loop, not at the speed of the model.
02
Every agent operation produces an auditable record: what was attempted, which decisions were made, which tools were called, what outcomes followed. The record is written to the client's internal audit function at delivery — not held behind a vendor portal, not summarized for a deck. The firm works on the assumption that someone on the inside will read the raw record, and writes accordingly.
03
Plan first: produce the plan, review the plan, approve the plan, then execute. No silent re-scoping mid-operation. No destructive action without a sanctioned plan behind it. When the plan is wrong, the engagement stops and re-plans — the firm does not improvise at the edge, and does not advance the plan through the pause.
The Practice
At a certain scale, a Claude deployment stops being a set of features and starts being part of the operating structure of the business. Agent systems begin to move money, sign communications, touch customer data, and operate on production infrastructure. The question is no longer whether the model works — it is who approves what, audited how, and reversible by whom. Keystone Meridian exists for that moment. We work with a small number of organizations at a time because depth of engagement is not negotiable.
Founder
John York
Founder
Keystone Meridian exists because I think the most consequential infrastructure being built right now — the agent systems that organizations are putting in front of their money, their customer data, and their public communications — deserves the same operating discipline that mission-critical hardware has lived under for decades.
I built the firm after two decades in electronics manufacturing and supply-chain program management, where shipping the wrong firmware or missing a component tolerance was never theoretical. It was a line-down event, and it was answered for the same day. That work shapes how I think about every Claude deployment we touch: the question is never whether a system can be built, it is what happens when it is in production and something has to be answered for.
The organizations I work with are past the experimental phase. Their Claude adoptions are beginning to move money, sign external communications, and operate on infrastructure the business runs on. At that altitude, the work the firm does is not advisory; it is structural. Approval gates, audit records, separation of planning from action — those become the operating model, not a layer on top of it.
I work with each engagement directly. The firm does not scale by adding accounts; it scales by deepening the work it does for the organizations already on the inside.
We work with a small number of organizations at a time. The right scoping starts with a named person on your side who can describe the adoption you have now, and the posture you need next.
Engage