Calyx has run inside a working regulated accounting firm since September 2025 — through a full filing season, a year-end close, a client dispute, and an insurance renewal. Plenty of people have opinions about governing AI in a regulated firm. Very few are operating it inside one.
The efficiency numbers are real, and they are not the point. The firm didn't get faster at what it already did. It started doing work it had never sold before — entity structure analysis, transaction advisory, ledger reconstruction, multi-entity planning — because the capacity to do that work stopped being a hiring problem.
That's a change in what business the firm is in. Not an efficiency gain.
Advisory revenue carries different margins than compliance revenue, which is why the third of those numbers moved at the same time as the first. Three new service lines were built and sold inside the engagement: a business situational awareness assessment, a personal one, and transaction advisory.
In September 2025 a long-established regulated accounting firm lost its key internal operator and two part-time staff simultaneously, with active client work in flight and filing deadlines fixed. What it needed was not a technology vendor. It needed someone to run the operation while the operation kept running.
Calyx deployed its platform and put people inside the firm — an architect and an operator, working the actual queue. That is the difference between a system that demos well and one that survives a real tax season, a real client dispute, and a real insurance renewal.
Emergency continuity during active filing pressure, then workflow restoration. Nothing was modernized until the work stopped being at risk.
Infrastructure modernization and governed platform operations, with client delivery running throughout rather than paused for a transition.
Four governed verticals plus the evidence layer, running in production across multiple client engagements with real audit, compliance, and underwriting exposure.
Scenario modeling, entity structure analysis, ledger reconstruction, transaction classification, and financial advisory deliverables.
Engagement letter development, scope definition, and liability limitation language across all return types.
Cyber and professional liability coordination, with AI governance language taken into underwriting.
The evidence layer. Provenance, audit trail, and integrity — every analytical output traceable and defensible.
The internal surface the operator actually worked from, integrating the verticals into one governed tool.
People placed inside the client environment, running the actual queue. Not advice, not software, not consultation — an integrated operating team that also proved the platform under live conditions.
Most AI vendors have never put their governance framework in front of a carrier. This one cleared commercial underwriting without exceptions, which is a material validation event and not a marketing claim.
New premium service lines created without new headcount. Advisory work became structured, repeatable, and profitable — which is what changed the revenue number.
The frameworks are in use rather than filed. That distinction is the difference between governance that is documented and governance that is provable.
Most firms don't need a deployment first. They need to know what AI is already touching the work, who is accountable for it, and what evidence exists that anyone reviewed it. The Situational Awareness Assessment maps that before any deployment decision — fixed scope, executive-ready findings.