Operator Engagement · Regulated Accounting Firm

Not a pilot.A year of operating.

Since September 2025, Calyx's founder has run operations inside a working regulated accounting firm, 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.

EngagementSept 2025 — ongoing
IndustryRegulated accounting & tax
ModelEmbedded operator
Firm nameHeld confidential

The firm's results stay with the firm. Its figures, and the evidence behind them, are shared under NDA and only with the firm's written consent.

Calyx Intelligence LLC was formed in May 2026. The work before then was done by its founder, Michael Lawrence, and it continues under Calyx.

What actually changed

A compliance firm now does advisory work.

The firm didn't just 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.

Three new service lines were built and sold inside the engagement: a business situational awareness assessment, a personal one, and transaction advisory.

How it started

The firm lost its operator in the middle of filing season.

In September 2025 a long-established regulated accounting firm lost its internal operator 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.

The founder stepped in as operator and architect, working the actual queue and building the governed tools around it. That is the difference between a system that demos well and one that holds up through a real tax season, a real client dispute and a real insurance renewal.

Stabilize first

Continuity during active filing pressure, then workflow restoration. Nothing was modernized until the work stopped being at risk.

Then rebuild

Infrastructure modernization and governed operations, with client delivery running throughout rather than paused for a transition.

What was deployed

Under live regulated conditions. Not a sandbox.

Governed tools plus an evidence layer, in daily use on real client work.

Numera

Scenario modeling, entity structure analysis, ledger reconstruction, transaction classification and advisory deliverables, each reviewed before release.

Juris

The firm's own engagement letter and scope templates, including AI-use disclosure, for the firm's review and use.

Insura

Coordination of the firm's cyber and professional liability renewal, including documentation of how AI is used in the work.

LedgerGuard

The evidence layer: where each output came from, who reviewed it, and when.

Operator workbench

The internal surface the operator actually worked from, bringing the tools into one governed place.

What the year produced

Work the firm can show.

What it shows

The things you can't learn from a demo.

The operator model works

Someone inside the client environment, running the actual queue. Not advice, not software alone — operations that also proved the tools under live conditions.

Every output has an owner

Each analytical output carries its source and a named reviewer, so the firm can show who checked what, and when.

Capacity becomes new services

Advisory work became structured and repeatable, and the firm sells it.

Compliance becomes operational

The frameworks are in use rather than filed. That is the difference between governance that is documented and governance that can be shown.

Start with where you actually stand.

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.