What does Formulation Labs actually do?

Formulation Labs designs, builds, and governs better systems for acquisition, follow-up, reporting, and operations.

Where do engagements usually start?

Most engagements start with a Systems Audit so the real leverage and sequence of work become obvious before implementation begins.

What is the difference between Systems Build and Systems Oversight?

Systems Build installs the approved operating layer. Systems Oversight governs and improves that layer over time so it does not drift.

AI & automation

AI operations and agent infrastructure

We built the tooling that lets an AI assistant safely run a company's advertising, email, and analytics across all of its platforms, instead of a person logging into each one by hand.

Six servers and more than 300 governed tools that let an AI assistant operate a company's advertising, email, call tracking, and analytics directly, instead of a person clicking through six consoles.

Context

The growth team managed advertising, email, call tracking, and analytics for a multi-location operator across half a dozen platforms, each with its own console and its own busywork. A routine change meant logging into six tools and clicking through the same steps by hand. They wanted to drive the whole operation through an AI assistant instead, but an assistant is only as capable as the tools it can safely call. We built that tool layer.

Constraints

Six platforms with completely different interfaces had to be exposed through one consistent, machine-callable surface.

An AI assistant cannot be handed raw credentials or unchecked write access, so every operation needed a typed, validated boundary.

Coverage had to be broad enough to actually run campaigns, not just read reports, which meant a large and carefully structured tool surface.

Credentials for six live advertising and analytics accounts had to stay out of the codebase entirely.

Work performed

Built six servers that expose Google Ads, Meta, Mailchimp, CallRail, GA4, and Monday to an AI assistant through one consistent interface.

Implemented more than 300 individual tools across those servers, the bulk of them in the two advertising servers, covering live campaign, budget, audience, and creative operations.

Enforced a typed, validated boundary on every tool, so the assistant operates inside guardrails rather than against raw account access.

Kept every credential out of the codebase, loaded at runtime from a separate configuration layer.

Outcomes

300+Governed AI tools

More than three hundred validated tools give an assistant real operating control, not just read access, over the company's marketing stack.

6Platforms, one interface

Google Ads, Meta, Mailchimp, CallRail, GA4, and Monday all answer to one consistent, machine-callable interface instead of six separate consoles.

0Hand-built API glue

Campaign, budget, email, and analytics work runs through one governed layer, rather than a drawer of one-off scripts maintained by hand.

← Back to work Apply