Growth Systems Engineering
Artificial Intelligence
AI is a layer of integration, not a strategy by itself. We build it into business systems the way we build everything: where it demonstrably serves the objective, on reliable data, with human judgment designed in.
Every few months the tools change names. The discipline underneath them does not. AI has crossed from novelty to useful instrument (for classification, extraction, drafting, decision support, and increasingly for agents that carry out multi-step work) and it now touches the systems we engineer often enough to deserve its own pages. Not because AI is everything, but because it is a layer of integration your business will be measured against.
Where AI sits in a systems practice
We are a systems engineering firm, not an AI company. That distinction is the value. An AI feature bolted onto disconnected systems inherits every one of their problems at machine speed: wrong records, stale data, duplicated identities, confidently automated. Built on the integrated, reliable infrastructure the rest of our practice creates, the same instrument compounds: cleaner inputs, supervised outputs, and automation that knows when to hand a case to a person.
The three applications we build
- AI in business automation
- Classification, extraction, drafting, and decision support wired into real workflows: the work described on AI in Business Automation. This is where AI most reliably earns its keep today.
- AI agents, supervised
- Multi-step work delegated to software with scoped tools, approval steps, and an audit trail: how that is engineered, and how MCP fits, on AI Agents & MCP.
- Agent-ready web presence
- Your customers are beginning to send AI assistants to websites on their behalf. Agent-Ready Websites & WebMCP covers how a site answers those agents with structure instead of leaving them to guess.
What we will not do
The position we published when this practice was one paragraph still holds: we will not rebrand the firm around this year’s model, and we will not hand an unsupervised process to a system that cannot explain itself. We add three more refusals here. We will not automate a process nobody understands. We will not point AI at data nobody trusts. And we will not remove the human from a judgment the business would regret delegating: money, medicine, and reputation stay reviewed.
The engineering that makes all of this dependable has its own page: Deterministic Systems. It is the argument underneath every one of these applications, which is that the model is a component and the system around it is the product.
Where to start
The same place every engagement starts: a systems assessment. AI questions are systems questions wearing a new name: what should happen, what data decides it, and who is accountable when it runs. If the assessment says the honest answer is “no AI yet, fix the data first,” that is the answer you will get. The engineering discipline behind all of this lives on Automation & Intelligent Systems.