The words we use, without the hand-waving.

A short working glossary — what each term means in our practice, not in a vendor brochure.

Term definitions

Digital FTE
An AI system that does a real job — not just answers questions — under human oversight. HR uses FTE for one employee's worth of work; a Digital FTE is software that carries that same unit of work, around the clock.
AI Worker
The same thing as a Digital FTE. The terms are interchangeable in our writing.
AI-Native company
A company whose operating model assumes AI workers do most of the execution — with people setting direction and judging outcomes. Removing the AI stops the process, the way removing employees stops a traditional one.
Forward Deployed Engineer
An engineer embedded with your team to ship AI systems inside your real process — reading your work, writing the spec, building the worker, deploying it, and leaving your team able to run it.
MCP (Model Context Protocol)
The open standard for connecting AI agents to external tools and data — repositories, ticketing systems, databases, browsers — through typed connectors the agent can call.
SKILL.md
A plain-text file that captures what an agent must know and how it must behave for a domain — the capability written down once so no run starts from zero. Skills port across coding agents.
Spec-driven development
Writing down the work, the quality bar, the constraints and the failure cases before building — then holding both humans and AI to that document. Wrong output means fix the spec, not the symptoms.
Human gate
A written decision point where the worker stops and a named human approves before any irreversible action — payments, data deletion, customer-facing messages, production changes.
Loop Engineering
Designing systems that run the work on their own: heartbeat, worktree, skill, subagents, connector, spine — plus a human gate. The discipline of building loops, not just prompting once.