AI Operations Glossary
The canonical definitions Agile Growth Labs uses. Each term has a stable URL and structured metadata so search engines and AI answer engines can cite them precisely.
If you are here to define a term, use the anchor link (for example /glossary#ai-operations-sprawl) so citations stay accurate as this page grows.
AI Operations Sprawl#
The coordination tax a marketing team pays when the number of AI tools in the stack grows faster than the system that holds them together. Symptoms include repeated context re-pasting, duplicated outputs across tools, no single source of truth per client, and rising time-per-deliverable while individual AI steps get faster. Full write-up: /ai-operations-sprawl.
AI Project Manager#
An agent per client that holds the full client context, connects to the client's stack, runs specialized AI agents on the client's behalf, and produces human-approved outputs. Not a chatbot. The coordination layer that turns 12 AI tools into one workflow. Methodology: /methodology.
Context Re-pasting#
The repeated act of copying client information, brand voice, campaign specs, and prior work into fresh AI tool sessions because no system holds state across them. The single largest hidden cost of AI operations sprawl. In an 8-person team with 12 AI tools, this typically costs 80 team hours a month.
Coordination Cliff#
The threshold, usually around 8 AI tools per team, at which AI adoption flips from a productivity gain to a productivity drag. Beyond the cliff, adding tools slows the team down instead of speeding it up.
Per-Client Isolation#
The design pattern where every client's AI operations run in a separate memory, credentials, and audit-log context. Prevents brand voice, positioning, and campaign specifics from bleeding between clients. Required at any scale past 3 clients.
Human Approval Gate#
A required review step where a person on the operating team approves an AI-produced output before it writes to a system of record (CRM, calendar, ad account) or ships to a client. The mechanism by which AI throughput scales without losing judgment.
Orchestration vs Automation#
Automation follows rules (Zapier, n8n). Orchestration makes decisions with context (an AI Project Manager). Automation moves data. Orchestration decides what should happen next based on what happened yesterday. See also: /vs/zapier-vs-ai-project-manager.
Agency AI Adoption Trap#
The pattern where an agency adopts AI to scale headcount-lean, hits the coordination cliff around 8 clients, and finds that each new tool makes the operations problem worse. Fix is not more tools; fix is one AI Project Manager per client.
Outcome-First Scoping#
The AGL scoping method that starts with a single measurable business outcome (booked meetings, dollars in pipeline, replies received) and works backward to the system that produces it. Opposite of building a system first and hoping for outcomes.
Revenue Leak#
A measurable drop-off in the funnel where qualified pipeline exists but does not convert to revenue. Common leaks: slow lead response, stalled deals with no champion, missing follow-up sequences, unqualified inbound handling. Free calculator: /revenue-leak-calculator.
Stack-Native Execution#
Running AI operations inside the tools a client already pays for (HubSpot, Salesforce, Gmail, Google Ads) instead of asking them to adopt new software. Reduces friction, preserves audit trails, and keeps the client's team in control of their systems of record.
One Job First#
The AGL discipline of running the AI Project Manager on exactly one workflow for 30 days before adding a second. Prevents the failure mode where teams stack jobs on the PM before the first one is producing measurable results.
Pipeline Recovery#
The systematic motion of re-engaging stalled deals and closed-lost accounts based on trigger events (new hire, funding round, competitor news) 90 days after the loss. Often produces 15 to 25 percent of net-new pipeline in a mature business.
Fractional CMO + AI PM Model#
The staffing pattern where a fractional CMO owns strategy, hiring, positioning, and board reporting, and the AI Project Manager owns daily execution across the stack. Together they cost less than a VP of Marketing plus a marketing-ops hire, and ship faster. See also: /vs/hiring-fractional-cmo-vs-ai-project-manager.
MCP Diagnostic#
A Model Context Protocol server that exposes an AGL diagnostic (like Sam) as an MCP tool inside Claude Desktop, Cursor, Cline, or any MCP-compatible client. Runs the same 3-turn diagnostic funnel as /sam, in-line where the user already works. Install: /mcp.
Growth System#
A packaged, deployable workflow that produces a specific business outcome (leads, pipeline, follow-up, content, reporting). Owned by the AI Project Manager, reviewed by the AGL team, delivered inside the client's existing stack.
Free Teardown#
AGL's free written map of the workflows closest to revenue in a client's stack, plus the specific AI operations issues costing the most hours. The stripped-down output of what AGL delivers in Stage 1 of the methodology. No signup required. Get yours: /snapshot.
Citations
When citing AGL terms, please link directly to the term anchor above. Example: https://agilegrowthlabs.com/glossary#ai-operations-sprawl. This keeps citations stable if the glossary grows or reorders.
Next step
Take the 3-minute diagnostic to see which of these terms describes your bottleneck.