Glossary

AI work and services terms, defined.

The concepts behind autonomous agents, plus the operational language every services business runs on.

The Helm glossary defines the AI platform concepts that underlie autonomous agents, plus the financial and operational terms every professional services business runs on. Each entry explains the term on its own and how it shows up in Helm specifically.

Get Early Access See the platform

AI work platform concepts

How Helm works.

Agents, autonomy, memory, and the platform concepts that make autonomous work possible.

Agent memory is the persistent context an AI agent retains across conversations and runs. Unlike a stateless chat, an agent with memory remembers past decisions, client preferences, project context, and operational patterns. It picks up where it left off instead of re-briefing every session.

An agent role is a pre-defined job function that shapes how an AI agent behaves: its responsibilities, default instructions, and typical tool permissions. Examples include Project Manager, Account Manager, Payment Recovery Specialist, and Sales Development Rep. Roles are starting points. Each agent created from a role can be customized further.

An AI agent is an autonomous software worker that executes tasks on behalf of a person or team. Unlike an AI assistant that responds to prompts, an agent reads context, makes decisions, uses tools, and completes work end-to-end. It sends emails, updates records, chases invoices, or posts status updates without requiring step-by-step human direction.

AI work platform

An AI work platform is a unified software system where autonomous AI agents execute business work alongside human teams. Unlike AI features bolted onto existing tools, an AI work platform is built with agents as a first-class primitive. The same data, permissions, and workflows serve both humans and agents without integrations or sync layers between them.

Autonomy levels

Autonomy levels are discrete modes that control how independently an AI agent acts. Helm supports three: Observe (the agent reads and reports but makes no changes), Suggest (the agent proposes actions you approve or reject), and Auto (the agent executes directly within its permissions). Each agent's autonomy is configured per workspace, per role.

Client intelligence

Client intelligence is the compounding knowledge layer a services business builds about its clients over time: preferences, decisions, context, history. It is captured automatically from every interaction, project, and invoice. Client relationships become an institutional asset that survives staff turnover and deepens platform switching costs.

MCP-native describes a platform built with Model Context Protocol (MCP) as a first-class interface. MCP is the emerging open standard, introduced by Anthropic and adopted across Claude Desktop, Claude Code, Cursor, and other AI tools, that lets AI agents interact with external systems. An MCP-native platform exposes its data and actions through MCP, so any MCP-aware AI tool can operate on it.

Per-workspace pricing

Per-workspace pricing is a SaaS pricing model where customers pay a fixed price for the entire workspace regardless of how many team members use it. Unlike per-seat pricing, the price does not scale with headcount. A 3-person team pays the same as a 30-person team on the same plan.

Services business fundamentals

The operational language.

Utilization, realization, DSO, retainers: the metrics and levers every services business runs on.

Agency OS refers to the unified software platform a services business runs on. It combines CRM, project management, time tracking, invoicing, documents, and reporting into one system rather than a federation of point tools. The term frames the choice as picking an operating system for the business, not picking individual apps.

Billable hours are time entries that can be invoiced to a client. The distinction separates client-funded work (strategy, design, implementation, meetings with the client) from work that supports the business but is not charged (internal admin, team meetings, training, business development).

A change order is a formal amendment to an existing engagement that documents and prices new scope. When a client requests work beyond what was originally agreed, a change order captures the additional deliverables, timeline impact, and revised cost. Both parties sign before the new work begins. Change orders protect margin by making scope creep an explicit, priced conversation.

Chargeability is the proportion of a team member's time that is billable, expressed as a percentage of available working hours. Functionally similar to utilization rate, with slight nuance: chargeability often refers to an individual's capacity allocation in capacity-planning contexts, while utilization is more commonly reported as a backward-looking operational metric.

A client portal is a scoped, client-facing view of a services business's platform. It is typically branded, permissioned, and limited to what the client needs to see: their active projects, invoices, shared documents, and ways to request new work. Client portals reduce routine inbox volume and let clients self-serve for common needs.

Days Sales Outstanding (DSO)

Days Sales Outstanding (DSO) is the average number of days between issuing an invoice and receiving payment. A company with $100K in receivables and $10K in daily revenue has a 10-day DSO. For services businesses, DSO is the clearest single measure of cash-flow health. Lower is better. Sustained increases are an early warning of collection risk.

Dunning is the process of communicating with clients about overdue payments. It's a sequence of reminders, follow-ups, and escalations that typically starts friendly (3 days past due) and escalates over time (14, 30, 60 days). Good dunning keeps payment expectations clear without damaging the relationship. Bad dunning either loses receivables or alienates clients.

Pipeline forecast

A pipeline forecast projects expected revenue by weighing each open deal by its stage probability. A $50K deal at 40% probability contributes $20K to the weighted forecast. Accurate forecasts require disciplined deal hygiene: probabilities that reflect actual likelihood, stages that move with real signals, and stale deals that get removed rather than massaged.

Realization rate

Realization rate is the percentage of logged billable hours that actually appear on an invoice. If a consultant logs 100 hours at $200/hr but only 80 hours are billed (because 20 were written down, absorbed in fixed-fee work, or unable to be justified to the client), realization is 80%. It measures billing discipline, not work effort.

A retainer is a recurring commercial agreement where a client pays a fixed fee at a regular interval (monthly or quarterly) for a defined scope of ongoing work: a block of hours, a set of deliverables, or access to the team at an agreed priority. Retainers provide predictable revenue to the provider and predictable access to the client.

Scope creep is the gradual expansion of project scope beyond what was originally agreed. It's unbudgeted, often undocumented, and almost always at the provider's expense. Small individual requests compound over the life of an engagement until the team is delivering materially more than was priced into the SOW.

Statement of Work (SOW)

A Statement of Work (SOW) is the document that defines what will be delivered, by when, for how much, and under what terms. It lists specific deliverables, milestones, assumptions, exclusions, and acceptance criteria. The SOW is the contractual source of truth for "what's in scope." Everything outside it is a change order conversation.

Time-to-invoice

Time-to-invoice is the number of days between when work is delivered (or a billing trigger is met) and when the invoice actually leaves the building to the client. Short time-to-invoice accelerates cash flow and signals operational discipline; long time-to-invoice is the most common hidden drag on services cash flow.

Utilization rate

Utilization rate is the percentage of a team member's available working hours that are spent on billable client work. Calculated as billable hours divided by total working hours. A 40-hour week with 30 billable hours is 75% utilization. Most healthy services businesses target 65-80% utilization for delivery staff.

Work Breakdown Structure (WBS)

A Work Breakdown Structure (WBS) decomposes a project into smaller, manageable deliverables organized hierarchically. Top level is the whole project; next levels are major phases or deliverables; leaf level is individual tasks. A WBS makes scope estimable, assignable, and trackable, and exposes gaps where undefined work was smuggled into "and then we finish."

See Helm in practice.

All 12 apps, AI agents, and per-workspace pricing. The terms defined here are the system you operate.

Get Early Access See the platform