Enterprise ITSM + ESM,
with the AI Workforce to run it

A credit is Atomicwork's unit of AI workforce consumption. Every time an AI co-worker completes a job — answering a question, processing a request — it draws down credits from your plan.

Credits are designed so simple work draws lightly and heavy work draws more.

Six factors drive how many credits a single AI co-worker execution consumes:

‍

1. AI Coworker complexity

A simple Q&A Coworker answering from your knowledge base does one job in one step. A vulnerability response co-worker investigates, retrieves data, calls tools, and writes a remediation plan — many steps, more credits.

‍

2. Number of steps

Each reasoning step the AI Coworker takes costs credits. A single-turn response is cheaper than a multi-turn investigation, which is cheaper than an end-to-end automation chain.

‍

3. Tool and API calls

When a Coworker reaches into another system — Slack, Okta, your ITSM platform, a custom CMDB — those tool invocations consume credits on top of the LLM work.

‍

4. Response length

Long-form output (a written report, a detailed answer) consumes more credits than a one-line acknowledgement. Output is the most expensive token type, so output length matters most.

‍

5. Multi-Coworker handoff

When two Coworkers chain on a single ticket (e.g., Onboarding Coworker hands off to App Provisioning co-worker), each Coworker execution counts separately.

‍

6. Model selection

Different jobs run on different models. A lightweight model handles simple routing fast and cheap. An advanced reasoning model runs deeper analysis — like reviewing your AWS dashboard for misconfigurations — at a higher credit weight. Atomicwork routes work to the right model automatically based on the Coworker's configuration and the task's complexity.

‍

Atomicwork credits map to four broad categories of work. Within each, the actual credits a single execution consumes depend on prompt length, conversation history, the model used, and how many tools fire. The examples below show typical executions at each end of the range. These credits are just estimations and not indicative of a maximum number.

‍

Knowledge / Q&A: 1-10

Simple knowledge query: 5

A user asks "What's our remote work policy?" The Coworker retrieves the answer from your KB and replies in one turn.

‍‍

Multi-turn with image analysis: 10

‍User uploads a screenshot of an error message. The Coworker analyses the image, identifies the issue, and walks them through the resolution across two or three turns.

‍

Access management AI: 5-25

Single-app access request: 8

‍A user requests access to Tableau. The Coworker validates their role, routes for manager approval, provisions the app, and notifies the requester.

‍‍

Multi-system access for a new project: 20

‍A user requests access to AWS staging, an S3 bucket, Tableau, and a new Slack channel. Multi-step: role validation, multi-approver chain, provisioning across systems, notifications.

‍

Workflows: 1-50

Deterministic ticket routing: 1

‍A pre-defined workflow classifies an incoming ticket by category and routes it to the right team. No AI reasoning needed — just rule-based logic.

‍‍

Sentiment-classified response: 20

A customer replies on an open ticket. The reply is classified by sentiment. If negative: escalate to a human and draft an empathetic response. If positive: close the ticket and trigger a CSAT survey.

‍

AI workforce: 20-100

Status Update Agent: 20

‍Pulls metrics from Jira, GitHub, and project tools each week. Drafts a status update for the project owner to review and send.

‍‍

Network Ops Agent: 50

‍Investigates a connectivity issue: checks logs, reviews network topology, looks at recent changes. Identifies likely root cause, drafts remediation steps, opens a ticket if needed.

Atomicwork offers three outcome-based tiers: Knowledge Support ($1/outcome), Access Automation ($2/outcome), and Service Acceleration ($3+/outcome). An "outcome" maps to a specific, completed AI-driven action: for example, answering a policy question, provisioning a software license, or resolving an IT incident end-to-end. A Knowledge Support outcome is a resolved knowledge query — answered from your internal docs, KB, or policies. An Access Automation outcome is one access request handled end to end — manager approval, provisioning, or deprovisioning. A Service Acceleration outcome is an AI Coworker achieving its defined goal on a ticket — NetworkOps, CloudOps, DeviceOps, Incident Management.

Yes. The Enterprise plan includes custom data residency options and advanced compliance and security features. Atomicwork holds SOC 2, GDPR, HIPAA, CSA STAR Level 1, ISO 42001, CASA, and Microsoft 365 certifications. That's why our customers in regulated industries like financial services, healthcare, and insurance trust us for their IT requirements.

Atomicwork is positioned as a holistic, agentic service management platform, not just ITSM. The Business and Enterprise plans include custom workspaces for every department, enabling HR, Finance, Legal, and other teams to run their own service operations. The AI Coworkers are pre-trained on internal docs, knowledge bases, runbooks, and policies, meaning they can answer domain-specific questions across functions, not just IT tickets.

Atomicwork natively integrates with popular IT and HR tools including SharePoint, Microsoft Teams, Azure Active Directory, Intune, Okta, ADP, Workday, and Slack as well as the broader Microsoft 365 ecosystem. It also integrates with other ITSM platforms (ServiceNow, Jira Service Management) for the "Bring Your Own Platform" use case. You can view all our supported integrations here.

Yes, Atomicwork offers a native MCP (Model Context Protocol) server that lets any MCP-compatible AI assistant connect directly to your Atomicwork tenant. That means tools your engineering and IT teams are already using — Claude, OpenAI Codex, Cursor, VS Code Copilot — can query tickets, create incidents, manage service requests, look up users, and browse your service catalog without ever leaving their workflow.