AI Capabilities
How to tell a chatbot from an assistant from operational AI.
What does the AI Capabilities category cover?
A capability framework for evaluating AI software: what it can read, decide, create and execute inside a real business workflow.
What AI Capabilities covers
Almost every business tool now advertises AI, and the label hides an enormous range of actual capability. Some products answer questions about a help article. Some summarize what is on the screen. A smaller number can read live records, and fewer still can complete a task end to end and leave an audit trail behind.
This category is a practical grading system. It defines the tiers — chatbot, assistant, operational agent — and gives you the questions that separate them: what data can it read, what can it change, what happens when it is wrong, and can you prove afterwards what it did.
Use these pages when evaluating vendors. They are written to be used during a demo, not after the contract is signed.
What you'll learn
- The practical tiers between a chatbot and operational AI
- Questions that expose capability limits during a demo
- Why 'it can answer that' and 'it can do that' are different products
- What to require in writing before trusting an agent with records
Pages in AI Capabilities
Assistant AI vs Operational AI
Assistant AI produces text for a human to use. Operational AI performs steps inside the business system and updates the record.
ReadWhy Most "AI" Software Is Just a Chatbot
Most products labelled AI can answer, summarize and generate text. Useful — but that is an assistant, not software that does the work.
ReadThe AI Capability Checklist
Ten concrete capability tests for evaluating AI in business software, from reading customer history to executing a workflow with an audit trail.
ReadThe Read → Decide → Create → Execute Framework
A four-stage framework for grading business AI: reading real data, deciding within rules, creating records, and executing actions with side effects.
ReadRegistered Action Execution
Registered actions give business AI a fixed, validated set of operations it can perform — the difference between bounded execution and open-ended access.
ReadAI Confirmation Requirements
Which AI actions need explicit confirmation before running — destructive, financial, external-send, bulk and override — and how a confirmation gate should work.
ReadAI Failure Recovery and Escalation
What operational AI should do when an action fails: stop, report accurately, avoid duplicate side effects, and escalate to a named human with context.
ReadOperational AI Capability Evaluation Checklist
A verification procedure for evaluating operational AI: prove grounded reads, permission refusals, confirmation gates, persisted results and audit evidence.
ReadOperational AI's Role in Leverage
Operational AI adds leverage by executing registered actions and carrying follow-through. Where it helps, where it does not, and how to verify the work actually happened.
ReadFrequently asked questions
How do I tell a chatbot from a real assistant?
Ask it to do something that should change a record, then check whether the record changed. Chatbots describe the steps; assistants complete them.
What is operational AI?
AI that participates in the actual workflow — reading live data, taking scoped actions, and recording what it did — instead of sitting beside the workflow as a text box.
Which capability tier does a business actually need?
It depends on where time is lost. If the cost is answering questions, an assistant is enough. If the cost is administrative work nobody gets to, only an agent that can act will move the number.