Why AI Approvals Need More Than a Yes Button
An approval button can make a workflow look controlled without making the approved action clear. The agent presents a summary….
-
-
How to Roll Out Enterprise AI Without Handing It the Entire Business
Going live should not mean granting the system every capability the team eventually wants it to have. I think an…
-
The Role of Deterministic Controls in Agentic AI
Some parts of an AI system should not need interpretation. If an agent is permitted to change one customer's record,…
-
Why Your AI Agent Needs a Stop Condition
An agent that keeps working is not necessarily making progress. It may be investigating a difficult problem. It may also…
-
How to Measure the Real Cost of an AI Workflow
A cheap model can sit inside an expensive operation. I think that is one of the most useful things to…
-
Why AI Agents Should Not Treat Every Document as an Instruction
AI prompt injection is a useful reminder that reading information and accepting instructions are different responsibilities. An agent may need…
-
The Architecture Behind Reliable AI Knowledge Retrieval
AI knowledge retrieval can produce an answer that looks well supported while the system has still selected the wrong information….
-
When One AI Agent Is Better Than a Multi-Agent System
A single AI agent can be the right architecture even when the technology makes a larger agent network possible. I…
-
Why AI Observability Needs to Follow the Entire Workflow
AI observability can give a company an impressive amount of information about its models while leaving it surprisingly uncertain about…
-
How to Choose the First Enterprise AI Workflow Worth Building
AI workflow selection is often treated as a search for the most impressive thing the technology can do. I think…