Articles
-
Do AI assistants recommend your product? How to measure AI visibility
Ask the assistants what your buyers would ask, many times over, and count how often you're named and cited. Start with questions about you, then how-to questions your docs should answer, then your category, and add the search engines' free reports.
-
Questions our customers ask about AI agents in production
What a rollback undoes, what happens to a waiting approval, who the audit trail is for, whether you bring your own model keys, and more. The questions our customers ask, each answered with an example.
-
How we write our posts
A coding agent drafts our posts, replies and articles, and backs every claim with our code and docs. Each one is gated: we review it, edit it and say yes.
-
A personal AI assistant or an AI agent your business runs: what changes
A personal AI assistant is safe because its owner watches it. An agent your business runs needs that watching written down, from who approves what to a record of every run and the version that ran.
-
Build vs buy for AI agents: what to build, and what to buy around it
Build what only your agent does. Buy what every agent needs around it, like approvals, an audit trail, evals and rollback. A coding agent can write those too, but owning them is the real cost.
-
Approval fatigue: keeping a human in the loop without a queue nobody reads
Approval fatigue starts when every agent action waits for a person. Match the check to the risk. Reads run, rules you can write down go in code, and a person decides only what needs judgment, seeing exactly what the call would do.
-
When a coding agent builds your AI agent, what should it build on?
Let the coding agent write what's yours, like your tools, your instructions and your rules. Have it build on a stack that already does approvals, the record of every step, versions and rollback, and one it can read as it works.
-
Upgrading the model under a live agent: test it before customers see it
A model change is a release. Make it on purpose, test it on cases your people already judged, roll it out to a few first, and keep rollback one step away.
-
AI agents give different answers each time: how to run them in production anyway
You can't make a model answer the same way every time. You can decide where variation is allowed, fence in the rest, and see what varied.
-
Workflow automation or an AI agent: which one fits the job
Use workflow automation where the steps and their order are the rule. Use an AI agent where a step needs judgment. Most real processes need both, in one flow.