The OpenTag blog

Notes on making AI useful at work.

Product thinking, field notes, and engineering choices from the team building an AI coworker that does the work—not just talks about it.

Written by the OpenTag team

From the team

Latest notes

Practical ideas for handing real work to AI while keeping context, control, and accountability intact.

Product5 min read

The best AI interface is already open

Work does not begin in a blank prompt. It begins in a thread, surrounded by the people, decisions, and tools that give it meaning.

Read article
Trust6 min read

AI work needs a handbrake, not a disclaimer

Confidence is not permission. Useful agents need hard approval gates before anything expensive, public, or irreversible.

Read article
Field notes4 min read

What a useful weekly report actually contains

A dashboard shows numbers. A useful report explains what moved, why it matters, and what the team should decide next.

Read article
Engineering5 min read

Model choice should be infrastructure

Teams should describe the work they need done. Choosing the right model for each step should happen underneath the product.

Read article
Product4 min read

From repeated request to reliable routine

The third time a team asks for the same outcome, the system should notice—and offer to take the next one.

Read article

Put the ideas to work

Less prompting. More finished work.

Add OpenTag to Slack, connect the tools your team approves, and give it the job that keeps getting pushed to next week.