Writing
What we learned building these systems, written by the people who built them.
Featured · How we run Kaizen
How we triage email with rules, AI and checks in code
How Kaizen sorts its own email: written rules first, an AI model only for mail that no rule covers, and checks in code that can only keep a message in front of a person.
· 9 min read
Our inbox that was automated mail
54%
Every new message, in this order
- 1Written rules decide first.
- 2AI reads only what no rule covers. Unsure means no change.
- 3Checks in code keep people's mail in front of a person.
- 4A person taps Send on every reply.
More posts
-
How should an AI agent prepare email and invoice actions safely?
A practical design pattern for letting AI prepare email and invoice actions while keeping authorization and execution under application control.
How we run Kaizen
· 5 min read -
How to automate follow-up when deadlines slip
Record where each deadline came from, check the current status before sending a reminder, and send only real problems to a manager.
How we run Kaizen
· 7 min read -
How we cut ten alerting agents down to one
We had ten AI agents, and most of them could interrupt us. Now one agent decides what is urgent, and the rest wait for the daily briefing.
How we run Kaizen
· 5 min read -
The parts of programming that did not get automated
An agent checked every field and reported success, but the feature was still broken. Five skills still matter when AI writes the code. Developers used to learn them by accident, and now someone has to teach them on purpose.
Engineering practice
· 10 min read -
An instrument finder that sends researchers to the right person
We built an instrument finder for a multi-site research program. Every claim links to its source, and every answer ends with the person who runs the instrument.
Research centers
· 7 min read -
Why our proposal advisor answers common questions without AI
We built a proposal advisor for a National Science Foundation program. Common questions get a stored answer written by an expert. AI answers the rest from the program guidance.
Research centers
· 6 min read -
Adding plain-English queries to a city data warehouse
We built a plain-English query tool on top of the DC Department of Public Works data warehouse. The tool was small. The years of data integration underneath are what made it possible.
Government
· 6 min read -
Why every score in our assessment tool comes with its evidence
We built a self-assessment tool for research institutions, with nine dimensions. Each score keeps the evidence behind it, so a reviewer can trace any score back to its source.
Research centers
· 5 min read -
How we cut an AI agent's running cost by splitting the work between two models
We helped a tax accountant in India put an AI agent into their filing work. It worked on day one but cost too much to run. Now a large model builds tested tools once, and a small model uses them every day.
Engineering practice
· 9 min read -
Polka Dot Sky Software is now part of Kaizen
In December 2022, Polka Dot Sky Software merged into KZN Consulting, Inc. (Kaizen). Its alerting products for local governments, Unified Alerts and MyDPW, continue as Kaizen service alerts.
How we run Kaizen
· 2 min read
Working on a similar problem?
In a 30-minute call, we will tell you whether it is worth building and roughly what it would cost.