AI adoption
2 min article
01
Check Your Connections Before Buying Another AI Tool
Before adding another AI subscription, look at the connections and tools already available in the software your business uses.
Check existing connections before paying for another AI tool.
-
Check existing connections before paying for another AI tool.
-
Consider the Microsoft tools your business already uses alongside ChatGPT and Claude.
Read article →
Make Your AI Workflow Usable by a Teammate
An unfamiliar CRM lookup shows why shared AI skills need to help someone beyond the person who created them.
A useful AI workflow should be reusable by a teammate.
-
A useful AI workflow should be reusable by a teammate.
-
A shared second brain can hold procedures and skills alongside knowledge.
Read article →
Before You Rebuild Your CRM, Map the Work
A Salesforce replacement request raised a bigger question: whether a new system was worth the maintenance, disruption and time running two tools.
Compare a replacement project with the other problems in your business.
-
Compare a replacement project with the other problems in your business.
-
Account for maintenance and the time people may spend using two systems.
Read article →
Business workflows
2 min article
04
Describe the Work Before You Build More Software
An AI agent can offer another way to carry out business logic. Start with the process you need done, including a simple task such as uploading information to a CRM.
AI coding tools are only part of the change Alek sees in business software.
-
AI coding tools are only part of the change Alek sees in business software.
-
An agent can take instructions in English to carry out a business process.
Read article →
AI implementation
2 min article
05
Your team's second brain needs someone to own it
Shared AI knowledge brings decisions about data structure, storage, security and connections. Give one person the authority to drive that work.
Employee training is only part of implementing shared AI knowledge.
-
Employee training is only part of implementing shared AI knowledge.
-
Data structure, storage, security and connections still need engineering judgment.
Read article →
Business operations
2 min article
06
Let Your Team Reuse What It Learns
A shared folder of company knowledge gives employees and AI agents somewhere to find procedures, skills and work that may already exist.
Useful automation needs to be accessible beyond one person's computer.
-
Useful automation needs to be accessible beyond one person's computer.
-
A second brain can start with directories and readable text files.
Read article →
I wanted to give AI a spec and walk away
Why I built Necronomicode around Codex, and why specs, handoffs and stop conditions still make it useful in my daily work.
A terminal Codex supervisor coordinates bounded work.
-
A terminal Codex supervisor coordinates bounded work.
-
Specs and explicit handoffs let the work continue without constant supervision.
Read article →
How I'd use local AI before sending a document to a cloud model
A proposed local preparation layer for cloud AI: selective context, consistent labels and checks before private material leaves the local system.
Local AI could help prepare the context a cloud model receives.
-
Local AI could help prepare the context a cloud model receives.
-
Replacing names is pseudonymization, not anonymity.
Read article →
AI workflows
2 min article
09
What ComfyUI taught me about image workflows
Learning image generation through connected steps, reusable workflows and the difference between correct wiring and a useful result.
A visual graph makes the steps in an image workflow inspectable.
-
A visual graph makes the steps in an image workflow inspectable.
-
Reusable builders let selected inputs change within a stable setup.
Read article →