Writing
Notes from the work.
What we have learned building software for real businesses, written for the people who run them.
How to tell whether a workflow is ready for AI automation
A short checklist we run against every candidate process before quoting. Volume, ambiguity, and reversibility decide far more than how interesting the problem sounds.
AI agents versus traditional workflow automation
They solve different problems and they are priced differently. Knowing which one a workflow needs is the difference between a two-week project and a six-month science experiment.
How to automate back-office operations without losing control
Control is not the opposite of automation. It is a set of design decisions about access, thresholds, reversibility, and what a human is allowed to override.
What a reliable AI agent needs beyond a language model
The model is the smallest part of the system. Retrieval, validation, tool boundaries, retries, and observability are what make an agent safe to leave running.
How we connect AI agents to the tools a business already uses
Most operations problems are integration problems wearing an AI costume. This is the order in which we check for an API, a webhook, or a last resort.
When a business should not use an AI agent
Sometimes a rule, a form, or a single hire is the right answer. We turn down work when automation would add cost, risk, or a maintenance burden the team cannot carry.
Start a project
Tell us what you are building or fixing.
hello@nexumlab.comInclude what you are building, your timeline and a rough budget. We reply within one business day. No forms, no chat widget, no sales sequence.