Start with the work. Define the evidence.
An implementation note for this portfolio, 13 October 2026. A practical AI implementation begins with an organizational problem and a way to assess whether the resulting workflow helps.
Published · By Creig Phiri
Specify the workflow
Describe who needs information, what decision they are making and which sources they can use. Agree on what the system may do and where a person must decide. A fluent answer is useful only when it fits that operating context.
Make context inspectable
Company Brain starts from a practical problem: company understanding is scattered across people and tools. The implementation uses stable records, explicit sources and change history so that context can be inspected and maintained.
That structure is a starting point for evaluating retrieval. It does not prove that an agent has found every relevant fact. Keep the question of missing or outdated context visible.
Separate a demonstration from an outcome
A demonstration can show that a workflow is possible. A business result needs additional evidence: a baseline, a defined measure, real use and a defensible account of what changed.
Company Brain remains an MVP for that reason. Its internal implementation demonstrates an approach to organizational context; measured revenue impact is a separate question.
Keep people accountable
Define who reviews the answer, who maintains the sources and who responds when something fails. Give those people enough visibility to challenge the result. Permissions, verification and feedback belong in the workflow from the beginning.
These ideas also appear in my Edmonton Startup Week teaching material: set it up, run it and keep it working.
contact.sh
Bring the problem.
Let’s talk implementation.
For practical AI and software implementation, contact Boring Tech Solutions with your workflow, constraints and what you need to demonstrate.
Contact Boring Tech Solutions