Human-centred AI for tomorrow's business
AmirahCo is where practical AI, clear workflows, and human control meet.
AI Workflow Setup
We help teams use AI for real work: sorting messages, drafting replies, summarizing meetings, organizing notes, and finding information faster.
Testing AI Tools
We test AI tools against real tasks to see what works, what breaks, and where human review is still required.
Simple Automation
We build small tools for repeated tasks that keep records, expose failures, and make work easier without removing human control.
Services that turn AI ideas into small, reviewable improvements.
Workflow Review & Process Cleanup
We review how a team works, identify useful and risky AI opportunities, then improve the prompts, instructions, checklists, and templates around the chosen workflow.
Internal Knowledge Setup
We organize notes, documents, and FAQs into a searchable internal workflow using tools you already have, with scoped access, clear sources, and reviewable answers.
Small Automation Prototype
We build a bounded prototype for one repeated task, including scoped permissions, logging, clear failure handling, and human approval where needed.
Projects and tools for more reliable AI-assisted work.
memory_Db
Public release v0.1.0 · MIT License. A memory system that helps AI coding agents retain project notes, decisions, bugs, setup rules, and past fixes during long builds.
Agent-Friendly Architecture (AFA)
Research & Development · A rules-and-templates framework for AI-assisted software work. It defines project structure, boundaries, documentation standards, and implementation constraints for safer AI-assisted development.
CJSON Framework
Exploratory · Compact AI Data Format. CJSON is an early idea for making AI outputs shorter and easier for software to check. It is meant to help AI send structured data without wasting as many tokens.
We keep projects small, clear, tested, and easy to review.
Build useful things. Start small.
Begin with a focused test and produce something people can use: a workflow, template, checklist, prototype, report, or guide. Scale only after the small version works.
Keep people in control. Build safely.
AI should support human judgment. Tools stay bounded with clear permissions, logs, review steps, and explicit responsibility for the final decision.
Write it down. Judge it honestly.
Record decisions, assumptions, failures, and known limits so others can review the work. Judge tools by usefulness, reliability, and fit, not hype.
Safer AI work needs clear access, approval, and output checks.
Access Levels & Limits
Exploratory research, not a legal standard, certification, or finished product. Define what information AI and human operators can access, and set explicit rules for what the system may do.
Human Approval Steps
Place clear approval points before sensitive actions or outputs are used, so a person remains responsible for the final decision.
Output Checks & Records
Check that AI output follows the rules, matches the task, and is consistent before anyone acts on it; retain enough context to understand what happened and why.
Custom AI workflow setup, from first conversation to daily use.
We design and configure practical workflows, support files, and step-by-step systems around the tools your team already uses.
001 · Understand The Workflow
A short consultation to learn how the work actually runs today, which tools are already in use, and where the friction is.
002 · Configure The Setup
A custom setup built around your existing business tools and processes, rather than a system your team has to work around.
003 · Roll Out And Use It
Guided implementation and deployment, ending in practical AI systems your team can use day to day.
Small, useful tools beat confusing AI hype.
Operating principle
AI should help people make decisions, not secretly replace them.
Human-control principle
Small tests, clear records, and tools people can actually use.
Delivery principle