Blog
Product notes, AI automation, comparisons, and implementation learnings behind the software and services we build.
AI Document Workflow Automation: Route Exceptions Before Export
Build an extraction-to-review workflow: invoice/order mismatch example, routing rules, human approval, export checks and tool selection.
How to design a business workflow with AI agents
Turn a business need into inputs, tools, permissions, review and tests.
Internal Knowledge Retrieval for AI Agents: When RAG Actually Helps
Internal knowledge retrieval helps AI agents when the source material, permissions, context, and citations are designed around real workflows.
AI support triage: how to prepare and check a first pilot
Design a test with fictional data, permissions, human review and metrics without promising safety or savings.
What an agent skill is and how to apply it to a business process
Design a reusable agent procedure with permissions and human review, without confusing skills and MCP.
MCP: how AI agents connect to tools and data
Learn what MCP provides, how it works alongside agents and skills, and what to check before connecting business tools.
Accessible webinars: how to prepare live captions that help
Plan a webinar with live captions: test the audio, ask about attendee needs and prepare support when it matters.
AI and accessibility: everyday uses, limits and choices
Explore captions, image descriptions and reading support using AI, and learn to check errors, privacy and alternatives.
How to code with AI while managing cost and risk
Evaluate AI coding tasks, API pricing, permissions and review without promising savings.
How to evaluate AI outcomes and costs in a business process
Define a baseline, measure quality and cost, and avoid attributing savings without evidence.
How to explain AI product privacy without promising more than you tested
Turn broad privacy claims into verifiable explanations of a product's features, network use and limits.
How to document a real use case in your portfolio
Turn an authorised observation into a case record with context, evidence and limits, without inventing customers.