Blog
Writing
Notes on AI and software.
AI Agent Implementation Timeline: How Long It Takes
A standard agent implementation for a service business takes 2–4 weeks from scoping to live. Integration complexity and process documentation drive timeline — not AI sophistication.Before You Hire an AI Agent Implementer
The businesses that reach a working agent fastest are the ones who documented their process before the first call. What preparation actually looks like — and why it cuts implementation time more than any other variable.Which Workflows to Automate First
Ranking workflows by integration overlap — not business value — cuts total build cost in half and makes each new agent faster than the last.What AI Agent Maintenance Actually Involves
AI agent maintenance is rewriting prompts when the business changes, patching integrations when tools update, and reviewing logs for silent edge cases.Is Your AI Agent Actually Working?
Without defined success criteria there is no way to tell a running agent from a working one — and degradation stays invisible until it's expensive.Your Second AI Agent Is Harder Than the First
The second build surfaces every undocumented decision the first one made, and those decisions become constraints you did not know you had chosen.Custom Agent vs Off-the-Shelf: How to Choose
Off-the-shelf agents bake in assumptions about approval chains, data formats, and output destinations. When those assumptions don't fit your workflow, workarounds compound — and maintaining them costs more than a purpose-built custom agent.Custom AI Agent Development: The Five-Stage Build
A custom agent build stalls at stage 1 — process documentation — more than any other stage. If the workflow isn't written down, it can't be built. Here is the full build process, stage by stage, with timelines and what to do when something breaks.Hermes Setup Guide: Deploy on Your Own Server
A Hermes setup needs defined work, scoped tool access and tested review rules. Separate installation, local skills and external AI processing before launch.How Hermes Learns: Saving and Reusing Skills
Hermes can save reusable task instructions as skills. Learn where they live, how sharing works, and how to check whether a revision improves results.What Is a Custom Agent? When to Build vs. Buy
A custom agent is an AI agent built for one business's specific data, processes, and output requirements. The model is a commodity — the integration work, approval logic, and workflow design are what make it custom. Here is when to build, when to buy, and what it costs.What Is an AI Agent? Definition, Types, and How They Work
An AI agent is software that takes actions on your behalf — reads inputs, makes decisions, and acts in external tools like Gmail, Slack, or HubSpot. Unlike chatbots, agents execute. Here is how they work, what they cost, and when they make sense.What Is Hermes? AI Agents and Reusable Skills Explained
Hermes is Nous Research’s open AI agent with reusable skills and 20+ messaging platforms. Learn how its learning, permissions and deployment actually work.What AI Agent Implementation Costs a Small Business
Most implementation quotes cover the build and nothing else. Integration and maintenance together often cost as much again over three years.What Building Your Own AI Agents In-House Actually Takes
In-house AI agent development is a staffing commitment with ongoing obligations, not a build project with a finish line. The economics rarely favour SMBs.Why Most AI Agent Projects Stall After Ninety Days
Most AI agent projects don't fail at the build. They stall after go-live, when ownership disappears and no one's job is keeping the agent running.How to Brief an AI Agent for Your Workflow
A useful AI agent brief defines triggers, inputs, outputs, exceptions, and escalation. Test the acceptance criteria before connecting it to live work.When Is a Process Ready for an AI Agent?
The readiness bottleneck is not technology. It is whether the process has ever been specified clearly enough to run without the person who knows it.AI Agent Limitations: What They're Actually Bad At
The failure mode most business owners don't expect: AI agents don't fail on hard tasks. They fail on vague ones. Here is what that looks like in practice — and how to screen for it before building.Have something in mind?
Tell us what you want built. We'll tell you what it would take.