Session brief · ready 10:30 · 60 min
Custom AI agents for advisors, coaches and consultants
Imagine having 5+ extra hours every week.
We review your typical week. I show you which repeated work your agent can safely take over, build it around your method, and work with you until you are confident using it, and happy with the time it saves you.
The 5-hour guarantee
After 14 days of real use, it must save you at least five hours a week. If it does not, I refund 100% of my build fee.
One repeated job. Built, tested and taught to you.
- Your typical week reviewed before the build
- 20 real situations tested before handover
- 5+ hours a week or my build fee back
Your first win is not an agent. It is knowing what to build.
Most people start with a tool. That is backwards. We start with the work you repeat and the result you want. In the first session, you learn what AI should do, why it matters and whether it is worth building.
Walk into every client call prepared
Turn intake forms, previous notes and current context into a focused agenda, better questions and the points that deserve attention.
Turn a conversation into action
Use notes or transcripts to create a client-ready summary, clear next steps and a useful follow-up email.
Help clients keep moving between calls
Give clients bounded guidance based on your method without pretending the agent is you or removing human judgment.
Stop starting from zero
Use your ideas, calls and frameworks to support research, proposals, onboarding, documentation and content without flattening your voice.
Custom only when it earns the premium
You may not need a custom build.
An existing tool is the smarter buy when the job is standard and the platform already does it well. Custom work earns its place when the value lives in your method, your workflow and the details generic software cannot know.
Use existing software when
- You need standard inbox, calendar, meeting or admin help
- A basic chatbot trained on your content is enough
- Your current platform already does the job well
- The workflow does not repeat often enough to justify custom work
Custom earns its place when
- The work crosses several tools or data sources
- The agent must apply judgment that is specific to your method
- Mistakes need clear limits, tests and human handoffs
- You need to understand, own and improve the core logic
One useful agent. Built for you. Taught to you.
You do not get a generic chatbot, a folder of prompts or a 40-page manual nobody wants to read.
You get one agent built around real work in your business. Then we use it together until you understand where it helps, how to use it and when a person still needs to take over.
Best-use-case map
A clear definition of the user, task, inputs, output, value, boundaries and first result worth measuring.
One working custom agent
Instructions, questions, decisions and response formats structured around your method and standards.
Method and knowledge setup
The useful source material organised for the agent, with irrelevant or risky material removed.
Limits and human handoffs
Clear rules for what the agent can do, what it should refuse and when you need to take over.
A 20-case acceptance test
Normal, unclear, difficult and out-of-bounds situations tested before the agent is trusted with real work.
Live training on your work
A working session using your own examples so you learn by doing, not by watching another generic AI tutorial.
Quick-start operating guide
A simple guide showing what to give the agent, how to use the output and how to spot when the answer needs human review.
Ownership handover
Your instructions, knowledge map, tests and documentation delivered in files you can read and keep.
30-day scorecard
A simple way to measure use, time saved, output quality, mistakes and the next improvement worth making.
You should not need me forever to use what I built.
By handover, you should be able to explain what the agent does, give it the right inputs, recognise a good output and know when human judgment is required.
You receive the instructions, knowledge map, tests and guide. The core logic is not hidden in a box only I can open.
- Learn using your own real examples
- Know how to spot a weak or risky answer
- Update the source knowledge yourself
- Keep the files that create the value
I've used AI agents for hundreds of hours to build software and systems for my clients.
Every build came with headaches. Each one taught me what breaks, what works and how to move faster. You get those lessons without paying for my mistakes.
Local Community Platform
An open-source platform for local communities to preserve posts, resources, member profiles and community-shaped events beyond a busy group chat.
What this proves Reusable product architecture, invite-gated membership, public participation, permissions and a real community running on the platform.
GitHub 
Command Center
The private system I use to manage projects, prospects, CRM, calls, content, publishing and AI agents in one place.
What this proves Operational context, agents connected to real work and a private system built around its user.
ServiceKit
A Jobber and ServiceTitan alternative with CRM, scheduling, dispatch, quotes, invoices, payments, timesheets, reviews and AI-assisted workflows.
What this proves Complete business workflows, multiple users, data, permissions and work that cannot fit inside a prompt.
Booking Kit
A branded booking system with availability, public booking pages, calendar connections, email, embeds, paid bookings and an owner dashboard.
What this proves Client experience, calendar integrations, availability rules and the operations that happen after the click.
PostKit
A self-hosted social scheduler for composing, queuing, publishing, media, accounts, retries and results.
What this proves Queues, APIs, states, retries and the unglamorous work that makes automation reliable.
Signal to Noise Ratio
A Chrome extension that tracks whether browser time is moving the work forward or leaking into distraction. All data stays in the browser.
What this proves Product judgment, local privacy and a small tool built to change one specific behaviour.
Media Downloader
A Chrome, Brave and Edge extension with a local yt-dlp server that detects media, downloads files and HLS or DASH streams, and keeps progress on the user’s machine.
What this proves Browser extension work, native messaging, a local server, stream handling, background state, retries, settings and desktop integration.
From AI confusion to one useful system.
Find the value
We identify the repeated work, the desired result and the first useful win. If an agent is overkill, I will tell you.
Build and test it
I turn your method into instructions, knowledge, boundaries and an agent that survives difficult cases.
Learn on your own work
We use the agent together with your actual examples. You see what changes, why it matters and how to improve it.
The commitment, made clear
Know exactly what you are buying before the build starts.
You get a plain-English scope before any build starts. It turns custom AI into one specific job with one specific plan.
Build
The job, inputs, outputs, integrations, boundaries, 20-case acceptance test, build price and delivery plan.
Run
Where it lives, how it starts work, expected platform, model or hosting costs, and which accounts stay in your name.
Maintain
Who updates knowledge, reruns tests, fixes defects or broken integrations, and how new capabilities are scoped.
A useful agent starts with real work and real expertise.
AI can extend a method. It cannot invent the method that does not exist yet.
This tends to work when
- You already have paying clients
- Your work follows a method you can explain
- One task happens often enough to matter
- You want to understand and use the agent yourself
- You are willing to test it on safe, real examples
- You want better delivery, not the disappearance of the human work
It is the wrong build when
- You want a magic lead machine without a strong offer or distribution
- You expect AI to invent expertise you do not have
- The agent would give unsupervised medical, legal, financial or therapeutic advice
- You want five agents before proving one is useful
- You want the words AI agent on your website more than a better business process
This service is new. The ability to build is not.
The software on this page is real. What I do not yet have is a wall of case studies from this exact service. I would rather say that plainly and earn the proof through useful work.
Start with one task you repeat
Find the AI use case worth building first.
Tell me where time disappears in your week. I will help you work out whether AI should handle it, support it or stay out of it.