01
Finding a useful starting point
We talk through the jobs you keep doing and choose one where an assistant could be useful.
Start here
AI Workbench
Set up AI around the way you actually work.
I build an assistant around your role, company knowledge and the work you repeat. We start with one useful task, then connect the tools or add the dashboard it needs. I test it with you and show you how to use it.
AI Workbench is part of the MorpheOS AI implementation service, and is priced for the work involved. It is separate from the IntelligenceOS download.
Your work and background
The information your assistant needsSkills
Useful tasks you can repeatSafety
What AI can draft, and what needs youHandover
How to use it without turning it into admin
One senior professional
Agreed recurring tasks
Connections and dashboard as needed
Human approval built in
The problem
A blank chat does not know your job.
A blank chat box does not know your role, your company, your audience, your tone, your recurring tasks or your standards. You end up explaining the same context again before it can help.
The practical fix is a small workbench containing the context it needs, the jobs worth repeating and the actions that still require your approval.
What you get
A working setup for the tasks you actually repeat.
The setup can work with your existing AI tool or a dedicated assistant workspace. We choose the approach around the task and the information it needs.
02
An assistant that understands the job
I bring together the background and examples the assistant needs, including how you like the work written and what information it can use.
Context
03
Less explaining next time
Once we’ve worked out a useful way to prepare a brief, review copy or turn meeting notes into actions, I make that something you can use again.
Reuse
04
Control over what it does
You decide what the assistant can access and which actions need your approval, before it is connected to anything.
Control
05
Help getting comfortable with it
I show you how to use it, we try real tasks together and I include one round of adjustments. Ongoing help is available separately.
Adoption
Who it helps
For people whose AI output still sounds generic or forgets the job.
The best first version is deliberately narrow: one senior professional, one workbench, a small number of high-use workflows, proper checks.
A good fit for
- Senior marketers, founders, consultants or operators.
- You already use AI, but the output still feels generic.
- You keep explaining the same tasks and background to AI.
- You know which company information you’re allowed to use with AI.
Not the right fit if you need
- A custom SaaS product beyond the agreed assistant setup.
- Unrestricted access to live systems.
- Use of company material without confirmed permission.
- Specialist legal, compliance or security sign-off.
How it works
We begin with one role and the work you repeat most often.
The sprint starts with basic but important questions: what are you allowed to use, what should AI never do, and which workflows are worth setting up first?
The information it can use
We check that the chosen tool is suitable for the company information you want to use.
The first version
I build the assistant’s context and reusable tasks, plus the connections or dashboard we have agreed.
Trying it on real work
We test the setup against useful examples, not perfect demo prompts.
Making it yours
You keep the setup and instructions, so you can use it and know where to start if something needs changing.
Built from real work
Janet: an assistant for a senior marketing role.
In June 2026, I built a dedicated assistant workspace with role context, reusable skills, a dashboard and rules for actions needing approval. Read the Janet implementation story to see the build.
We agree what runs automatically and what waits for your review.
The assistant should show its sources and flag what it cannot verify.
We check which company information can be used before adding it.
Tool connections and dashboards are included only where we have agreed to build them.
See the output
An original comic, paired with technical analysis.
For my own publication, Swarm Signal, I paired an original four-panel story with an article about agents using several tools to finish a task.
The result is Nell, a fictional office clerk, and a cupboard full of plans. The four-panel story gives a dry British take on a familiar problem: lots of activity, and very little finished.
The illustration has an editable vector source and a full-size image. It offers a fictional way into the same question: what actually got done?
Read the Swarm Signal article · Open the editable illustration · Get the A4/A3 print files on Etsy
Working together
First, we understand the work and what AI is allowed to see.
I build this around your role and the tools you use. We can start with a conversation about the work you repeat and where AI currently falls short.
I’ll explain what I’m building and flag any hosting, AI subscription or API costs before we start. Start with a description of the task; keep confidential company or client material out of the first email.
Reporting, campaign workflows and website journeys can connect to the same work. See how the MorpheOS service fits together.
Before work begins
What the work will cost.
Before we start, you’ll know what I’m doing, the price, what I need from you and when we expect to finish. If the work changes, we agree the next step first.
Get in touch
Tell me what you want AI to help with.
Send me your role, how you use AI now and two or three pieces of work you repeat. Keep confidential employer or client material out of the first note.