Handover Method
How I implement AI and automation
Five stages. An idea gets assessed, tested against your real data, built into something that runs every day, and then handed to your team. The last stage is the point: most projects stop at deployment, which is exactly where they start to rot.
Why this method exists
AI projects rarely fail for technical reasons. They fail for these:
- Someone picked the tool before anyone looked at the process.
- It was tested on tidy made-up examples, then met the real data.
- It works, but only the person who built it understands it.
So each one gets its own stage. The process comes before the tool, the pilot runs on your data rather than a demo, and handover is real work with its own deliverable instead of a final email.
Stages
Each stage answers one question and leaves you with something you can hold.
Assessment
Where is the most manual work today?
- What happens
- I look at how the process actually runs: who does what, in which systems, how long it takes, where errors repeat. I talk to the people who do the work. You can run a first, self-serve version of this stage yourself: the free Oopsify audit shows in 10 minutes what in your business is worth automating.
- What you get
- A process map and a list of the places where automation would pay off most.
- Who is involved
- The manager and one or two people who run the process.
Priorities
What to automate first, and why?
- What happens
- I rate each opportunity by benefit, risk, and how much data you already have. We pick one or two places to start.
- What you get
- A ranked list with reasoning, and a brief for the first solution: what it will do, what it will not do, how we will measure it.
- Who is involved
- The manager, the person who makes the decision.
Pilot
Does it work on your real data?
- What happens
- I build a working prototype and run it on your real data and real cases. We measure where it fails and how much time it saves.
- What you get
- A working prototype, pilot results, and a decision: build the system or stop.
- Who is involved
- The people who run the process every day.
System
How does it run every day, without me?
- What happens
- The prototype becomes a system: integrations with your tools, error handling, monitoring, documentation. It runs in your environment, in your accounts.
- What you get
- A running system, its documentation, and a plain description of what to do when something breaks.
- Who is involved
- Your IT person or provider, if you have one.
Handover
Can your team run and extend it on its own?
- What happens
- I train the team to use, maintain, and change the system. Together we review what to automate next and how to evaluate it without me.
- What you get
- A trained team, access and code in your hands, a list of next steps.
- Who is involved
- Everyone who will use the system.
What it is not
Four ways this method differs from typical AI agency work.
Not a template.
The solution is built for your process, not adapted from a catalogue.
Not tool first.
Tools are chosen after the assessment, not before it.
Not a rental.
Code, access and documentation stay with you. You can continue without me.
Not deploy and leave.
Handover is a separate stage with its own deliverable, not the last day of the project.
Where each offer fits
The seven offers use different stages of the method. The Custom AI Solution is the only one that runs all of them.
| Stage | Custom AI Solution | Personal AI | Audits | Team training | Consultation | AI Partner |
|---|---|---|---|---|---|---|
| Assessment | Yes | Yes | Yes | Brief | Yes | No |
| Priorities | Yes | Yes | Yes | No | Yes | No |
| Pilot | Yes | With you | No | Exercises | No | No |
| System | Yes | Yes | No | No | No | Ongoing |
| Handover | Yes | Yes | A plan | The whole stage | Next steps | Ongoing |
What you have when it is done
A running system your team uses every day and knows how to change.
Documentation that a non-programmer understands. Access and code in your own hands. And most of all: your people can evaluate the next AI idea themselves, they know what to ask, how to test on real data, and when to say no.
Bring me a problem.
A free 15 to 20 minute intro call. You tell me what you want to achieve, I tell you whether I can help and what it would take. If AI is not the answer, I say so.
No pitch. A diagnostic conversation.
