Your automations, off the laptop and under control.
How we take it on
Nothing is switched off, rebuilt or rushed. We find what you rely on, move it over one job at a time, run each one beside the original until it's proven, and then keep watch for good. You'll notice the change when the morning digest arrives.
The three steps
Connect.
A free stack check finds what you rely on; then we connect StackProbe to your systems with your own accounts.
Set your rules.
You decide what the AI may do, system by system. Every client starts in Assisted mode.
StackProbe runs and watches.
Your jobs run in your sealed workspace, every action checked and recorded, with our team watching.
The timeline, stage by stage
Stage 1: Stack check
Free. About five minutes of your time, then a few days for us.
You answer a short questionnaire about what runs automatically in your business, where it runs and whose logins it uses. You see a summary on screen straight away. Within a few business days we send a plain report: what you depend on, what's fragile, and what we'd move first.
What you do: Answer the questions. Forward the report to whoever else should see it.
Stage 2: Scope and quote
A short call, then a written quote.
We go through the report together, agree which jobs to move and in what order, and write down the rules for each one: which systems it may touch, what it may do, and who approves what. You get a quote based on what you actually run: a fixed monthly fee plus usage.
What you do: One call, then say yes or no. No commitment until you sign.
Stage 3: Connect
Usually a day or two.
We set up your sealed workspace and connect it to your systems using your own accounts, asking only for the permissions each job needs. Passwords and keys come off the laptop and into encrypted storage. Your staff get access through Teams, email or their Claude app, signing in with their work account.
What you do: Approve each connection from your side (you'll get a normal sign-in prompt from Microsoft, Google, Shopify and so on). Nothing to install.
Stage 4: Move over, side by side
One job at a time. Days to a few weeks, depending on how many.
Each automation is moved onto StackProbe and run beside the original. Both run; you compare. When the StackProbe version has matched the original for long enough that you're comfortable, the original is retired. Existing scripts are moved as code and keep working the way they did, with monitoring added.
During this stage every job is in Assisted mode: the AI drafts, a person approves. You'll see the first approval requests arrive in Teams or your mailbox, and the first entries appear in the record.
What you do: Compare outputs, approve what's correct, tell us what isn't. Decide when each original can be switched off.
Stage 5: Running and watched
From here on.
Your jobs run on schedule in your workspace. Health checks run every five minutes; failures alert our team, and we fix them. You get a morning digest and a monthly summary. When a job has proven itself and you want it to act without approval, we write down its limits and move it to Autonomous mode, one job at a time.
What you do: Read the digest. Approve what's waiting. Change the rules whenever you like.
What we need from you
Less than you might expect.
- About an hour, total, across the stack check and the scoping call.
- Sign-ins, when we connect each system. You do these yourself; we never ask for your passwords.
- One or two approvers: the people who'll say yes or no to what the AI drafts.
- Honest feedback during the side-by-side stage. If an output is wrong, say so; that's how the rules get tightened.
You don't need to document anything first. Finding out how things currently work is our job, and it's part of the stack check.
What you get at each stage
| Stage | You get |
|---|---|
| Stack check | A plain report of what's fragile |
| Scope and quote | Written rules per job, and a price |
| Connect | A sealed workspace, your logins off the laptop |
| Side by side | Proof each job works, before anything is switched off |
| Running | A morning digest, a monthly summary, and someone responsible when it breaks |
What if…
…the person who built the automation has already left?
That's one of the most common reasons people call. We work from what's left: the files, the logins that still work, the output everyone's used to. Often we can rebuild from the output alone.
…I don't know what I'm running?
Most owners don't, fully. That's what the stack check is for. Answer what you can and tick "not sure" for the rest.
…an automation is in Zapier, Make or n8n?
We can leave it where it is and watch it, or move it, depending on how fragile it is and what it touches. The stack check report says which we'd recommend and why.
…something goes wrong during the move?
The original keeps running until you say otherwise, so nothing stops. That's the point of side by side.
…I want to stop?
Stop. Export everything (files, memory, skills, rules, history) at any time. We'll help you move it.
Start with the free stack check
Five minutes. A plain report of what's fragile. No commitment.
Start the free stack check