Turn your automation idea into a spec any coding AI or no-code tool can build.
Turns one chosen automation idea into a build spec a coding assistant or a no-code tool can build from. You get the trigger, the steps, the prompt, failure handling, test cases and the running cost.
Free to run · Works in ChatGPT, Claude and the AI you already use
How to ask for it
Ask your AI in your own words.
Once Loadout is connected, your AI picks this power when you ask for the job. This request runs it, with the underlined parts swapped for yours.
Use Loadout to write a build spec for an AI receptionist at a dental clinic in Leeds
It also runs when you say things like
- spec this automation
- write the build spec
- plan an AI receptionist
- scope this build
- give my coding AI something to build from
What you get
Six pieces of finished work.
A build spec with steps, the AI prompt, failure handling and test cases.
- 01The steps, the trigger and what goes in
- 02The AI part written out, with its fixed output
- 03Human checkpoints, and who owns each one
- 04Failure handling, with an alert for every quiet failure
- 05Test cases, privacy limits and the monthly running cost
- 06A brief ready to hand to whoever builds it
Your route
Seven ways to run it.
Your AI asks where you're starting from and picks the route that fits. You can always choose another.
AI receptionist or booking assistant
When calls and messages ask the same things, and calls go missed after hours.
Lead follow-up and qualification
When form or phone enquiries wait hours for a reply.
Quote or estimate drafting
When quotes take days and come from a price list and notes.
Review reply drafting
When there are many reviews and few replies.
Inbox triage
When one shared inbox mixes enquiries, invoices, suppliers and spam.
Document and invoice processing
When staff type data from invoices, receipts or forms into a system.
Internal knowledge assistant
When staff ask the same questions and the answers live in people's heads.
What it asks you
- The opportunity and its success measure
- Systems, access and volumes
- Who does the task today, and who builds it
- Three recent real examples, one that went wrong
- The rules
- Privacy limits
The live data it reads
Scan website
Audits a website, covering technology, speed, AI crawler access, structured data, contact details, page weight and the layout at each screen width, with the top fixes.
What it costs
Free to run.
Free tools only. Your AI shows the cost before it starts, and nothing is spent without your yes.
Not made for
It does not choose what to automate, do the build, or cover large software projects.
The work around it
Part of the AI Automation playbook.
Find the work AI can take over, then build it. Every piece below comes from one run of the playbook, for one business. Pick another to see its set.
The whole AI Automation playbook
Find the work AI can take over, then build it.
Offer sheet
Packages, prices and a guarantee, pressure-tested.
Prospect list
Businesses that need it, with the owner's contact.
Outreach, call brief and proposal
A sequence, a call brief, a proposal to sign.
Opportunity note per prospect
The hours AI could give back, priced.
Opportunity map and build spec
The pilot, written so it can be built.
Go-live checks and handover pack
Live, tested, and handed over to the team.
Monthly impact report
What it saved this month, in hours.
Client file per client
Starting numbers, what was agreed, every month.
Opening 12 November
Join the waitlist.
Loadout opens to everyone on 12 November. Leave your email and we'll tell you the moment it does.
Goes well with
The powers that go with it.
A playbook runs its powers in order, from finding the client to the monthly report. Every power and playbook is in the one plan.
Questions
Before you run it.
How much does Automation Spec cost?+
Nothing. It runs on what you give it and free checks. Paid data, if you choose it, shows its cost first.
What does Automation Spec hand over?+
A build spec with steps, the AI prompt, failure handling and test cases. The steps, the trigger and what goes in. The AI part written out, with its fixed output. Human checkpoints, and who owns each one. Failure handling, with an alert for every quiet failure.
What do I need to give it?+
The opportunity and its success measure, Systems, access and volumes, Who does the task today, and who builds it, Three recent real examples, one that went wrong, The rules, Privacy limits. Your AI asks for anything missing.
Which AI apps does it work in?+
ChatGPT, Claude, Claude Code, Codex, Grok and Muse. You connect Loadout once and the same account works in all of them.