.avif)
Sends go up, list size goes up, connections go up, and pipeline stays flat. Every team in that position starts optimizing the layer they can see, which is usually the wrong one.
A GTM engine has four layers, and each one only works when the one below it works: data, demand, conversion, retention. When several are broken, the answer is always the lowest one, however much the others hurt today.
This skill runs that diagnosis as a conversation. It asks one question at a time, adapts to your answers, and refuses to recommend a tool before it knows the state of your data.
Callout, Uncomfortable truth: Most audits end with a shopping list, so the team buys the most visible thing instead of fixing the cheapest one. This skill ends with a section called "Do not do this yet".
What is inside: a structured sales process audit your AI assistant can run
- One markdown file, ready for Claude, ChatGPT or any assistant that reads markdown
- Four audit blocks: data, demand, conversion, retention, each with an opening question and the follow-ups
- A reading key per block: what a given answer tells you about the layer
- Three scoring levels, with the rule that "runs unattended" requires a named owner
- The four build paths, so you see where you stand rather than only what is broken
- A one-screen report format: levels, path, weakest link, fix order, what to postpone
- A first step you can take this week without buying anything
- The rules that keep the audit honest, including no pitch and no tool talk before the data is known
A preview: how the audit reads your answers
Four layers, audited in dependency order
The audit starts at data: structure, cleanliness, currency and the rule for what counts as qualified. Everything above it runs on guesses until that is in place. Then demand, which is outbound, inbound and signals, and which sends messages to the wrong people at the wrong time when the layer below is broken. Then conversion: routing, pipeline and everything between a lead and a call, where leads wait and deals stall quietly. Last, retention, the lost opportunities and existing accounts that leave the system for good when nobody owns them.
A broken data layer under a sophisticated outbound motion is a second floor on no foundation. The skill will say so, and it will not let you start at layer 02 because that is where the pain shows.
Three levels, and the one that needs a name
Does not exist. Runs manually. Runs unattended. No points, no percentages, because they imply a precision an interview does not have.
"Runs unattended" only counts when you can name the person who maintains it. A process that works today and has no owner degrades within a quarter, so an unowned workflow scores as manual.
Key rule blockHeader: One rule the audit never breaksRule designation: Rule 1 of 8Quote: "Never propose a tool before you know the state of the data."Attribution: GTM Engine Audit, section 3
How to run it
Claude: drop the file in as a skill, or paste it into a project. ChatGPT: attach it to a custom GPT or paste it at the start of a chat. Then write one line about your setup and answer the questions as they come. It takes about fifteen minutes and at most twenty questions.
Who this is for: teams that cannot say which layer is broken
- Activity is rising and pipeline is flat, and the reason is not obvious
- You are about to automate, and you want to know whether the data underneath is ready
- Your CRM is messy and three people have three different answers about where a contact lives
- You want a second opinion before you brief a partner, hire an SDR, or buy another tool
If you already know which layer is broken and what to do about it, skip the audit and read the GTM Engine playbook instead.
Run it yourself, or have someone run it with you?
The skill is written to be run without us, and it is deliberate that it ends without a link or an offer. The output is a diagnosis and an order of work, and plenty of teams can execute both alone.
It stops being enough when the fix spans several layers at once, when the data work needs an owner your team does not have, or when the first step turns into a quarter of work. We compared both models, with the numbers, in GTM engineer: in-house or agency.
We have run 250+ campaigns and built 50+ revenue engines. The framework the skill audits against is the one we build from.
FAQ: the GTM engine audit
What is a GTM engine audit?
A structured review of the four layers a go-to-market system runs on: data, demand, conversion and retention. It places each layer on a three-level scale, names the lowest broken one, and puts the fixes in dependency order.
What is a skill file, and which tools does it work with?
A markdown file with instructions an AI assistant follows. It works in Claude as a skill or inside a project, in ChatGPT as a custom GPT or a pasted prompt, and in any other assistant that reads markdown.
How long does the audit take?
About fifteen minutes. The skill asks one question at a time and stops at twenty questions, because past that it is collecting detail it will not use.
Do I need a CRM to run it?
No. The skill is written not to assume one. A spreadsheet and a shared inbox are a valid answer, and they place your data layer on the scale like anything else.
What do I get at the end?
A one-screen report: a level per layer, the build path you are on, the weakest link, the fix order, what to postpone, and one first step you can take this week without buying anything.
Will it recommend tools?
Only after it knows the state of your data, and never an AI agent for a process nobody has run by hand. Tool recommendations made against unknown data are guesses.
Is it free, and can I change it?
Yes to both. Download it, edit it for your own stack, share it with your team.
What if I disagree with a finding?
Tell it what it got wrong. The skill is instructed to ask, listen, and revise the level when the correction is a fact.
.avif)
