AI doesn't replace CRO judgment, it removes the busywork around it: diagnosing which funnel stage actually leaks, drafting the page variant, mapping the activation flow, and setting up a test with proper sample size math instead of eyeballing it. gtm-skills' Experimentation Lead pack is 10 free, open source Claude Skills covering acquisition, conversion, and activation. Retention has no skill yet - that gap is named honestly below, with a churn-prediction tool coming next.
There's no shortage of "will AI replace CRO" takes online. We checked whether anyone's actually searching for that question - every phrasing of it, run twice, across two separate research passes. Every single one came back with zero measurable search volume. Nobody's typing that into Google. So this post isn't going to manufacture a threat that isn't there.
What is real, per DataForSEO keyword-volume data pulled for this piece: "ai conversion rate optimization" gets searched about 50 times a month. Modest, but it's a signal that people want to know what actually changes in the day-to-day CRO job, not whether the job disappears. Here's the practical version. AI is good at the diagnostic grind - reading funnel data against benchmarks, drafting a landing page variant, mapping a flow. It's not good at deciding whether a result is trustworthy or whether a price change is worth the risk. That split is what this pack is built around.
TL;DR
- gtm-skills' Experimentation Lead pack is 10 free skills: funnel leak diagnosis, activation mapping, landing page generation, pricing strategy, personalization routing, Bayesian test design, onsite search merchandising, ad spend waste triage, checkout auditing, and product page review.
- Install once with npx skills add sidchaudhary/gtm-skills.
- The pack covers acquisition, conversion, and activation. Retention has zero coverage - that's named honestly, not hidden.
- Customer churn prediction (320 searches/mo, $18.85 CPC per DataForSEO) is the best-supported "coming next" skill in this whole 7-post series.
How to install these in about 60 seconds
- Run npx skills add sidchaudhary/gtm-skills.
- Create a Claude project per skill - "Leak Finder," "Hypothesis Engine."
- Paste the skill's SKILL.md into the project instructions.
- Start a chat with your funnel data, activation metric, or pricing question.
The 10 skills, mapped to the funnel
10 skills now, mapped to the funnel stages a testing program actually touches: acquisition leak diagnosis, three conversion-side skills (page generation, personalization routing, pricing), activation flow design, Bayesian test design underneath it all, and four newer additions for onsite search merchandising, ad spend waste, checkout UX, and product page review. Retention has no skill yet, and that gap is called out below instead of hidden.
| Funnel stage | Skill | What it does |
|---|---|---|
| Acquisition | The Leak Finder | Diagnoses funnel drop-offs against real conversion benchmarks and ranks root causes |
| Conversion | The Page Shipper | Generates a conversion-optimized landing page as deployable HTML and Tailwind |
| Conversion | The Variant Router | Maps audience segments to content variants, with a condition, an experience, and a measurement plan attached |
| Conversion | The Price Point Finder | Works through value metric selection, tier structure, and price-increase timing |
| Activation | The First Mile Mapper | Designs the post-signup flow to the "aha moment" and diagnoses drop-off at each step |
| Testing itself | The Hypothesis Engine | Designs Bayesian A/B tests with sample size math, guardrails, and exit criteria |
| Retention | None yet | Zero coverage - the honest gap, covered below |
Acquisition: The Leak Finder
Diagnoses funnel drop-offs against real conversion benchmarks, then ranks root causes instead of just flagging where the numbers look bad. This is where testing should start - know the actual leak before you design a test around a guess. Read the skill.
Conversion: The Page Shipper
Generates conversion-optimized landing pages and outputs deployable HTML and Tailwind CSS - not a wireframe you still have to build. Point it at a leak the-leak-finder surfaced and get a page ready to put in front of traffic. Read the skill.
Conversion: The Variant Router
Builds rules that map audience segments to content variants, with a condition, an experience, and a measurement plan attached, ordered by priority with a default fallback so nobody falls through the cracks. For personalization that needs actual logic behind it, not a coin flip. Read the skill.
Conversion: The Price Point Finder
Works through value metric selection, tier structure, actual price points, and price-increase timing. Pricing is a test with real financial stakes, and it deserves the same rigor as a landing page test - just with tighter guardrails. Read the skill.
Activation: The First Mile Mapper
Designs the post-signup activation flow: what has to happen before someone hits the "aha moment," in what order, and how drop-off at each step gets diagnosed. Use this when people are signing up but not sticking. Read the skill.
Testing itself: The Hypothesis Engine
Designs Bayesian A/B tests with Thompson sampling, proper sample size calculations, guardrails, and exit criteria. This is the skill that turns "we ran a test" into a result you can actually trust, instead of stopping the moment the numbers look good. Read the skill.
Conversion: The Checkout Auditor
Audits your actual cart and checkout flow from screenshots or a walkthrough plus your policy details, and names the specific friction: form fields nobody needs, missing payment methods, absent trust signals, one step too many. For when abandonment is high and you want the reason rather than the benchmark. This is a direct UX audit - the-leak-finder is the one that compares your funnel against benchmarks. Read the skill.
Conversion: The PDP Reviewer
Reviews a product page you already have - using the page, its reviews, and real buyer questions - and returns a prioritized edit brief covering clarity, trust, proof, and unanswered objections. Run it before sending more traffic at a page that isn't converting. It reviews and prescribes; it doesn't generate new page code, which is the-page-shipper's job. Read the skill.
Demand: The Search Merchandiser
Turns your onsite search logs, including the zero-result queries, into specific merchandising fixes: synonym rules, naming corrections, and the collection pages you're missing. This is demand signal from people already on your site telling you what they want in their own words - considerably better than a keyword tool guessing. Read the skill.
Spend: The Spend Waste Finder
Triages paid spend across channels, campaigns, and audiences using ROAS, CAC, and spend concentration, to find what to cut before you add budget. It works one level up from the-leak-finder: that one diagnoses a specific page or funnel step, this one tells you which campaigns and audiences are burning money in the first place. Read the skill.
Just need the sample-size math, not the full test design? The AB Test Calculator does that one calculation in a browser tab, free, no account needed. The-hypothesis-engine is for designing the whole test - Thompson sampling, guardrails, exit criteria - not a replacement for the quick check.
The gap: retention has zero coverage
Acquisition, conversion, and activation are covered. Retention isn't. That's not an oversight we're glossing over - it's the honest state of a 10-skill pack, and it's worth naming plainly instead of stretching one of the existing skills to pretend it covers churn.
It's also the best-supported gap in this whole series. DataForSEO keyword-volume data pulled for this piece found "customer churn prediction" pulling 320 monthly searches at an $18.85 CPC - solid volume and strong commercial intent, stronger than most of what we found across the other six GTM roles. A churn-prediction skill is coming. It isn't built yet, and this post won't pretend it is.
The order to build them in
Start with the-leak-finder. Every test that follows should answer a question the leak-finder actually raised, not a hunch. From there, go to whichever funnel stage it points at: the-page-shipper or the-variant-router for a conversion problem, the-first-mile-mapper for an activation problem. Bring in the-hypothesis-engine once you have a real test worth running properly. Save the-price-point-finder for when pricing is the actual lever, not a default place to look because everything else feels tested out.
What to never hand over
Calling a test significant before it's actually reached its planned sample size. It's the single most common way testing programs fool themselves - peeking early, seeing a lead, and shipping it. The-hypothesis-engine sets the sample size and exit criteria up front for exactly this reason, and the discipline to wait it out is still on you.
The actual pricing decision. The-price-point-finder gives you the value metric, the tier structure, the analysis. Whether to actually ship that price change this week, given a renewal that's closing or a competitor move you know about and the skill doesn't, is a call a human makes with context no skill has access to.
When you outgrow the manual version
Intempt's Experience Optimizer agent runs this same work continuously against live traffic: CUPED variance reduction to reach significance faster without inflating false positives, and a one-click promotion from winning variant to permanent personalization rule. Same jobs, running with live guardrails instead of a spreadsheet you check by hand.
Frequently asked questions. Answered.
Six free Claude Skills for CRO and experimentation work: funnel leak diagnosis, activation mapping, landing page generation, pricing strategy, personalization routing, and Bayesian test design. Install with npx skills add sidchaudhary/gtm-skills.
