AI Labs · The machine room
Don't take my word for it. Run the machine.
Three real business pipelines — outbound, lead scoring, listing optimization. Run them here and read the reasoning, not a brochure. Each run says whether it came from the live model or a stored example; this is the room where the work happens.
The honest question
“Isn't this just ChatGPT with a good prompt?”
The model? Yes — and that's the point. Models are a commodity now; nobody should pay for access to one. What you're testing here is the thinking around the model: the pipeline, the structure, the reasoning it's forced to show.
And the part no chat window can give you: the production version runs without you — on every lead, every listing, every night at 2am — wired into your CRM and inbox, with QA gates and a human approval step. A chat answers when you ask. A system works while you sleep. The lab shows the brain; what you'd hire me for is the body.
Lab 01 · outbound
Cold Email Lab
See how AI writes outreach that sounds like you actually researched the buyer — because it did.
What actually happens
- 01
Understand
Reads what you sell and who buys it
- 02
Personalise
Anchors every line to the buyer's world
- 03
Sequence
Drafts Day 0 / 3 / 7 with one CTA each
- 04
Annotate
Explains why each choice was made
In production → this runs on every new lead automatically — researched, personalised, sent, and followed up from your inbox, with a human approving the first send.
zenji · idle — describe your business above and run the pipeline. Nothing is stored.
How to use this lab
- 1.Describe what you sell + who buys it, in one or two lines (role, company size, country help).
- 2.Hit Run — in seconds you get a 3-email sequence with the reasoning behind every line.
- 3.Run it again with your real product and compare against your current outreach.
Lab 02 · pipeline
Lead Scoring Lab
Watch messy inbound leads become scored, routed decisions — with the reasoning in the open.
What actually happens
- 01
Parse
Reads leads in any messy format
- 02
Score
Weighs role, intent, timing, fit
- 03
Route
Hot → call · Warm → nurture · Cold → archive
- 04
Explain
Shows which signals moved each score
In production → this sits on your inbound form and CRM — every lead scored and routed in seconds, your closers only ever see the hot ones.
zenji · idle — describe your business above and run the pipeline. Nothing is stored.
How to use this lab
- 1.Say what you sell, then paste 1-3 leads — any messy format is fine.
- 2.Hit Run — each lead comes back scored and routed (call / nurture / archive), with reasoning.
- 3.Try 2-3 real leads from your inbox and watch how the agent weighs them.
Lab 03 · e-commerce
Listing Optimizer Lab
The same class of system as Shakti — which lifted marketplace revenue 20–30% monthly. Try it on your listing.
What actually happens
- 01
Audit
Reads your title & description like a buyer
- 02
Research
Maps the keywords buyers actually type
- 03
Rewrite
Keyword-rich but human, trust markers added
- 04
Justify
Ties every change to ranking or conversion
In production → this rewrites entire catalogues on schedule, tracks rank movement, and re-keywords listings when competitors shift — Shakti did exactly this.
zenji · idle — describe your business above and run the pipeline. Nothing is stored.
How to use this lab
- 1.Paste a listing's title + description (naming the marketplace helps).
- 2.Hit Run — you get the rewritten listing, with every change tied to ranking or conversion.
- 3.Compare the keyword line against your live listing — the gap shows itself.
More machines being built — WhatsApp triage · research briefs · document extraction. A lab ships only once it genuinely runs.
Liked what the machine did with 4,000 characters?
Imagine it wired to your real CRM, inbox, and catalogue. A 30-minute call scopes exactly that.