AI Systems & Workcraft for Entrepreneurs | LachlanCB
48 capabilities · 6 gaps

The Receipts

The Receipts is a public capability record. Every skill sits on one of four tiers, and every claim carries the artifact behind it. If a claim cannot survive a follow-up question, it is listed as a gap instead of a skill.

I ship production software by directing AI tooling. That makes me an AI-native operator rather than an engineer, and the page says so plainly where it matters.

The tier scale exists because a normal skills list flattens everything into bullets. Having used a tool and having run it with money at stake are different claims. Only one of them belongs on a CV without a qualifier.

$4K → $30KMonthly revenue for a home-services client, in 12 months, as sole operator
Profitability on a seven-figure marketplace brand, in six months
16Live production systems I run and maintain
1,000+Cold calls placed personally, funnel instrumented end to end
How to read this · 04

Every row below can be opened, watched, or checked.

  • PractitionerDone repeatedly in production, with money or reputation at stake. I can own it and lead others through it.
  • WorkingDone end to end for real at least once. I can do it again unaided, sometimes with the docs open.
  • FamiliarUsed it and understand the model. I would need ramp time before owning it.
  • Won't claimNo real evidence behind it. Listed anyway, because the gaps are what make the rest checkable.
AI and automation · 10

The part I am actually hired for. All of it runs daily rather than sitting in a demo.

  • LLM APIs in production

    Practitioner

    Content engine across six brands, plus 30+ daily automations. See it

  • Prompt engineering and system design

    Practitioner

    300+ custom Claude Code skills, with hook-based self-correction turning failures into standing rules.

  • Workflow automation (n8n)

    Practitioner

    Production workflows, webhook routing, and a written record of every gotcha that cost me a day.

  • AI agent frameworks (LangGraph)

    Working

    Agent stack on my own VPS, exposed as run endpoints.

  • RAG and vector search (Qdrant)

    Working

    A live memory collection with semantic recall across the knowledge vault. See it

  • Voice AI (ElevenLabs)

    Working

    Eight production voice agents covering scheduling, intake and overflow.

  • Human-in-the-loop design

    Working

    Content pipeline with a deliberate approval gate before anything publishes. Not full auto, on purpose.

  • Data pipelines from wearables

    Working

    A webhook receiver processing 900+ transcript segments a day through AI summarisation.

  • Observability (Langfuse)

    Familiar

    Deployed and running. Light use so far.

  • Model fine-tuning and training

    Won't claim

    Never done it.

Technical · 10

What I can build and operate without an engineer, and the exact line where that stops.

  • AI-assisted delivery as the primary build method

    Practitioner

    Every product listed on this site was built this way, including the site itself. See it

  • Docker and self-hosted VPS operations

    Practitioner

    16 containerised services and 22 scheduled jobs on one production box, self-healing and self-managed.

  • REST APIs, webhooks and third-party integrations

    Practitioner

    Live integrations across CRM, CMS, automation, messaging and Search Console.

  • Next.js, React and Vercel

    Working

    Six production sites, including this one at roughly 150 routes. See it

  • Supabase and PostgreSQL

    Working

    Multi-tenant SaaS with row-level security enabled at table creation.

  • Infrastructure monitoring and alerting

    Working

    Uptime and performance monitors feeding an auto-generated dashboard every 30 minutes, failures alerting to Slack.

  • Git and GitHub, branch to PR to main

    Working

    Standard workflow across every repo I own.

  • DNS, Cloudflare and deployment

    Working

    Multiple domains across three hosts, with the failure modes documented.

  • Browser automation (Playwright)

    Working

    Site checks and render pipelines that run unattended.

  • Python and TypeScript

    Familiar

    I direct and debug both. I cannot read code unaided, and I will not pretend otherwise.

Growth and commercial · 12

Ten years of this would be the normal claim here. What I have instead is a small number of channels run properly, with the failures kept in.

  • Marketplace advertising and retail media

    Practitioner

    400 to 500 active campaigns, roughly half a million dollars in lifetime managed spend, peak 15× return.

  • P&L and unit economics ownership

    Practitioner

    Owned the KPI set on a seven-figure brand and doubled its profitability in six months.

  • SEO, technical SEO and structured data

    Practitioner

    This site: ~150 routes, page-level structured data on all but four, 3,000 impressions in month one on a new domain. See it

  • Local SEO and Google Business Profile

    Practitioner

    The core agency motion. Profile work plus suburb and service page architecture.

  • CRM and lifecycle automation

    Practitioner

    Full delivery stack: pipelines, email and SMS nurture, lead scoring, missed-call text-back.

  • Cold calling, setting and closing

    Practitioner

    1,000+ calls placed personally. A conversation every 10 to 15 dials, measured rather than estimated.

  • Content systems and publishing

    Practitioner

    45+ articles across three owned properties through one AI pipeline. See it

  • Client delivery management

    Practitioner

    A repeatable five-phase pipeline, tracked in Linear, handed over with an operations handbook.

  • Conversion optimisation

    Working

    Funnel and product-page work, landing builds, A/B tests, session replay. See it

  • Cold email and SMS systems

    Working

    Systems built to roughly 4,000 emails a month. The 5,000-email test returned close to zero replies, and killing it is the credential. See it

  • Paid search

    Working

    Client campaigns, run and reported.

  • Creative production

    Working

    Ad copy systems, video editing and graphic design. Enough to ship, not a design hire.

Measurement and attribution · 09

The most under-sold group on my CV. Most marketers stop at the tag manager.

  • GA4 and Google Tag Manager

    Practitioner

    Standard on every property I run.

  • Session replay and heatmaps

    Practitioner

    Used daily rather than installed and forgotten.

  • Dashboard design and instrumentation

    Practitioner

    Operations, sales and client-facing dashboards, generated live. See it

  • Server-side tag management

    Working

    Real implementation, not a tutorial. Rare in marketing hires.

  • Conversions API

    Working

    Server-side conversion forwarding in production.

  • Call tracking

    Working

    Necessary for local services, where the conversion is a phone call rather than a click.

  • Offline conversion imports

    Working

    Closing the loop from CRM outcome back to the ad platform.

  • Looker Studio reporting

    Working

    Client-facing reporting, refreshed automatically.

  • Mixpanel and Amplitude

    Won't claim

    Never used either. Same event-model thinking from GA4, but I would be learning the tool in week one.

Product and ways of working · 13

Product work done as the only product, design and build resource. That is a strength on scope and a gap on team process.

  • Customer discovery at volume

    Practitioner

    1,000+ direct conversations with the actual buyer, instrumented rather than remembered.

  • KPI definition and instrumentation

    Practitioner

    Every launch measured before it ships, not after it disappoints.

  • Killing your own work on evidence

    Practitioner

    A public kill log with the reasoning, the post-mortem and the rule carried forward. See it

  • Stakeholder and client communication

    Practitioner

    Sales through delivery through retention, as the only point of contact.

  • Async remote across timezones

    Practitioner

    Based in Thailand, working with Australian, UK and US clients. See it

  • Requirements, specs and acceptance criteria

    Working

    Written for my own builds. Not yet inside a team PRD process.

  • Roadmap and backlog prioritisation

    Working

    Phased delivery tracked in Linear. See it

  • Experiment design

    Working

    Field tests with the numbers published, including the ones that failed. See it

  • Agile delivery

    Working

    Phased and tracked. Not formal Scrum ceremonies, and I would not claim to have run them.

  • Line-managing engineers

    Won't claim

    I direct AI tooling. I have not managed an engineering team.

  • Jira and ClickUp

    Won't claim

    Not used. Linear and Notion daily.

  • Figma

    Won't claim

    I design through built components rather than in a design tool.

  • Scrum certification

    Won't claim

    No certification and no ceremony ownership.

The kill record · 02

The hardest thing to evidence is having shut your own project down.

I have killed a private-label product, a faceless YouTube channel, an agency positioning and a SaaS idea. Each one has a written post-mortem and a rule I carried into the next build. One of them cost about seven thousand dollars of my own money and taught me not to buy inventory before proving demand.

None of that is retold here, because it already has its own page. The reasoning is public, dated, and open to being checked.

Read the kill log
History · 06
  • Founder and growth operatorClearScale01/2025 – present
  • E-commerce brand managerAmazon aggregator, multi-brand portfolio11/2023 – 01/2025
  • Freelance e-commerce consultantMarketplace brand04/2024 – 06/2024
  • FounderOwn private-label brand03/2023 – 04/2024
  • Customer service consultantTelstra2020
  • EarlierLogistics, hotel guest services, hospitalitypre-2020

Diploma of Business, TAFE Queensland, 2020. Everything technical after that was self-taught and is listed above with its evidence.

Questions · 05

Want the long version?

The portfolio walks through the whole stack, including a 13-minute video. The systems themselves are on the apps page, the client work is on services, and the reasoning behind what I stopped doing is in the vault.