I build agentic systems and analytics for small businesses.

This site walks through one real engagement: a solo-built platform for a DTC beverage brand that automates customer service, delivery exceptions, fulfilment, and marketing-spend decisions — end to end, in production. Every workflow below is a scripted replay of the real system; nothing here is a live agent or real customer data. See under the hood for why.

Case study: a DTC beverage brand

A small, solo-run online beverage business needed customer support, fulfilment, and marketing decisions handled without hiring a team. The platform below is what got built: five agentic workflows, a marketing-mix model, and the engineering discipline (testing, incident response, evaluation) that keeps it trustworthy in production.

Agentic workflows

Customer service email

Refunds and returns, classified and resolved automatically — with an honest look at what happens when confidence is low.

Delivery exception auto-resolution

Shipping problems detected, customers notified, reshipments created — no human touch on the happy path.

Sample-pack zero-touch fulfilment

From nineteen identical manual requests to a fully automated intake-to-invoice flow.

Self-modifying playbook

A plain-language Telegram message becomes a committed, deployed policy change — or a filed issue if it needs real code.

Evaluation framework

How the agent's behaviour gets tested before it ever reaches a customer.

Analytical use cases

Inventory-stretch via marketing throttle

Using the marketing model backwards: finding the spend that slows demand to match supply.

ROAS optimisation

Reallocate spend across channels interactively and compare against the model's own optimum.

The Solver

Try the interactive spend-scenario planner — built on the real fitted marketing-mix model.