HomeHow it worksWorkflow SprintCase studiesBlogContact
Case studies

Systems we have built, run, and learned from.

These projects show the workflow, the authority given to agents, and the decisions that remain with a person.

Start a conversation See how it works →
What they show

Three systems with different jobs and risks.

Lobi Eco covers the daily work behind an Amazon FBA store. The trading system shows how authority can be separated in a high-risk workflow. Costfeed shows how agents can support the recurring work behind a data product.

Each case explains what the system handled, where a person approved the work, and what the project taught us.

The domain changes. Clear roles, approval boundaries, and logs remain essential.

Systems we run and systems we have built.

AllWorkflow systemsGovernanceWeb/product
Featured · 01Amazon FBA
Cloudcor operation · Multi-agent · Approvals · Content ops
Agent-assisted e-commerce operations

How Lobi Eco Runs Its Amazon Store With Agents

Lobi Eco uses specialised agents to prepare listings, content, creative, and ad changes. Consequential changes go to the operator for approval.

What it highlights
  • Workflow decomposition
  • Specialised agent roles
  • Human approval architecture
  • Cross-functional hand-offs
  • One system spanning content, optimisation, and growth
Why it matters

It shows how one owner can direct the store while agents prepare and carry out the recurring work.

Read case study →
Case · 02
Case 02System build
Governed autonomy

How a Trading System Separates Advice, Risk, and Execution

A governed trading system with separate roles for analysis, risk control, execution, observation, and market news. No single agent holds the full chain of authority.

What it highlights
  • Explicit control architecture
  • Layered supervision
  • Risk gating
  • Advisory vs. execution separation
  • Inspectable, governable autonomy

It shows how authority can be designed into the workflow before the system handles a high-risk decision.

Case · 03
Case 03Live product
Agent-built and -managed digital products

How costfeed.co Runs as an Agent-Assisted Data Product

Costfeed is a UK price-tracking platform. Agents support data collection, reporting, publishing, and routine site operations while a human operator handles decisions.

What it highlights
  • Live product operations
  • Recurring data workflows
  • Structured documentation
  • Reporting
  • Updates and maintenance

It shows the recurring operational work required to keep a data product current after launch.

Public · costfeed.coRead case study →
Shared method

What the three systems have in common.

The systems handle different work, but the design questions stay consistent. We define the workflow, divide responsibility, place approvals beside risk, and review what the system records.

Structure

the workflow clearly

Separate

responsibilities where it improves reliability

Define

where oversight belongs

Operate

with logs, owners, and ongoing review

  • Workflow-first
  • Role-aware
  • Approval-conscious
  • Operationally grounded
  • Designed for iteration
From proof to engagement

Start with one workflow.

The Agent Workflow Sprint maps one high-value workflow and turns it into a working first system with clear approval points.

Use the first version. Review the evidence. Decide what deserves to come next.

See the Agent Workflow Sprint Or read how Cloudcor works →

Which repeated workflow should we examine first?

Start a conversation Read the blog →