Customer Stories|Fuel retail and distribution
Workflow Automation Using AI
Our engineers join your team’s workflow, follow real work from start to finish, and build around what they find, with your owner deciding every rule.
Every business has a way of working that only its own people fully understand: the file that always arrives late, the exception everyone knows about, the rule that lives in one person’s head. That is where we start.
We work as forward-deployed engineers. Instead of handing you a product and a manual, our engineers join your team’s workflow in a shared channel, follow real work from start to finish, and build around what they find, in short loops, with your owner deciding every rule.
Leading AI models (e.g. Claude, OpenAI) help us write, test and check the software that takes the repetitive matching and checking off your team. MCP connectors let approved AI assistants read your systems safely, read-only. And we hand over skills, the packaged know-how the AI follows, and agents that do multi-step work and pass anything that needs judgment to a person. The stories below come from our work with fuel retail and distribution businesses.
Three customer stories
From the till to the bank, from the tank to the monthly report, and from the supplier invoice to the debit file.
Fuel retailFuel sales checking
Every station checked from till to bank, every day. No more afternoons lost chasing a gap.
Read the Story
Fuel retailFuel stock control
Fuel losses caught the day they happen, not at month end. A drifting tank opens its own investigation.
Read the Story
Any industryBilling & Payments Operations
Month-end in one flow: invoices, fees, incentives and bank debits checked before any money moves. No more re-typing.
Read the Story
How an engagement works
Five steps, from the first real day with your team to the next short cycle.
Discover your workflow
We follow one real day, or one real month-end, from start to finish with the people who do the work. We collect sample files and write down every rule, including the unwritten ones.
Embed engineers
A small team of our engineers joins yours in a shared channel. Each issue gets its own ticket. When a rule is unclear, your owner decides and we record the decision.
Build and connect
We build in short loops with AI coding agents (e.g. Claude, OpenAI). MCP connectors link AI to your systems, read-only first, with a separate access key for each assistant and a log of every call. Each change goes to a test copy first, passes its tests and an independent AI review, and goes live only when you say yes.
Hand over skills and agents
We package your team’s know-how into skills, and give you agents that handle routine multi-step work and pass anything that needs judgment to a person.
Run and improve
Watchdogs flag late or failed data, and your team’s requests go into the next short cycle.
Glossary
The few terms these stories use, in plain words.
- Forward-deployed engineering
- Our engineers work inside your workflow, not from a specification handed over a wall. They learn the real process with your team, your owner decides the rules, and nothing goes live without your yes.
- MCP (Model Context Protocol)
- An open standard for connecting an AI assistant to your systems through a controlled connector. The connectors in these stories work like a viewing window: the AI can look up what it is allowed to see, but cannot change anything. In the distributor finance story, each assistant also has its own access key, and every call is logged.
- Skill
- Packaged know-how (written steps, rules and checks) that an AI follows, so it does a task your way, every time.
- Agent
- An AI that carries out a multi-step task on its own, using only the tools and skills it is allowed, checking its work and handing over to a person when judgment is needed.
Ready to build AI into your workflow?
Tell us how your team works today. We will come back with a plan and a quote.