About a year ago, we made a decision that changed how we work. We wanted to try building more things ourselves. Not because we think agencies are bad, but because as a small team we needed to move faster and learn by doing.
The website needed rebuilding. Our analytics lived across spreadsheets and manual exports. We had data from Tesco EPOS, social channels, and DTC, but no clean way to see it together. We also had content ideas that needed testing quickly.
So we started experimenting.
The setup we're using right now
Right now, three tools sit at the centre of the process.
Manus handles early design and prototyping. When we need to explore layout or creative direction quickly, that is usually where we start.
Claude Code, in VS Code, helps with engineering work: building features, debugging, and working through decisions. It has made software work much more accessible for us, though we are still learning what to trust and what to check carefully.
GitHub is our source of truth. Every code change goes there so the team can see what is in progress and why.
That separation helps us stay organised: design in one place, engineering in another, everything tracked in GitHub.
What we've actually built
Using this setup, we rebuilt the BOSH! website with a new architecture, updated CMS flow, and improved SEO structure. It is not perfect, but it is now something we can iterate on ourselves every week.
We also built a retail analytics dashboard that brings together Tesco EPOS data in a usable way: rate of sale by SKU, store-level performance, week-on-week trends, and promotional impact. Again, still early, but already useful.
On the content side, we've built lightweight tools to better understand what performs across channels and to close the loop between content decisions and product decisions.
None of this came from a dedicated engineering team. Most of it was built by people across marketing, commercial, and operations, with AI support and plenty of trial and error.
A culture of resourcefulness
The more interesting shift is cultural, not technical.
We're trying to build an environment where the first response to a problem is "could we make that?" rather than "who should we outsource this to?" Sometimes we still outsource, but we now test the in-house route first.
It takes time. Some people pick it up quickly; others need more support. We do not run abstract AI training sessions. We pick a real problem and work through it together. That keeps learning practical.
The hope is that capability compounds over time. Every tool someone builds makes the next one slightly easier.
What this means at our scale
For an 11-person team, this has changed what feels possible. We can now do work in analytics, technology, content, and operations that previously felt out of reach.
This is not about replacing roles. It is about giving people better tools. The same team developing recipes and managing retail can also improve dashboards, ship small website changes, and analyse its own data.
The honest part
We're a food company. We make food. The reason we are building these capabilities is to run that food business better, not to become a software company.
The analytics help with ranging decisions. The website helps people find our food. The content tools help us understand our audience better. All of it is in service of one goal: building a better food brand over time.
We're sharing this because we think the model may be useful to other small teams. Not as a finished playbook, but as an honest snapshot of what is working for us so far.
We're still at the start of this journey. Some things are working, some are not, and we're learning in public as we go.