How We Use AI Inside Tiny Spark to Deliver More, Faster
By Sam St Aubyn
Putting what we've learnt into action
There’s a lot of noise around AI right now. Most of it focuses on what AI might do in the future. We’d rather talk about what it’s doing for us today, inside Tiny Spark, on real projects, for real clients.
We’re not an AI company. We’re a Bristol-based agency that builds websites, eCommerce stores, interactive experiences and digital tools. But over the past year, AI has quietly become part of how we work every day, not as a gimmick, but as a practical layer that makes our team faster, sharper and more focused on the parts of the work that actually matter.
Here’s how we incorporate it into our daily workload.
Turning Meetings Into Action Without the Admin
Every project has meetings. Most meetings produce actions that someone has to write up, assign, chase and track. That process is slow, and things slip through.
We’ve built an internal automation pipeline that takes recorded Teams transcripts, pulls out the key actions and decisions, and routes them to the right places for the team to review. A human signs off before anything is logged, but the legwork is done in seconds rather than hours.
The result is that our project coordination is tighter, our team spends less time on admin, and clients get faster follow-through on the things we discussed. It’s not glamorous, but it’s the kind of operational improvement that compounds over time.
Faster Briefs, Cleaner Scopes
One of the most time-consuming parts of any project is translating a requirements document into a structured plan. A PDF full of stakeholder notes, feature lists and competing priorities doesn’t automatically become a clear scope of work. Turning it into one used to take a senior person half a day.
We now pipe that kind of document through a workflow that reads it, identifies deliverables, flags ambiguities and drafts a structured breakdown ready for review. It’s not perfect out of the box, but it gets us 80% of the way there in minutes, and our team refines the rest.
For clients, this means less back-and-forth at the start of a project and a shared understanding of what we’re building from day one.
A Knowledge Base That Actually Gets Used
Studios accumulate knowledge fast. Past projects, solved problems, vendor notes, process docs. The problem is that this knowledge lives across shared drives, chat histories and individual memory, and most of it is effectively invisible when someone needs it quickly.
We’ve been building a self-hosted AI system that indexes our internal content and lets the team query it in plain language. Instead of searching through folders or asking a colleague, someone can ask a question and get a useful answer in seconds, with a source they can verify.
This is particularly useful when onboarding, when picking up someone else’s work, or when we need to check how we handled something similar on a previous project.
Speeding Up Development Without Cutting Corners
A significant chunk of our work is technical: building and customising websites on WordPress and BigCommerce, developing React front-ends, creating interactive experiences. AI-assisted development tools have become a standard part of how our developers work.
The value isn’t in AI writing code we don’t understand. It’s in reducing the repetitive parts of development: boilerplate, documentation, initial drafts of components that we then review and refine. Senior developers spend more time on the architectural decisions and the tricky edge cases, and less time on the mechanical parts of writing code.
For clients, this translates to faster turnaround on builds without sacrificing quality or oversight.
What We Don’t Let AI Do
We’re fairly deliberate about where AI sits in our process, and where it doesn’t.
Client strategy, creative direction and the judgment calls that define what makes a project actually good. Those stay with our team. AI is useful for processing information, reducing friction and accelerating execution. It’s not useful for having opinions, understanding context, or knowing what a client actually cares about.
We’ve also been careful about data. Our internal AI tooling is self-hosted, which means client information stays within our own infrastructure rather than feeding into third-party systems. That matters to us, and it matters to our clients.
The Practical Upside
If we’re honest about what AI has changed for us, it comes down to a few things: less time on low-value tasks, faster delivery on the things clients are waiting for, and a team that can take on more complex work without burning out on the coordination overhead.
None of it happened overnight, and it required real investment in building and testing these workflows properly. But the compounding effect of small efficiency gains across a team adds up quickly.
If you’re curious about how we approach this, or whether similar thinking could work inside your organisation, we’d be happy to talk.
