AI tools built around your business
Strategy, prototyping and implementation for businesses ready to use AI where it actually helps.
What I build
Chatbots and assistants
"Your team answers the same questions repeatedly"
I build trained assistants that handle routine enquiries using your own content, product data and policies. They escalate to humans when the conversation gets complex and improve over time as you feed back on their responses.
- Customer service bots trained on your knowledge base
- Internal HR assistants for policy questions
- Sales qualification bots that book meetings
- Onboarding assistants for new starters
Content automation
"Content production is slow and expensive"
I build workflows that shift the bottleneck from creation to editing. AI handles the first draft, humans refine and publish.
Internal business tools
"Your data sits in systems nobody searches"
I build AI-powered interfaces that let your team ask plain-English questions across documents, databases and knowledge bases. Internal search that actually understands what people are asking.
Website AI features
"Your website treats every visitor the same"
Small changes that improve conversion without redesigning anything. AI works behind the scenes to personalise the experience based on what each visitor actually needs.
Technology and tools
OpenAI
Anthropic
Pinecone
n8n
Make
Supabase
Vercel
Netlify
Opinionated choices
I pick specific tools for specific jobs rather than defaulting to whatever is popular this week.
OpenAI and Anthropic for language tasks. OpenAI for structured output and function calling. Anthropic for nuanced analysis and long-context work. Most projects use both.
Pinecone for vector search. When your chatbot or internal tool needs to find relevant information from your own content, Pinecone stores and retrieves it fast.
n8n and Make for automation workflows. n8n when you need full control and self-hosting. Make when speed matters more than customisation.
Supabase, Vercel and Netlify for hosting AI tools. Supabase handles databases and auth. Vercel and Netlify deploy frontends and serverless functions with minimal infrastructure overhead.
How I work
Assessment
You describe the problem. I assess whether AI is the right answer or whether a simpler solution gets you there faster.
Prototype
A working proof of concept in days, not months. You see it work before committing budget.
Build
Integration with your existing systems. Tested, documented and ready for real users.
Refine
Prompt tuning and edge case handling as real usage patterns emerge. AI tools improve with use.
AI reality check
Where AI does not help
- Your data is messy, incomplete or unreliable
- Accuracy is critical and errors carry real cost
- Volume is low enough that manual processes work fine
- Users expect human interaction and trust matters
- Regulatory requirements prevent automated decisions
Where AI delivers value
- Repetitive tasks that follow patterns but need language understanding
- Drafting content that humans review and publish
- Searching and summarising large volumes of existing content
- Routing and classifying incoming requests or enquiries
- Extracting structured data from unstructured sources
Common questions
Do I need a strategy or can you just build something?
Both. Some clients know exactly what they want and I build it. Others need help identifying where AI fits. The assessment step covers this — if AI is not the right answer, I will say so.
Should we build custom tools or use off-the-shelf products?
Start with existing APIs and services wherever possible. Custom tools make sense when off-the-shelf options do not fit your workflow or when you need control over the data. I will recommend the simplest approach that solves the problem.
How much does this cost?
Implementation is quoted as fixed-price project work. Ongoing API costs depend on usage volume — typically between a few pounds and a few hundred pounds per month. I estimate both before work begins.
What about data privacy?
Every project starts with a clear understanding of what data goes where. Some providers offer private instances. Some use cases work with anonymised data. I match the approach to your sensitivity and compliance requirements.
Can AI replace our content team?
No. AI drafts faster but humans still review, edit and publish. The value is in reducing production time, not eliminating people. Fully automated publishing creates quality and accuracy risks that are not worth taking.
What happens when models change or costs increase?
I build with flexibility in mind. Switching between language models is straightforward when projects are designed properly. You are not locked into a single provider.
Ready to explore what AI can do for your business?
Describe what you want to achieve. I will tell you whether AI is the right approach and what it would take to build.