AI Consulting & Automation

Most businesses don't need a moonshot AI strategy. They need the boring, repetitive work taken off their team's plate, and a clear answer to 'where would this actually help us?' That's what I focus on.

I start by listening — understanding your business, walking through your day-to-day processes, and identifying where AI can genuinely save time versus where it's just a distraction. Then I build and integrate the practical pieces: automations, LLM-powered features inside your existing tools, and the guardrails that keep them reliable. I build AI products myself, so the advice is grounded in what actually ships, not in slideware.

What's included

Process and workflow analysis to find high-value automation opportunities
Task automation — document handling, data entry, summarisation, routing
LLM and AI tool integration into your existing software and workflows
Custom AI agents with domain-specific prompts, rules, and guardrails
Choosing the right model for the job, balancing capability, cost, and privacy
Team enablement — getting your people confidently and safely using AI tools

How it works

Listen. I learn how your business actually runs and where the repetitive, time-draining work lives — before recommending anything.
Identify. Together we pick the opportunities with the clearest payback, and I'm honest about where AI won't help.
Build & integrate. I build the automations and integrations with the right guardrails, and test them against your real data, not a demo.
Enable & measure. I get your team comfortable with the new tools and we measure the time saved so the value is provable, not assumed.

Typical tech stack

Claude (Anthropic) · Local LLMs via Ollama · Go · Retrieval / vector search · Tool-calling agents · OpenSearch

Who it's for

Your team spends hours on repetitive copy-paste, lookups, or summarising.
You want to add an AI feature to your product but aren't sure how to do it reliably.
You've tried AI tools but can't tell what's genuinely useful versus hype.
Data privacy matters and you want to explore AI that runs on your own infrastructure.

Frequently asked questions

We're not a tech company — is AI automation realistic for us?

Absolutely. The biggest wins are usually in ordinary back-office work: handling documents, entering data, drafting replies, summarising long threads. You don't need a data-science team — you need someone to wire the right tool into your existing process.

How do you keep AI features reliable?

With guardrails: domain-specific prompts, validation of the model's output, and falling back to deterministic logic where correctness is non-negotiable. I've written about how differently models behave in practice — reliability comes from engineering around them, not trusting them blindly.

Can the AI run privately, without sending our data to a third party?

Yes. Where privacy is critical, I can build with locally-hosted models so nothing leaves your infrastructure — that's exactly what my Jeeves local-agent project explores.

Which AI models do you use?

Whatever fits the job. For most hosted work I reach for the latest Claude models from Anthropic; for private or cost-sensitive cases, local models via Ollama. The model is a means to an end, chosen per task.

Related work & reading

OOATS — Ad Campaign Platform — portfolio project
Jeeves — Local AI Agent Playground (v1) — portfolio project
Six Local Models, Six Behaviours: Lessons from Building Jeeves — from the blog
Senior Engineers + AI: Why the Combination Outperforms Both Alone — from the blog