AI automation where it actually earns its keep
The useful question isn't "where can we use AI?" but "which repetitive, judgement-light work is costing us time?" That's where automation, with AI where it helps, pays off.
Where AI helps, and where it doesn't
AI is good at reading, summarising, classifying and drafting. It is not a substitute for clear processes, clean data or accountability. Many of the best automation wins involve no AI at all, just well-designed workflows. The rest use AI for a specific step, with a person reviewing anything that matters.
What I build
Document-aware assistants
Assistants that answer questions from your own policies, manuals or project documents using retrieval-augmented generation (RAG).
AI inside existing workflows
Classification, extraction or drafting steps added to CRM, helpdesk or approval flows.
Pre-sales and document automation
Turning enquiries, RFPs or forms into structured data and first drafts for review.
Process automation
Rule-based automation and integrations that remove manual copying between systems.
How an engagement runs
- Find the work: identify repetitive tasks with clear inputs and outputs.
- Prove it small: a focused pilot on real examples, with results measured against the manual process.
- Build it properly: integrate into existing tools, add review steps, logging and access control.
- Hand over: documentation and guidance on monitoring quality over time.
Recent work
I'm currently building an AI pipeline that turns client RFPs into requirement models, BRDs, implementation plans and cost proposals for review, and I use AI-assisted development daily in engineering work.
Questions
Is our data safe with AI tools?
Data handling is designed in from the start: which data is sent where, what is stored, and who can access it. Sensitive use cases can use providers and settings that don't train on your data.
Do we need a large dataset?
Usually not. Most business automation uses your existing documents and records with retrieval, rather than training a custom model.
What does a pilot involve?
A small, time-boxed build on real examples from one process, so you can judge the value before investing further.