AI that does the work. You stay in charge.
Agentic systems for small sustainable teams that want to get ahead, built by someone who cares about your product as much as the model.
AI Solutions
For teams whose work runs on ratings, reports and rules.
Approval queue Example
Supplier questionnaire Waiting for you
Footprint update Waiting for you
Client email Approved
1 AI drafts The repetitive part: reading, checking, writing the first version. 2 You approve A person reviews every output. Nothing skips this step. 3 Then it goes Only approved work is sent, filed or published.
Send automatically
AI Solutions is one of three areas. It sits alongside ESG and founder operations, and most of my work crosses between them.
Is this you?
You know AI could save you hours But no-one on the team has the time to work out where to start.
Start with an audit You tried a chatbot and it didn’t stick A generic tool, no process around it, and nobody owning it.
Team rollout Your work is rules-heavy Ratings, reports and questionnaires: exactly what a well-built agent is good at.
Agentic workflows
Workflow audit Where the repeated hours are, and which are worth automating.
Agentic workflows Multi-step systems with a person approving every output.
Team rollout Prompt library, SOPs and training so the whole team uses it.
ESG data automation Reporting and assessment logic kept current, with a person signing off.
Prototypes & internal tools Clickable products and tools, shipped to a real URL.
Governance & footprint EU AI Act duties, data handling and the carbon cost of your AI use.
How it works
1 Audit Map the work and find the repeated hours.
2 Pilot One workflow, human-approved, on real data.
3 Embed SOPs, prompt library and training.
4 Govern Review, footprint and a roadmap.
Confidential · Own build · 2026
An AI system that runs a rules-heavy sustainability workflow, with a person approving every output.
Deliverable
Proposal and business case; the demo stays private under confidentiality.
See the build
Own build · 2026 – now
An offline-first workout tracker, live on the web. I wrote the spec and made every call; AI wrote the code.
See the build
Own build · research system · 2026 – now
A reusable way to make AI a standing researcher: tiered sources, graded evidence, a knowledge base that grows instead of resetting.
Deliverable
Claude skill plus two live knowledge bases.
See the build
Own build · tools · 2026
The tools I built so my own consulting runs faster: Claude skills for SEO/GEO audits and client onboarding, a visual sitemap tool, and a design reference library.
Deliverable
Claude skills and two internal tools.
See the build
Resting Reef · Founder’s Associate · freelance · 2026 – now
Rebuilt a marine start-up’s sales-funnel data and led its AI rollout, as Founder’s Associate.
Deliverable
Tracking spec, CRM build guide, reporting dashboard, prompt library and SOPs.
Read the case study
All AI Solutions projects
Filter the projects
Engagements
Three lengths. Each ends with something you own.
3 months 6 months 12 months
Focused project Prove it works on one real process.
What you get
Workflow audit One agentic workflow, human-approved Team trained on it
You end with
One process that takes minutes, not hours.
Questions
Is our data safe? Systems run in your own accounts, with providers whose terms exclude training on your data, and nothing goes out until a person approves it.
Do we need technical people? No. I build it, document it and train the people who’ll use it.
Why a sustainability specialist? Because the hard part isn’t the model. It’s understanding the ratings, reports and rules your business runs on.
What does it cost to run? We size the running cost in the audit, before anything is built, so you know it up front.
Know your numbers. Answer the question that just landed.
Carbon baseline Example
Scope 1 · Direct 9%
Scope 2 · Energy 14%
Scope 3 · Value chain 77%
Example data · GHG Protocol · DEFRA factors
Hand it over. Get it back done.
Your handover Example
Learn how you work
Take one thing off your plate
Write it down as SOPs
Take the next thing
Let’s look at your situation.
A 30-minute call about the work you’d automate. You’ll leave with a clear next step, whether or not it’s with me.