Four areas of work, from training to product
They can be booked separately or in sequence. Most people start with one and expand once it works.
AI training and adoption
Buying licences is not adopting AI. I work with the people: training by job role, internal champions, short lessons and an honest measurement of what is being used and what is not.
Who it is for
Teams with AI licences and little real usage.
Weekly active use
The main indicator
Time recovered
Measured per task
Champions trained
One per area
What it includes
Concrete examples
Related package
8-week adoption programme
From underused licences to measured weekly usage, with internal champions who can keep it going.
- Initial maturity and tooling assessment
- General training for the whole organisation
- Workshops for specific areas
- Initial prompt library
- Weekly lessons and internal newsletter
- Coaching sessions for the champions
- Adoption measurement and final report
Internal knowledge and prompts
A company's knowledge lives in manuals, emails, procedures and in a few people's heads. I turn it into assistants, search and libraries that answer with the source attached.
Who it is for
Organisations with plenty of documentation and no good way to query it.
Answers with a source
Always cited
Search time
From minutes to seconds
Reusable prompts
By role and process
What it includes
Concrete examples
Related package
Internal AI portal
One place where every employee finds prompts, assistants, training and use cases.
- Internal portal with search
- Prompt library by department
- Use case catalogue
- Assistant connected to your documentation
- Training and lessons section
- Usage metrics and most-consulted content
Agents and automation
I build agents that do real work, connected to the systems you already run, with human review where there are consequences and the cost per conversation in plain sight.
Who it is for
Operations, customer service and back office with high volume.
Tasks without intervention
Measured by volume
Response time
From hours to seconds
Cost per conversation
Monitored
What it includes
Concrete examples
Related package
Pilot agent
One specific case in limited production, with metrics to decide whether to expand it.
- Case selection and scoping
- Conversation and flow design
- Integration with one data source
- Working prototype
- Testing with real cases
- Basic metrics dashboard
- Recommendations for going to production
Software development with AI
I build custom applications, integrations and assistants, with the practices that keep software alive: specification, review, tests, observability and maintenance.
Who it is for
Product and engineering teams with a specific problem.
Pilot in production
In about 4 weeks
Automated evaluations
From the start
Cost and latency
Instrumented
What it includes
Concrete examples
Related package
From idea to pilot in four weeks
Something working with real users before committing a larger budget.
- Week 1 · Discovery and specification
- Week 2 · Clickable prototype and base integration
- Week 3 · Building the full case
- Week 4 · Evaluation, observability and handover
- Code, documentation and a plan for what comes next
Work with a fixed scope and timeline
Four formats with defined deliverables, so you know what you get before signing. If you need something else, we adjust from here.
Opportunity assessment
Knowing which processes AI helps with, and which it does not, before investing in any of them.
- Interviews with the areas involved
- Current process map
- Use case catalogue
- Prioritisation by impact, effort, feasibility and risk
- Six-month roadmap
- Impact and cost estimate
At the end · A roadmap you can make decisions with.
Copilot adoption programme
Going from underused licences to measured weekly usage, sustained by people inside.
- Initial maturity assessment
- General training for the whole organisation
- Hands-on workshops by area
- Corporate prompt library
- Weekly lessons
- Internal newsletter
- Coaching sessions
- Adoption measurement
At the end · Measured adoption and the internal capacity to sustain it.
Pilot agent
One case working with real users, and data to decide whether to expand it.
- Case selection and scoping
- Conversation and flow design
- Integration with one data source
- Working prototype
- Testing and evaluation
- Basic metrics dashboard
- Recommendations for production
At the end · An agent in limited production, measured and documented.
Monthly support
Having someone looking after the company's AI without building a department for it.
- Monthly internal newsletter
- Short training lessons
- Employee support
- Prompt library maintenance
- Idea generation and prioritisation of new cases
- Review of tools on the market
- Usage best practice
- Building small agents
- Adoption tracking
At the end · AI capability sustained month to month.
Shall we talk about your case?
Tell me which process takes the most time. In half an hour I will tell you whether AI solves it, how I would approach it and what it would take.