Answers are inconsistent
Ask the same question twice and get two different answers, in two different tones.
AI prompt engineering
AI that answers the way your organisation would.
I design, test and refine the instructions behind your AI tools, so assistants, chatbots and workflows give accurate, on-brand and safe answers, every time.
In short
What it is Designing and testing the instructions that make your AI tools accurate, consistent and safe.
Who it's for Medium-sized and larger organisations using AI with customers, clients or staff.
What it costs A fixed price, agreed after a short call. All prices exclude VAT.
Who it's for
Does this sound familiar?
It makes things up
Your AI tool sounds confident, but invents dates, prices or policies.
It doesn’t sound like you
Replies feel generic or robotic, not like your organisation.
Worried about risk
Sensitive topics, personal data and compliance need careful handling.
Use cases
Where prompt engineering makes the biggest difference
Examples from different industries where accuracy, tone and safety really matter.
Hospitality
Voice booking assistant
Answers the phone for a busy restaurant, takes and changes bookings, handles dietary questions, and passes large parties or complaints to a member of staff.
You get Voice prompts tuned for natural conversation, with clear booking rules.
Healthcare
Patient enquiry assistant
Handles bookings and practical questions for clinics, never gives medical advice, and hands over to a person when needed.
You get Safety rules and hand-over triggers, tested on real enquiry types.
Legal & professional services
Document summaries and first drafts
Summaries of contracts, reports and proposals in your house style, with references back to the source.
You get Prompts that follow your templates and flag anything uncertain.
Retail & e-commerce
Product content at scale
Consistent product descriptions, FAQs and customer replies across hundreds of products, all in your brand voice.
You get Brand-voice prompts and checks for accuracy and tone.
Education
Student and parent support
Answers questions about courses, deadlines and policies, and passes wellbeing or safeguarding concerns straight to staff.
You get Tested prompts, escalation rules and approved answers.
HR & internal teams
Staff policy assistant
Answers staff questions from your own handbook and policies, so people get the right answer first time.
You get Prompts grounded in your documents, with clear limits.
How I engineer prompts
More than clever wording: a tested, documented system
Every prompt I build has five parts, and none goes live until it has passed a test set built from your real questions.
- Role and goal Who the AI is, who it’s talking to and what a good answer looks like.
- Context and knowledge Grounded in your own documents and facts, so it doesn’t have to guess.
- Rules and guardrails What it must never do, what it must always check, and when to hand over to a person.
- Format and tone Your tone of voice, length and layout, so every answer looks and sounds right.
- Testing and scoring Real questions and tricky edge cases, scored for accuracy, tone and safety.
Example test scorecard Illustration only, not a real client
Customer service assistant · 40 test questions · first draft → final prompt Accurate answers29/40 → 39/40
Right tone of voice24/40 → 40/40
Safe hand-over to staff31/40 → 40/40
Correct format22/40 → 40/40
No made-up facts34/40 → 40/40
What's included
What you get
Every quote lists exactly what's included, so there are no surprises.
Discovery workshop
We agree what your AI should do, who it serves and what must never go wrong.
Engineered prompts
Clear, structured prompts for each task, written for the AI tools you use.
Test set and scorecard
Your real questions and edge cases, with results before and after.
Guardrails and hand-over rules
Clear limits for sensitive topics, personal data and escalation to staff.
Prompt library
Every prompt documented and version-labelled, so your team can maintain it.
Team handover
A session for your staff on how the prompts work and how to update them safely.
How it works
Discover, design, test, hand over
- 1 Discover We look at where you use AI, what it gets wrong and what good looks like.
- 2 Design I write structured prompts with the right context, rules and format.
- 3 Test Each prompt is tested against your real questions and refined until it passes.
- 4 Hand over You get the prompts, the test set and the know-how to keep them working.
Why work with me
Experience where it counts
- Real-world experience I’ve designed AI assistants for private healthcare, where accuracy and safety really matter.
- Tested, not guessed Every prompt is measured against your real questions before it goes live.
- Safe by design Clear guardrails, careful handling of personal data, and people in charge of sensitive decisions.
FAQ
Questions owners ask
Can't see yours? Ask my AI assistant, or send me a message.
Message me on WhatsAppIt’s the craft of writing and testing the instructions that tell an AI tool how to behave: what it knows, what it must never do, and how it should answer. Good prompts turn a clever tool into a reliable one.
The main ones organisations use, such as ChatGPT, Microsoft Copilot, Google Gemini and Claude, as well as chatbots built on them. The approach works across tools.
With a test set of real questions and tricky edge cases. Each answer is scored for accuracy, tone, format and safety, before and after the prompts are improved.
Yes. Many projects start with a review of existing prompts, then focus on the problems that matter most.
No. Everything is documented in plain English, and the handover session shows your team how to make safe updates.
I design prompts to use only the information they need, and agree clear rules for personal and sensitive data with you from the start.
Discovery call
Make your AI tools reliable.
Tell me where you use AI today, and I'll suggest how to make it more accurate and safer.
