What we offer.
Test business AI assistants using a defined set of prompts to investigate incorrect answers, inconsistent responses, unsafe assumptions and poor customer interactions.
AI agencies, chatbot developers and companies already operating customer-facing AI assistants.
What's included in the pilot.
- Up to 40 agreed test prompts spanning accuracy, consistency, limitations and customer-facing behaviour
- A repeatable test log recording expected results, observed outputs and reproducible issues
- A prioritized report of findings with severity, limitations and practical recommendations
- A handover discussion explaining what was tested and what remains untested
Pilot pricing & scope.
Pilot pricing applies only to a supported assistant, a defined testing scope and up to 40 test cases. Remediation, model/API fees, integrations, ongoing monitoring and additional rounds are quoted separately.
Introductory pilot estimates only. The exact scope, price, taxes, third-party costs and deliverables will be confirmed in writing before any work begins.
How the engagement works.
- 01
Define
Agree the AI system, permitted access, risks and representative customer scenarios.
- 02
Test
Execute documented test cases and capture outcomes where permitted.
- 03
Evaluate
Group findings by reproducibility, potential impact and severity.
- 04
Report
Provide actionable recommendations and the limitations of the assessment.
Important limitations.
This is a sample-based assessment, not an official AI safety certification, penetration test, compliance audit or guarantee of future model behaviour. Testing requires appropriate access and permission. Repeat runs may produce different outputs.
We assess each enquiry against available skills, technical requirements and applicable obligations before agreeing to undertake the work.
Discuss your requirements.
Tell us about the existing system, intended users and what a useful pilot would demonstrate.