We build AI into products for a living, so we owe you a clear account of how we do it. These are practices, not certifications.
AI only where it earns its place
We use AI where it adds value, and we'll tell you when it's the wrong answer. Many good products need very little of it. When a simpler, more predictable approach works, we recommend that instead.
Humans stay in the loop
Nothing reaches a customer without a human check by default. Internal steps, such as classifying or routing, can run automatically. Anything that sends, publishes or promises something to a person outside your business is reviewed first. Autonomy is extended one step at a time, once the evidence shows a step is reliably right.
Every step is logged
Automated steps record what they received and what they produced, so decisions can be traced, evaluated and replayed. When something goes wrong, you can see where.
Evaluate before relying on it
We test AI behaviour against realistic cases before launch, not after customers find the gaps. The evaluation is designed around what "good" means for your use case.
Choosing models
We work with models from several providers, including OpenAI, Anthropic, Google and Mistral, and choose per task based on quality, cost and data requirements. No single-vendor lock-in by default.
Data
We use only the data a project needs, and agree how it will be handled before work begins. [to confirm: commitment on client data and model training, e.g. "we never use your data to train models without your written agreement". Owner to confirm]
Honest with the people who use it
We design products so people can tell when they're dealing with AI, and can reach a human when it matters. [to confirm: owner to confirm this as a standing commitment]
No manipulation
Engagement features are designed around motivation (autonomy, competence and relatedness), not tricks. No dark patterns, and no rewards bolted onto everything to keep people hooked.
Regulation
Rules for AI are developing quickly, including the EU AI Act. We factor the rules that apply to each project into its design from the start. [to confirm: lawyer/owner to confirm the regulatory wording]
Questions
If you have questions about how we'd use AI in your project, email mohammad.ashraf@nuvai.qa.