Responsible AI

Our approach

Eightech builds AI into products and client systems where it removes real work. We start from a business problem and judge the result by whether that problem got smaller — not by whether AI was involved.

This page sets out the principles we apply. It is written to be useful to someone evaluating whether to let an AI system near their operations.

Human oversight

AI systems we build are designed to stay under appropriate human supervision. In practice that means:

• A person remains accountable for outcomes the system affects.
• Consequential actions are designed for review or confirmation rather than silent execution.
• There is a route for a human to take over — our own product hands a conversation to a person with full context rather than trapping the customer with a bot.
• The level of autonomy is a decision made with the client, not a default we set.

Transparency

People should be able to tell when they are dealing with an automated system. We encourage clients to disclose AI use where it affects the people they serve, and we build the capability to do so into what we deliver.

Reliability and limitations

AI outputs can be wrong. Language models can produce text that is fluent, confident and incorrect. They can misread context and behave unexpectedly on inputs unlike their training data.

We treat that as a design constraint rather than a caveat: systems are scoped, tested and given safeguards appropriate to what a wrong answer would cost.

Data

Client data used with AI systems is handled under the terms of the engagement — see data processing. How data may be used, including whether it can be sent to a third-party model provider and on what terms, is agreed with the client before implementation rather than assumed.

Third-party model providers

AI systems we build commonly use models operated by third parties. Eightech does not own, build or control those models, and we do not present them as ours.

Their capabilities, limitations, terms and availability are set by their providers and can change. Where a provider's terms matter to your assessment, we will tell you which providers an implementation depends on.

Responsible use

We will not knowingly build AI systems designed to deceive people about who or what they are dealing with, to manipulate vulnerable people, to carry out unlawful surveillance, or to make consequential decisions about individuals without human review.

Use of our website is also governed by our terms of use.

Questions

If you are evaluating an AI implementation and need specifics about oversight, data handling or model providers, ask us: get in touch.