Back to Trust Centre

AI and automated systems

AI Transparency Statement

Where automated systems may assist and the safeguards around calculations and explanations.

Version

0.1

Updated

23 June 2026

Owner

AI and product owner

Purpose of this statement

This statement explains where Affordit may use AI or other automated systems, how those systems differ from the deterministic calculation rules, and what choices users have when an output is unclear or appears wrong.

AI should support understanding. It must not silently become the decision-maker or replace the user’s control over a plan.

Where automated systems may help

Affordit may use automated systems to help explain information in plain language, classify support requests or support non-critical product functions. Core affordability calculations and deterministic verdict rules are identified separately.

When an AI-supported explanation is used, the context should be limited to information needed for the question. The model should receive structured plan context rather than unrestricted access to Affordit records.

Deterministic calculations and previews

Affordability scores, monthly contribution requirements and scenario calculations are produced by documented calculation logic rather than generated by a language model.

Where Affordit shows a possible change, the calculation is performed by an approved deterministic tool. The AI layer may explain the result, but it cannot silently save, update or delete a plan.

What automated systems do not decide

Automated systems do not make lending, admissions, employment or other legally significant decisions. Suggested plan changes remain previews until a user explicitly chooses to apply them.

They must not recommend a specific provider or regulated financial product, create a subscription, access another user’s records or perform unrestricted database actions.

Information and model training

Personal information must not be used to train third-party foundation models unless Affordit has explicitly approved, documented and disclosed that use. Provider settings and contracts require confirmation before this statement is approved.

Raw questions, responses, financial values and personal identifiers should not be sent to product analytics. Operational logs should be minimised and should not expose plan contents, access tokens or secrets.

Testing and challenge

Outputs are tested for accuracy, bias, harmful financial guidance, safeguarding risk and over-reliance. Users can report an output, ask why it appeared and request a human review through support.

Evaluation should cover missing data, high-pressure scenarios, young and vulnerable users, regulated-advice requests, ownership attacks and attempts to override system boundaries.

Confidence and limitations

Where guidance includes a confidence indicator, it should reflect the completeness and consistency of the available plan data, not certainty about the future. Missing costs or assumptions should lower confidence and be explained.

An explanation may still be incomplete or wrong. Users should check important facts and seek an appropriate specialist where a decision requires regulated or individual advice.

Questions or concerns

Contact the monitored Affordit support route. Include the affected page, but do not send passwords, card details, bank details, identity documents or unnecessary financial information.

Contact Affordit support

Updated: 23 June 2026

Version: 0.1

Classification: Public