We believe you deserve to understand the AI you're making business decisions with. This is our complete disclosure: the model we use, how answers are generated, our accuracy rates, our methodology, and our regulatory compliance. Last updated: April 2026.
This Transparency Centre covers 17 articles across 6 sections. It is the primary way we fulfil our obligations under EU AI Act Article 13 (transparency for limited-risk AI systems), and our own commitment to you as a user. Every article is reviewed quarterly. If something here is unclear or incomplete, email support@askbiz.co.
The technology behind AskBiz — which AI model we use, how we process your data, how answers are generated, and the limits of what the AI knows.
AskBiz uses Claude by Anthropic. Here's what that means — the model's capabilities, its limitations, how Anthropic handles data, and why we chose it.
Step-by-step: what happens between you asking a question and receiving an answer. Data retrieval, prompt construction, AI reasoning, confidence scoring, and output formatting.
Exactly which data from your business is included in AI prompts, what is never sent, how we minimise data exposure, and how prompt data is handled.
The AI's knowledge cutoff, what it can and cannot access, why it sometimes gives wrong answers, and how to work with these limitations productively.
How accurate AskBiz AI is, how we measure it, what causes errors, how to flag incorrect answers, and how flags improve the system.
What AskBiz's High, Medium, Low, and Estimate confidence ratings mean, how they're calculated, and how to use them to make better decisions.
What AskBiz's AI accuracy rates are across different question categories, the most common error types, what causes them, and how we track and reduce them.
How to report an AI answer you believe is wrong, what happens after you flag it, and how your flags improve AskBiz for everyone.
How AskBiz's core intelligence features work under the hood — Business Pulse scoring, anomaly detection, churn prediction, and export market scoring.
How the Business Pulse 0–100 score is calculated. The five dimensions, their weights, how each dimension is scored, and the statistical model behind the composite score.
How AskBiz detects anomalies in your business data. The statistical model, thresholds, seasonality handling, and why certain events trigger alerts.
How AskBiz's Churn Intelligence model identifies at-risk customers. The RFM model, machine learning approach, score calculation, and known limitations.
Data we publish openly about how AskBiz operates — accuracy metrics, enforcement statistics, law enforcement requests, AI model versions, and improvement logs.
Current AI accuracy rates across all question categories, confidence level distribution, flagging rates, and quarter-on-quarter improvement trends.
Statistics on AskBiz policy enforcement actions — warnings, suspensions, terminations, and appeals — published quarterly for accountability.
A running log of all significant changes to AskBiz's AI model, system prompting, confidence scoring, and business intelligence methodology.
How user feedback, flags, and usage patterns drive continuous improvement to AskBiz's AI — the feedback loop, what we do with flags, and how we test improvements before deploying them.
How AskBiz continuously improves AI accuracy through user feedback, automated benchmarking, and structured review cycles — and what role you play in making it better.
How AskBiz conducts product research and experiments — what we test, how we select participants, what data we use, and your right to opt out.
Our compliance position across key regulations — EU AI Act Article 13, GDPR, UK OSA — and our commitments to users about how we operate within these frameworks.
How AskBiz fulfils the transparency obligations of EU AI Act Article 13 for limited-risk AI systems. What we disclose, where, and how.
AskBiz's commitments to users about what we will always disclose, what we will never hide, and how we hold ourselves accountable.
If something in this Transparency Centre is unclear, incomplete, or you believe it is inaccurate, we want to know. Email us and we will respond within 2 business days.