AI Transparency Statement
Effective Date: October 10, 2026
1. Overview and Scope
This AI Transparency Statement explains how RingFuze Technologies LLC (“RingFuze,” “we,” “us,” or “our”), doing business as Calladhere, uses artificial intelligence (“AI”) in the Calladhere platform (the “Service”). It describes what our AI features do, which providers power them, how humans stay in control, the limitations you should understand, and how we handle the data involved.
Calladhere is a business-to-business compliance and quality assurance platform for U.S. Medicare and ACA insurance agencies and their partners. Our AI features are designed to support, not replace, the judgment of your compliance, quality assurance, and management staff.
This Statement supplements our Terms of Service, Privacy Policy, and, where applicable, the Business Associate Agreement described on our HIPAA & Business Associate Agreement page. If there is a conflict, your signed agreements with us control.
2. How Calladhere Uses AI
Depending on your plan and configuration, the Service uses AI for the following purposes:
Transcription. Call recordings that you upload or sync are converted from speech to text, including speaker separation, so they can be reviewed and audited.
Audit scoring. A first-pass AI model evaluates each transcript against the audit questions and policies your organization configures (for example, CMS and TPMO-style disclosure requirements) and identifies potential violations, along with supporting excerpts and explanations.
Second-model verification. A separate AI model independently re-checks audit findings without seeing the first model’s conclusions (a “blind” verification pass). Where the two models disagree, the finding is flagged as uncertain for human review.
Assistants. In-app AI assistants help authorized users ask questions about calls, audits, and compliance results, and draft content such as question wording or summaries.
Reporting agent. Where available, an AI reporting agent helps users explore audit and performance data and generate reports using natural-language requests.
Form import. When you upload a PDF or image of an existing form, AI converts it into an editable form template that you can review and adjust before use.
Data extraction. AI extracts and checks structured details mentioned on calls (for example, whether information stated on a call is consistent with the customer record) to support audits and data quality.
Knowledge base search. Documents you add to your organization’s knowledge base may be converted into numerical representations (embeddings) so that AI features can locate relevant passages when answering questions.
3. AI Providers We Use
We build our AI features on commercial AI and speech-recognition services offered by established providers, including speech-to-text providers for transcription and large language model providers for audit analysis, verification, assistants, reporting, extraction, form import, and knowledge base search. We access these providers through their commercial API offerings.
These providers act as our subprocessors and are bound by written agreements and confidentiality obligations. We notify customers before a new subprocessor begins processing Customer Personal Data, as described in our Data Processing Addendum.
4. Human Oversight and Customer Control
Calladhere is built so that people, not models, make the final call on compliance outcomes.
Uncertain flags. Findings where our verification pass disagrees with the first pass, or where the evidence is unclear, are marked as uncertain so your reviewers can prioritize them.
QA review. Your QA reviewers, managers, and administrators can open any audit, listen to the recording, read the transcript, and confirm or dispute each finding.
Reversals. Authorized users can reverse AI findings they believe are incorrect. Reversed findings are recorded so that reports reflect the human decision.
Customer-configured rules. Your organization decides which audit questions apply, how they are worded, and how strictly or leniently each question is evaluated. The AI applies your configuration; it does not decide on its own what your compliance standards should be.
Role-based access. Access to AI features and their outputs follows the roles and permissions your administrators assign.
5. Calibration and Continuous Improvement
When your reviewers confirm or reverse AI findings, that feedback helps us improve the Service. We use reviewer feedback to refine our audit prompts, instructions, and question wording so that future audits better reflect how human reviewers interpret your rules. This calibration changes the instructions we give to AI models; it does not retrain the underlying models.
No foundation-model training on Customer Data. We do not use Customer Data to train our own or third-party foundation models, and we use AI providers under commercial/API terms that do not permit them to use Customer Data to train their models.
6. Accuracy, Limitations, and Known Failure Modes
AI output can be wrong. Transcripts may contain errors, and audit findings may miss real issues (false negatives) or flag issues that did not occur (false positives). Common causes include:
Transcription errors. Misheard words, names, numbers, plan names, or Medicare-specific terms can change the meaning of a statement.
Accents and speech patterns. Strong accents, fast speech, mumbling, or speech in languages other than the one expected may reduce transcription accuracy.
Crosstalk. When speakers talk over each other, words may be dropped or attributed to the wrong speaker.
Audio quality. Background noise, low volume, compression, dropped segments, hold music, or truncated recordings can produce incomplete transcripts.
Ambiguous phrasing. Agents may paraphrase required disclosures, split them across the call, or use wording that reasonable reviewers could interpret differently.
Missing context. A single recording may not capture everything relevant (for example, disclosures given on a separate call or in writing), and the AI only evaluates what it is given.
Configuration effects. Question wording and strictness settings directly affect results. Vague or conflicting questions produce less reliable findings.
Accuracy also varies by call type, recording source, and configuration. We monitor quality, but we do not guarantee that any AI output is complete, accurate, or suitable for a particular purpose.
7. Not Legal Advice
Calladhere is a decision-support tool. AI findings, scores, summaries, suggested question wording, and reports are not legal advice and are not a determination that any law, regulation, CMS guideline, or carrier requirement has or has not been violated. You remain responsible for your compliance program and for interpreting the requirements that apply to your business. Consult qualified counsel or compliance professionals for legal questions.
8. Customer Responsibilities
By using Calladhere’s AI features, you agree to use them responsibly, including by:
Reviewing before adverse action. Having a qualified person review the recording, transcript, and finding before taking disciplinary, compensation, termination, chargeback, or other adverse action against an agent or employee based on AI output.
Complying with AI and employment laws. Considering whether laws governing automated decision-making, AI use in employment, workplace monitoring, or consumer privacy apply to your use of the Service, and meeting any notice, assessment, or opt-out obligations they impose.
Configuring thoughtfully. Keeping your audit questions, policies, and strictness settings accurate and current.
Providing notice and obtaining consent. Giving any required notices to, and obtaining any required consents from, call participants and your workforce. See our Call Recording & Consent Notice.
Using outputs appropriately. Following our Acceptable Use Policy and not relying on AI output as the sole basis for decisions with significant effects on individuals.
9. Quality Evaluation
To measure and improve accuracy, authorized RingFuze personnel may create evaluation test cases (fixtures) from selected audits and use them to compare AI results against human-reviewed outcomes. These fixtures are used solely for quality assurance and accuracy evaluation of the Service, under access controls, and are handled in accordance with our agreements with you, including any Business Associate Agreement.
10. Model and Provider Changes
AI technology changes quickly. We may change the AI models, model versions, or providers used for any feature, for example when a provider retires a model, when a newer model performs better in our evaluations, or for reliability, cost, or security reasons. Changes may affect results, including scores and the wording of explanations.
We evaluate significant model changes before rolling them out broadly. When a change involves a new subprocessor, we will notify customers in advance as described in our Data Processing Addendum.
11. Data Handling and Protected Health Information
Call recordings, transcripts, and customer record fields processed by our AI features may include protected health information (“PHI”) about insured individuals. We process this data on your behalf, as your service provider and business associate, to deliver the Service. Where you have entered into a Business Associate Agreement with us, PHI is handled as described in that agreement and on our HIPAA & Business Associate Agreement page.
AI providers receive only the data needed to perform the relevant feature, such as audio for transcription or transcripts and relevant record fields for audit analysis. Data is transmitted over encrypted connections (TLS). For more about our safeguards, see our Security Overview. For details on how personal information is collected, used, and retained, see our Privacy Policy and Data Processing Addendum.
12. Feedback and Reporting AI Issues
We want to hear about AI results that seem wrong, biased, or harmful. The fastest way to correct an individual finding is to dispute or reverse it in the app, which also feeds our calibration process.
To report a broader concern, such as a pattern of incorrect findings, a transcription problem affecting many calls, or output you believe is inappropriate, contact support@calladhere.com. For regulatory or compliance concerns about our AI features, contact compliance@calladhere.com. Please include the affected audit or call identifiers (not raw PHI in the email body, where possible) so we can investigate.
13. Contact Us
If you have questions about this AI Transparency Statement, please contact us:
General support: support@calladhere.com
Compliance and HIPAA: compliance@calladhere.com
Privacy: privacy@calladhere.com
RingFuze Technologies LLC (Calladhere)
3770 Shaw Ave, Suite 102-226
Clovis, CA 93619
USA