Question 1

Does Heravox store raw customer feedback?

Accepted Answer

No. Raw text is processed only transiently, in memory, for a single purpose: detecting and removing personal data. Once anonymization completes, the raw text is discarded — only the anonymized version is written to durable storage. The pipeline is fail-closed: if a record cannot be anonymized, it is not processed or published at all. Logs never contain raw text or personal data either.

Question 2

Is our data sent to third-party AI models?

Accepted Answer

Only in anonymized form. Anonymization is performed by Heravox's own Turkish-language NLP model — developed in-house and running entirely on our own infrastructure, with no third-party API involved. Complementary rule-based and machine-learning layers detect personal data: national IDs, phone numbers, emails, IBANs, names, addresses, sector-specific identifiers such as policy or claim-file numbers, and sensitive health information. Only anonymized text reaches the LLM classification layer, which runs on an enterprise-grade managed cloud AI service whose provider does not retain submitted content and does not use it for model training. Identifiable personal data is never transferred abroad.

Question 3

Where is our data stored?

Accepted Answer

On the standard Cloud deployment, everything that could relate to your customers is processed and stored on infrastructure hosted in Türkiye, operated by an ISO 27001-certified provider. Each organization runs in its own isolated tenant. Data is encrypted in transit (TLS 1.2+) and at rest (AES-256), and backups are encrypted and kept in Türkiye as well. On Hybrid and On-premise deployments, storage lives entirely inside your environment — you control residency, retention, and access.

Question 4

How does Heravox handle KVKK, GDPR, and the EU AI Act?

Accepted Answer

Compliance is built into the data path, not bolted on. Anonymization is a precondition of every processing step, which satisfies KVKK's data-minimization and anonymization principles — including Article 6 special-category data — and means identifiable personal data is not transferred abroad; the same principles map to GDPR. For the EU AI Act, Heravox maintains an AI usage inventory covering every AI component of the platform: analytical outputs are advisory, no automated decisions are made about individuals, and AI-generated content is always labeled as such. Policies, the AI inventory, and data-processing agreements are available to customers.

Question 5

How is our data isolated from other customers?

Accepted Answer

Physically, not just logically. Each brand's data lives in its own dedicated database — two customers' records never share a table — and cross-tenant access is blocked at the application layer. Platform access is protected by multi-factor authentication and least-privilege, role-based authorization. Per-brand encryption keys make the right to erasure real: destroying a brand's keys renders its encrypted data permanently unreadable.

Question 6

Can you support our vendor security assessment?

Accepted Answer

Yes — it is a standard part of our enterprise onboarding. We complete security questionnaires (including CSA CCM-based assessments) and share our information-security policy set, AI usage inventory, and independent penetration-test summaries under NDA. The platform has been penetration-tested by an independent firm, and further operational detail — audit logging, release integrity, hardening — is shared as part of the assessment process. Our infrastructure provider is ISO 27001 certified, and Heravox's own ISO 27001 certification is on the roadmap.

Question 7

Can Heravox be deployed on-premise or in a hybrid setup?

Accepted Answer

Yes. Cloud is the standard managed deployment — zero setup on your side, working on publicly available content, hosted in Türkiye. Hybrid keeps raw data and the dashboard in your environment while classification runs on Heravox's AI layer; it can also connect to your internal systems (CRM, NPS, surveys, call center). On-premise runs the full platform — collection, LLM-based classification, dashboard — inside your infrastructure, with nothing leaving your network. Hybrid and On-premise are offered for regulated industries and strict data-residency requirements.

Question 8

Is your model trained on our data?

Accepted Answer

No. Neither the anonymization model nor the classification layer is trained or fine-tuned on customer data, and the cloud AI provider is contractually committed to not retaining submitted content or using it for training.

Question 9

Is any setup, integration, or IT involvement required?

Accepted Answer

The standard deployment is deliberately integration-light: no API integration, SDK, or system-to-system connection is required to start — a smaller integration surface is also a smaller attack surface. Once your organization is onboarded, Heravox collects and classifies feedback from the selected channels automatically. Hybrid and On-premise deployments, and optional internal-data connections (CRM, NPS, call center), are run as a defined rollout together with your team.

Question 10

Which channels does Heravox cover?

Accepted Answer

Heravox tracks the public channels where customers actually talk about your brand — complaint platforms, app stores, social media, review sites, and public discussion forums, both local and global. Coverage is tailored to where your customers are most active. On Hybrid and On-premise deployments, internal channels — CRM records, NPS responses, surveys, call-center transcripts — can be brought into the same platform, so external and internal customer voice are analyzed together.

Question 11

How quickly do we see results after onboarding?

Accepted Answer

Historical reviews from the selected channels are ingested and classified during onboarding, so the dashboard is already populated within the first few days. After that, new reviews are continuously collected and classified, and the dashboard always reflects the current state.

Question 12

Does Heravox replace our VoC, NPS, or survey tools?

Accepted Answer

It depends on the deployment. On Cloud, Heravox is an external voice-of-customer layer that complements your VoC, NPS, and survey systems by surfacing public customer voice. On Hybrid or On-premise, it can also ingest your internal VoC, NPS, CRM, and call-center data — giving you a single platform for external and internal customer voice combined.

Question 13

Can we customize categories and taxonomy?

Accepted Answer

Yes. Heravox ships with a sector-specific taxonomy tuned during onboarding — for insurance, for example: claims, policy, call center, mobile app, pricing, and renewals — and your team can add custom categories that reflect how your business thinks about the customer journey.

Question 14

Can we export analysis results?

Accepted Answer

Yes. Analysis outputs export as PDF reports directly from the dashboard, and scheduled report deliveries by email — daily, weekly, monthly — are built in. Additional export formats such as Excel are on the roadmap and enabled per customer need.

Question 15

How accurate is the classification?

Accepted Answer

Heravox runs an agentic classification architecture: a network of specialized AI agents — each focused on a dimension (topic, sub-category, sentiment, urgency, risk) — processes every review, cross-checks each other's output, and produces a consolidated classification validated by rule-based checks. Combined with a sector-specific category system, this yields high accuracy on the categories that matter operationally — in insurance, for example: claims, policy, customer service, digital channels. The goal is not to resolve individual reviews automatically, but to reliably route them to the right focus area so teams know where to look first.

Question 16

What does the AI detect on each review?

Accepted Answer

For every review, Heravox identifies the topic and category, the sub-category, sentiment (positive / neutral / negative), urgency level, and intent (complaint, suggestion, question, praise). Reviews that carry churn, reputation, or legal risk are flagged as critical.

Question 17

Can we ask questions about our data in natural language?

Accepted Answer

Yes. Hera, the platform's analysis assistant, lets authorized users ask questions in plain language — “what drove the negative trend in claims this month?” — and answers grounded in your classified data. Like every AI output in Heravox, Hera's responses are clearly labeled as AI-generated and are advisory: they support your team's judgment rather than replace it.

Question 18

Which teams typically use Heravox?

Accepted Answer

Customer experience, digital channels, marketing, operations, and executive teams. CX teams use it for trend and root-cause analysis, digital teams monitor their apps and websites, marketing tracks perception shifts, and executives get a periodic portfolio-level view.

Question 19

Is Heravox a reporting tool, or an operational tool?

Accepted Answer

Heravox is a decision-support layer. It continuously collects and classifies customer voice, turns it into prioritized categories and trends, and makes shifts in customer perception visible — so your teams spend their time acting on insight instead of hunting for it across channels.

Question 20

How much operational work does Heravox require from us?

Accepted Answer

Day-to-day operation is effectively zero. Data collection, classification, and dashboards run automatically. Your team logs in when they want insight, and receives digest emails — daily, weekly, or monthly — based on their preferences.

Question 21

How do alerts and email digests work?

Accepted Answer

Each user configures their own digest cadence (daily, weekly, or monthly) and timezone. Digests summarize the most relevant movements in volume, sentiment, and categories, and surface critical reviews that need attention — so stakeholders stay informed without anyone having to build manual reports.

Question 22

Can we run a pilot before committing?

Accepted Answer

Yes. We run short, well-scoped pilots — a defined set of brands and channels — under a proper contractual framework including a data-processing agreement. Your team evaluates classification quality, dashboard fit, and workflow value on real data. If you decide not to continue, all data is returned and verifiably destroyed at the end of the pilot.

Question 23

What happens if there is a security incident?

Accepted Answer

Heravox operates a defined incident-response process. Affected customers are notified within the contractually agreed window through pre-designated contacts, together with an impact assessment and remediation steps. Notification obligations and timelines are fixed in the data-processing agreement — they are commitments, not intentions.

Frequently Asked Questions

Couldn’t find your answer?