Frequently asked questions.On the product, security and deployment.

Call transcripts, call notes, emails, surveys and complaint records. For calls, most teams continue with their contact center's ready-made transcripts; if you prefer, Heravox transcribes the recordings with the best-performing Turkish speech recognition models. You can also bring social media and app store reviews into the same analysis.

Customer experience, contact center and operations, product and digital channels, and leadership. Each user sees only the data their role needs, and tasks land with the team that owns them.

Heravox was built for Turkish. Inflections, typos and everyday speech are handled together with the terms of your industry and your company. Labels are written to a category tree built around your own journeys and products.

On more than ten dimensions: main category, subcategory and detail; root cause; journey type, phase and step; emotion and intensity; behavioral intent (e.g. switching provider, formal complaint); urgency; and the source of the problem (the company, a partner provider, an agent).

Yes. The category tree is built with your team during setup, around your journeys, products and terminology. When a new topic appears in the data, Heravox suggests it; once you approve, it joins the tree.

We measure accuracy on your own data. During setup, a sample of your records is checked together with your team and the category definitions are sharpened against it. Records the model is unsure about are flagged separately for your team to review.

Yes. Calls, emails, surveys and complaints come together on the same customer's timeline, so their whole history from first contact to today is on one screen. Heravox detects the contacts that belong to the same journey and groups them, and a customer who comes back about the same issue is flagged as a repeat contact.

Which step of the customer journey concentrates negative experience, and that step's root causes, for example the assessment step of an application journey and its most frequent cause. After an action, before and after for the same step sit side by side.

If the model runs inside your environment (a model on your own GPUs managed by Heravox, or your organization's own model), data goes nowhere and no masking is needed. For the Heravox TR model in Türkiye, masking is optional. If you choose AWS Bedrock abroad, personal data is always masked before the request leaves. Customer data is never used to train any model; each organization's data is kept in its own database.

No. Customer data never trains any model, including Heravox's own.

With an on-premises installation, entirely in your environment. With a managed cloud installation, data and backups are kept in Türkiye. Data is encrypted in transit and at rest.

No. The installation package ships with every image, and the licence is verified offline. In the on-premises profile, every path that could send data outside your environment is switched off.

Each organization's data lives in its own dedicated database; two organizations' records never share a table, and cross-access is blocked at the application layer.

We have no standing access to your environment. Administrator accounts belong to your system administrators. When support is needed, access is opened only through your privileged-access process and only for the duration of the request.

Yes. Heravox connects to your identity provider over SAML 2.0, OIDC, CAS and LDAP / Active Directory. Multi-factor sign-in is mandatory for local accounts. Each user sees only the data and actions their role allows.

Yes, we operate in line with them. Personal data is processed only as far as needed; if the model sits outside Türkiye, it is masked before the request leaves; data stays in Türkiye or in your environment. AI output is advisory, and no automated decisions are made about individuals.

Yes. We answer your security questionnaires and share our policies and the annual independent penetration test report under NDA.

We handle incidents through a documented response process. If an incident affects you, we notify you in writing within the period set in the contract.

Hera is Heravox's AI agent. It answers plain-language questions with charts and numbers, turns the answer into an executive report, edits the report on request, opens tasks and cases, tracks cases until they close, and alerts you when a threshold is crossed. Nothing opens without your approval.

In Jira, your CRM or Heravox's own case module. The task reaches the owning team with its root cause and linked customer records.

You set a threshold, in plain language if you like: “Tell me if the complaint rate crosses the threshold.” When it's crossed, Hera alerts you. Each user can also pick a daily, weekly or monthly email digest that leads with the most important changes and the records that need attention.

Yes. Reports, including executive reports prepared with Hera, export as PDF, and tables export to Excel.

Yes. With an on-premises installation, data stays entirely in your environment and under your control. You choose the language model to fit your resources: a model on your own GPUs managed by Heravox, your organization's own model (bring your own LLM), the Heravox TR model hosted in Türkiye, dedicated to your organization with zero data retention, or the AWS Bedrock integration. A managed cloud installation is also available; data is stored in Türkiye.

Your team doesn't write connectors one by one. Connections for contact center, CRM, survey, email and complaint systems come ready with Heravox, so what was said, who said it and what happened next meet in the same record.

No, it works alongside them. It reads the free text and calls they collect, links them to root causes and journey steps, and opens tasks back in their system.

Your historical records are processed during setup, so the dashboard doesn't start empty: it opens with your history on day one. After that, new records are processed as they arrive.

Yes. We run a pilot limited to specific channels and a period, under an agreement that includes data processing terms. Your team evaluates the results on its own data; if you don't continue, the data is returned and destroyed as the agreement sets out.

Sending text to a general model is a start; enterprise use needs the rest: a data integration for every source, a Turkish, industry-specific vocabulary and consistent labeling, personal-data masking, cross-channel customer matching, repeat-contact rules, a journey model, an action and before/after loop, an audit trail and on-premises installation. On top of that, language models change every month, so evaluation sets, model migrations and vocabulary upkeep never end. Heravox comes with all of it and owns the upkeep.