Implementing Voice AI Correctly in 5 Steps
"Press 1 for sales. Press 2 for support. Press 3 for..."
No one likes hearing that voice on the phone anymore. 89% of service professionals say conversational AI increases the self-service resolution rate (Salesforce State of Service, September 2025; n=6,500). By 2027, 50% of all service requests are expected to be resolved by AI, up from 30% in 2025 (Salesforce State of Service, September 2025).
The question is no longer whether voice AI is coming. The question is how you implement it without ending up running two separate systems that know nothing about each other. This article shows five steps that determine success or failure.
1. Choose the right provider
The first step is choosing the right voice AI solution. Four providers are shaping the German market in 2026, each with a different focus.
Pendura is a phone solution for first-class customer service and a direct branchly partner. Pendura handles the phone side, call answering, routing, and customer outreach. The integration with branchly is coordinated, so you don't need to set up a custom integration project. Pendura on the phone, branchly as the database and dashboard, done. Particularly suitable for mid-sized companies that want to manage their website AI and phone AI from a single source.
The Red Phone from London replaces rigid IVR menus with natural-language routing. The mission is clear: fix the first 30 seconds of every call. Native audio voice model with under 600 ms latency, 99.9% uptime, 30+ languages. Integrations for ServiceNow, Twilio, Five9, Salesforce, Genesys. Suitable for companies with an existing contact center that want to end IVR pain without switching the entire platform.
Fonio is the volume player in the German market. 9,500+ customers, 4.9 stars on Trustpilot, "#1 Software Tool of the Year 2026" by OMR out of over 10,000 companies (OMR, 2026). AI phone assistant for small and mid-sized businesses, setup in minutes. Proprietary orchestration layer, servers in Nuremberg. Suitable for SMEs, self-employed people, and tradespeople who can't afford to miss calls but don't run a contact center.
Newcall is an AI phone assistant for tourist information centers and destinations in the DACH region. The AI assistant Mira answers guest inquiries 24/7 in 25+ languages, fully grounded in the destination's data. Servers in Germany, GDPR compliant, EU AI Act compliant. Official destination.one partner with direct DMS access. Go-live in one week, full service including setup and ongoing optimization. From €650 per month (new-call.de, 2026).
Which provider is right for you depends on your industry, your volume, and your existing infrastructure. What all four have in common: they are direct partners of branchly and can be natively connected to the branchly platform.
2. Avoid a duplicate database and set up a central knowledge base
The most common problem when implementing voice AI is not the technology. It is the architecture. Companies buy a voice AI solution, already run a chat on their website, and end up with two systems running side by side.
The problem: the voice AI on the phone knows different information than the chat on the website. A customer calls to ask about opening hours and gets a different answer than on the website, because the two systems draw from different sources. Every change to the knowledge base has to be made twice. Every FAQ answer has to be entered into both systems. This dual maintenance costs time and leads to inconsistent answers. Compliance has to be clarified separately for each system.
The solution is a central knowledge base. You set up your content once: website texts, FAQs, product data, PDFs, event calendars, external databases. This knowledge base then applies to every module: the chat on the website, AI Search, Advisor, Navigator, Forms, and the Voice Agent on the phone.
If you update an FAQ entry, it is immediately available on the phone and on the website. If you add a new product, the chat, search, and voice AI all know about it. One setup, six modules, one data foundation. That is the difference between "buying voice AI" and "strategically implementing voice AI".
3. Ensure GDPR compliance from the start
Voice AI in customer service falls under the scope of the EU AI Act, which has been in force since August 2024. High-risk system requirements take effect in August 2026, with fines of up to 7% of annual turnover. The BFSG (Accessibility Strengthening Act) and WCAG accessibility requirements have been mandatory for B2C services since June 28, 2025.
Thinking about these rules from the start instead of tacking them on later saves time and money. The questions you should clarify before implementation:
Where is the data processed? US servers are a procurement blocker for DACH companies. GDPR, data processing agreements, Schrems II. Whoever offers EU-native hosting wins. Fonio has servers in Nuremberg, Newcall in Germany, Pendura in the EU, The Red Phone in UK/EU with SOC 2. Branchly hosts on Microsoft Azure EU for LLMs and OVH Cloud for user data. Nothing leaves the EU.
Is there a DPA? A data processing agreement must be in place from day one. If it is missing, the data processing is not legal. Branchly provides a DPA in accordance with Art. 28 GDPR.
Is the solution EU AI Act compliant? The Act requires transparency, logging, and human oversight. Those who cannot prove this risk fines. Branchly is EU AI Act ready.
Is the voice AI solution WCAG compliant? Accessibility has been legally required for B2C services since June 2025. Not every tool delivers this out of the box. Branchly is WCAG 2.1 AA compliant for all modules, out of the box.
4. Use call insights to identify content gaps
Calls are a data source that most companies leave unused. Voice AI changes that, but only if the data flows into a central dashboard.
If 100 customers a month call about opening hours, that is no coincidence. It is a signal that your website doesn't make this information prominent enough. Frequent call topics are FAQ candidates. Instead of guessing what customers ask, you see it in the data.
Callers say things they would never have written on the website. "Product X doesn't work with Y." "I wish there were Z." That is free market research material if you collect and analyze it.
If callers are frustrated about a particular topic, the dashboard shows it. You can react before the topic turns into a complaint.
The prerequisite for these insights: the voice AI data flows into a central dashboard that also aggregates the website data. Two separate dashboards will never show you the full picture.
5. Connect voice AI to your website AI via branchly
The last step is the connection. Branchly does not offer its own voice AI. Branchly is the central database and dashboard that connects voice AI solutions like Pendura, The Red Phone, Fonio, or Newcall to your website AI. The voice technology comes from the partner; the data foundation and insights come from branchly.
All four providers are direct partners of branchly and can be natively connected to the branchly platform. The voice AI on the phone then accesses the same knowledge base as the chat on the website, AI Search, Advisor, Navigator, and Forms.
You set up the knowledge base once in branchly. Create, structure, and index content. It then applies to every module:
The chat on the website answers questions from the knowledge base.
AI Search finds relevant content via hybrid search.
Advisor recommends suitable products from the catalog.
Navigator guides visitors through processes.
Forms intelligently pre-qualify inquiries.
The Voice Agent delivers the knowledge base to the phone via the integrated voice AI solution.
What you get from the connection:
No conflicting information. Phone and website draw from the same knowledge base. A customer gets the same answer on the phone as in the chat. No contradictions, no outdated information on any channel.
No duplicated maintenance effort. You maintain one knowledge base, not two. Every update applies to all channels immediately.
One dashboard. All voice and web insights in a single interface. What topics are asked on the phone? Which ones on the website? Where are the content gaps?
One compliance standard. GDPR, EU AI Act, WCAG, solved once, not per channel. One platform, one DPA, one audit trail. Microsoft Azure EU hosting for LLMs, OVH Cloud for user data. Nothing leaves the EU.
No dual solution. You don't have to maintain two platforms, manage two contracts, or evaluate two dashboards. One platform, two interfaces, one data foundation.
Summary
Successfully implementing voice AI means more than buying a tool. It means choosing the right provider, avoiding a duplicate database, considering GDPR from the start, using call insights for content gaps, and connecting voice AI with your website AI via branchly.
Whoever takes these five steps does not operate an island. They operate a platform that gets better with every interaction.
