Parloa has developed an AI agent management platform that uses generative AI to automate customer interactions in contact centers, offering both fully autonomous AI agents and real-time AI assistance to human agents. The Berlin-based scale-up’s platform integrates with enterprise telephony systems, CRM tools, and the Microsoft Azure ecosystem, supporting voice, chat, email, and messaging channels. Key enterprise clients include Allianz, Booking.com, SAP, HealthEquity and Swiss Life.
“Every Fortune 500 company wants to integrate AI to drive business outcomes,” says CEO and co-founder Malte Kosub. “Customer support is one area that is working at scale.”
Large enterprises in every sector maintains a contact center and virtually all of them are looking for ways to reduce cost, reduce wait times, and scale support without proportionate headcount growth. It is no surprise then that Gartner forecasts that agentic AI will autonomously resolve 80% of common customer service issues without human intervention by 2029.
Parloa is gunning for a large percentage of that business. “We have one clear goal to be number one globally in this space,” Kosub says. The enterprise AI platform, founded in 2018 by Kosub, Tilmann Böhme, and Stefan Ostwald, has raised a total of $562 million in funding across five rounds — most recently a $350 million raise that tripled its valuation to $3 billion. The company reports annual recurring revenue of over $50 million.
Prior to Proloa, co-founders Kosub and Ostwald, who serve as CEO and Chief AI Officer respectively, co-founded Future of Voice, a conversational AI consulting agency in Germany, and were among the first developers in the country building applications for Amazon Alexa. That hands-on grounding in voice technology — in how real users speak, how real systems break, and how real enterprises struggle to deploy AI at scale — shaped everything that followed.
Parloa was incorporated with a specific mission: to help large enterprises reinvent customer service using AI. Early products focused on voice automation for German-speaking markets. By the time the company raised its $66 million Series B in 2024, it had already begun to outgrow its European roots, opening a New York office and signing its first American clients.
The pivotal product launch came with the introduction of the AI Agent Management Platform, or AMP — described by the company as the industry’s first agentic AI platform purpose-built for enterprise contact centers.
“When it comes to outcomes customer support is challenging: you can’t find enough people, it is a frustrating job and there is high turnover,” says Kosub. “When AI takes over a lot of those problems are fixed. It is highly reliable and there are cost efficiency gains.”
How The Technology Works
Understanding what makes Parloa’s platform distinct requires an explanation of how it differs from the AI customer-service tools that came before it.
Classical contact-center automation was built on decision trees: a customer calls, the system presents a menu, the customer presses a number, the system routes accordingly. The introduction of natural language processing brought voice bots that could understand spoken queries, but they were typically trained on narrow domains and broke badly when customers deviated from expected phrasing. The next wave introduced large language models, which dramatically improved language understanding but introduced new challenges around reliability, hallucination, and enterprise governance.
Parloa’s AMP is specifically designed to address those governance challenges. The platform is structured around four stages: define, test, scale, and optimize. In the define phase, teams configure AI agents using natural language — what the company calls ‘briefings’ — rather than rigid scripts or code. An agent learns which processes, policies, and data sources it should draw from, making setup accessible even to non-technical teams. Before any agent goes live, it passes through a rigorous simulation and testing environment: data isolation, content filtering, and behavioral monitoring all happen in a sandboxed environment before customer contact. Parloa says setup takes weeks rather than months.
Once deployed, agents operate with what the company describes as a latency as low as 700 milliseconds — fast enough to maintain the rhythm of natural conversation. They support ‘barge-in’ capability, allowing customers to interrupt mid-sentence without losing context, and can handle dual-tone multi-frequency input for customers who prefer keypad navigation. Crucially, they can access multiple enterprise systems simultaneously: CRMs, ERPs, and knowledge bases, pulling live data rather than relying on static training corpora.
A 2025 feature called Agent Composition extended the platform’s global reach significantly. Rather than building separate agents for each market, companies can now build a single agent and refine it across regions, languages, and channels — maintaining consistent brand identity while adapting to local dialects and regulatory environments. This addresses one of the most persistent headaches for multinationals deploying conversational AI: the cost and complexity of localization, says Kosub.
The Competitive Landscape
Competitors include well-funded European challengers as well as deep-pocketed American incumbents.
PolyAI, founded in London in 2017, is arguably Parloa’s most direct competitor in the voice-first enterprise space. The British company’s proprietary Owl speech recognition and Raven reasoning models have earned it a reputation for natural-sounding conversations — handling interruptions, context switches, and informal speech with notable fluency. In real deployments, PolyAI has reported call containment rates above 80% without human escalation. Where Parloa emphasizes the breadth of its management platform, PolyAI stakes its claim on the quality of the voice interaction itself.
Sierra AI, based in San Francisco, has differentiated itself through a focus on brand alignment and emotional intelligence. Its agents are trained not only on product knowledge but on a company’s tone, values, and policies, aiming to make every automated interaction feel consistent with the brand’s broader customer promise. Emotion detection — the ability to identify frustration or satisfaction in a caller and respond appropriately — is presented as a core capability. Sierra operates primarily in the North American market and targets sectors including retail, financial services, and technology.
Parloa competes not only with AI-native startups but with the established contact-center-as-a-service players and CRM giants that have incorporated AI into their platforms. Genesys, Avaya, and Salesforce Service Cloud all offer AI functionality, but they are primarily platform providers: AI is a layer on top of existing infrastructure rather than the foundational architecture. Customers consistently describe these solutions as less capable in pure AI performance than the dedicated players, though they benefit enormously from installed-base inertia. Parloa’s strategy of integrating with these systems, rather than displacing them, is designed to circumvent the switching costs that protect the incumbents.
There is also the question of competition from foundation model providers themselves. OpenAI, Google, and Anthropic are developing increasingly capable voice models — and offer them through API. Parloa’s answer, implicit in its product architecture, is that the hard problem is not the AI model itself but the enterprise infrastructure around it: governance, testing, integration, security compliance, and multi-market scalability. The AMP is positioned as the system of record for that infrastructure — a layer that retains value regardless of which underlying model is best at any given moment.
The sector’s recent history provides cautionary examples. Klarna, the Swedish buy-now-pay-later firm, became a prominent case study in AI customer service after deploying AI agents at scale — but later acknowledged it may have moved too quickly, having cut thousands of jobs and then revisited some of those decisions. The episode highlights a genuine risk: AI customer service works, but the human and reputational costs of not calibrating deployment well can be substantial. Parloa’s emphasis on simulation testing, gradual rollout, and human-in-the-loop governance is designed to help companies avoid missteps.
What Comes Next
Geographic expansion is the immediate focus: new offices in San Francisco and Madrid will bring the company closer to its largest existing and prospective clients in North America and Southern Europe. The company has also launched a formal Partner Program, working with systems integrators and technology partners to extend its reach beyond direct sales.
Product development will center on expanding the AMP’s capabilities — particularly in analytics, continuous learning, and what Parloa calls the ‘Optimize’ phase of its platform lifecycle. The company has also introduced the Parloa Promise: a public commitment to agent reliability, continuous innovation, and human-centric responsible AI. Another plus is “data privacy is in our DNA,” says Kosub.
The partnership with HealthEquity — the largest Health Savings Account administrator in the United States — and a multi-million-dollar contract with TP, one of the world’s largest customer experience outsourcers, signal Parloa’s ambition to move beyond individual enterprise contracts toward platform-level relationships with the companies that manage customer service on behalf of others. If TP and its peers begin embedding Parloa’s technology into their outsourced contact-center operations, Parloa’s addressable market could expand dramatically.
This article is content that would normally only be available to subscribers. Become a subscriber to see what you have been missing
