Cognigy: a strong platform that assumes you have a team
A low-code conversational platform from Düsseldorf with real enterprise depth across voice and chat. The line nobody puts in the brochure is that it supplies the tool, not the work, and that assumption decides whether it fits.
Cognigy
Cognigy is a Düsseldorf-based platform for building conversational systems across voice and chat. It is technically strong, it hosts in Germany, and it carries the certifications that shorten a security review in regulated sectors.
The thing worth being clear about before an evaluation is what kind of purchase it is: Cognigy sells the tool, not the build.
What it is
A low-code environment for designing conversation flows, with multi-channel support, analytics and the integrations enterprise contact centres need. You build the flows; the platform runs them, scales them and reports on them.
The strengths are real and they are enterprise strengths: scale, channel coverage, analytical depth, and a governance story that a compliance function can follow. In its own segment it is among the best available in Europe.
The assumption in the brochure
Cognigy assumes an internal team that builds, operates and develops the bots. Not a developer team necessarily, since the environment is low-code, but people whose job includes conversation design, maintenance and iteration.
That is not hidden and it is easy to skim past. A platform is leverage on a team you have; it is not a substitute for one. A company that buys Cognigy without naming who owns the flows ends up with an excellent environment containing three prototypes, which is the most common ending for a well-funded AI project.
The sizing question is not “can we afford the licence.” It is “who spends a meaningful share of their week in this.” If no name comes to mind, the platform is the wrong shape regardless of price.
Where it fits
Fifteen to twenty service staff and a few thousand contacts a month is where the economics start working.
Our own threshold from deployments in this segment, not a vendor figure
Contact centres with volume. Above that line the depth gets used rather than admired.
Multi-channel by requirement. If the same conversation has to work on the phone, in chat and in messaging, a platform that treats them as one design is worth substantially more than three separate tools.
Regulated sectors that need German hosting and a certification list. This is where the DACH advantage is concrete rather than rhetorical.
Where it does not
Below the threshold. A team of four in service does not need a conversation design environment. They need one process automated in the tools they already have, which is a different purchase and roughly an order of magnitude cheaper; we set out the comparison (auf Deutsch).
When the answer needs your systems. A platform answers from its knowledge base. The high-value support questions (where is my delivery, what did we agree, why was I charged this) need a live lookup in your ERP or CRM. That is buildable in any platform and it is not what the platform gives you for free, and it is usually where the automation rate actually comes from.
Text-first situations. If most of your inbound is email and forms rather than voice, the platform’s main strength is idle. Voice is where this class of tool earns its price.
What to verify
- Where processing happens, not only where data rests: two separate answers
- Which model provider sits underneath, named in the DPA as a sub-processor
- Retention for transcripts and recordings, as a number of days including backups
- Who inside your company owns the flows, by name, before the contract
- The escalation design, before the automation: what happens to what the bot cannot answer, and how fast a person sees it
The last one is not a platform question and it is the one that decides the outcome. Klarna’s public reversal was a lesson about exactly this: the exit was removed, not the agents.
And since 2 August 2026 there is a hard rule on top: anyone interacting with an AI has to be told. In a voice channel that belongs in the opening, not in a privacy notice. What came into force and what was postponed (auf Deutsch) is currently the most commonly muddled point in this market.
What we would actually do
For a mid-sized company: automate the text channel first, in the systems you already run, and measure it. If the volume and the results justify voice, evaluate Cognigy then, with a named owner and a real answer on the escalation path.
That order is unpopular because voice is the visible channel and email is the boring one. Email is also where the repetitive volume actually sits in most Mittelstand companies, and it is where a result arrives in six weeks rather than after a six-month procurement.
Frequently asked questions
Is Cognigy suitable for a mid-sized company?
Above roughly fifteen to twenty service staff with meaningful voice volume, yes. Below that, the platform assumes a team and a scale that are not there.
Does Cognigy host in Germany?
Yes, and it carries the certifications that regulated buyers ask for. Confirm the processing location per deployment, not only the storage region.
Do we need developers?
Not necessarily: the environment is low-code. You do need someone whose job includes owning and iterating the flows, which is a role rather than a task.
Can it answer questions from our own systems?
Yes, via integrations, and that is usually where the real automation rate comes from. It is a build, not a switch, and it belongs in the project plan.
What should we do first?
Automate the highest-volume text channel in the systems you already have, measure it, and let that result decide whether a voice platform is the next purchase.
Sources: Cognigy product material and a DACH vendor comparison, checked July 2026.
Strategy Call