CogniVis is a ready-to-use system—you can configure and launch it yourself, but our implementation methodology involves close collaboration with your team, so you can start making full use of its capabilities as quickly as possible.
The initial CogniVis setup is like a factory machine with an engine and default settings. It is fully functional – you could probably build a working assembly line around it. However, our implementation methodology means that CogniVis adapts to your processes, not the other way round.
Each system component, such as the Chat Widget, review support and Profitroom integration, is implemented in several stages to enable deep personalisation.
01
PREPARATION
Trial
Organising the implementation, agreeing the meeting schedule and knowledge scope, identifying your system providers and aligning expectations. This stage concludes with preparation of the initial solution.
02
TESTING
Trial
Initial testing with a small group of your employees. We conduct these tests through specially secured channels and do not interfere with your service operations. This stage concludes with documented feedback from you.
03
REFINEMENTS
Trial
CogniVis applies refinements based on the feedback provided. This stage concludes with approval of the revised solutions and authorisation for publication.
04
PUBLICATION AND MONITORING
SUBSCRIPTION
Publication for the intended audience, such as website users, guests staying at the property or the entire internal team. We closely monitor the initial results and, where necessary, apply hotfixes for non-critical adjustments. This stage ends one week after the component is published.
05
DEVELOPMENT AND SUPPORT
SUBSCRIPTION
Continuous development and improvement of the solution based on observations, changing needs and evolving technology. At this stage, we also deliver new features, more advanced customisations and integrations with PMS, CRM, HiS, Booking Engine and other systems. This stage continues for as long as you wish to use CogniVis.
RAG search
CogniVis is built on Retrieval-Augmented Generation (RAG), an approach that combines the power of large language models (LLMs) with precise information retrieval from your own resources. This enables us to provide responses that are not only relevant, but also based on up-to-date, verified data. We have been developing and refining our RAG solution since 2023, with a focus on the specific needs of the hotel industry.
No hallucinations thanks to strict contextual grounding
Full auditability – we always know where a response comes from
Greater relevance through hybrid search and re-ranking
Security – permissions control and data masking
01
Retrieval
Vectorising content and selecting the most relevant passages from the sources.
Hybrid search: BM25 + vectors
Permission and context filters
02
Augmentation
Normalisation, deduplication and compression – the model receives only what is relevant.
Re-ranking and relevance scoring
Protection against data leakage
03
Generation
Responses created by an LLM, with optional citations and no hallucinations.
Citations with paragraph numbers
Response mode aligned with company rules
Simplified CogniVis pipeline
1User query
2Normalisation and semantic expansion
3Intent recognition, relevance and permission classification
4Hybrid search and re-ranking
5Context merging and compression
6Generating a relevant response or handing over to a human
Example response
No hallucinationsSource-based
Every bathroom has a hairdryer[1]. These hairdryers are usually wall-mounted or placed in a drawer beneath the washbasin[2]. If you need an additional hairdryer or have any special requirements, please contact the hotel reception, who will be happy to help.
source 1source 2
*Visible source citations are an optional feature that can be disabled after the trial period. Responses will still be based on sources, but without visible citations.
Flexibility in choosing an LLM
CogniVis can integrate with various large language model (LLM) providers, including OpenAI, Anthropic and others. This flexibility enables cost and performance optimisation, as well as rapid adaptation to changes in technology.
Compatibility with future models
We design integrations with the future in mind. When a more powerful model becomes available, you can enable it without rebuilding the system.
AI development means better performance at your property
New generations of models provide more accurate responses, better contextual understanding and greater task automation.
No dependence on a single provider
If Anthropic releases a model that is better than OpenAI’s, you can switch operations with a single decision – without migrating content or experiencing downtime.
Greater cost efficiency
Choose the provider offering the best balance of quality and price. Use premium models for complex tasks and cheaper ones for routine work.
Reduced risk of outages
If your primary provider experiences an interruption, we route traffic to an alternative model. Service continuity without manual intervention.
Experiments and comparisons
Run A/B tests and benchmarks across models. Base your decisions on quality, speed and cost metrics.
Supported providers
CogniVis + LLM on your infrastructure
We can deploy both CogniVis and an LLM locally (on-premise) on your own infrastructure, without requiring a public cloud. This gives you full control over your data and enables you to meet even the strictest requirements.
OpenAI
Strong language understanding and a stable tool ecosystem. Best for most use cases.
Anthropic
More expensive models focused on precision and maximum relevance.
Amazon
A wide range of models and integration with AWS. A good choice if you already use Amazon’s cloud.
Google
Cheaper, lighter and faster models with strong multimodal capabilities.
Meta
Local models that we can deploy directly on your infrastructure, without a public cloud.
Microsoft
Models from Azure OpenAI Service, providing the highest cloud security standards.
Deepseek
Very affordable, lightweight models that can also run locally without the cloud.
Mistral
A European provider offering highly competitive prices and powerful models.
Frequently asked questions
What are the key stages of implementing AI in a hotel or accommodation property?
+
Implementing AI in a hotel or accommodation property involves several key stages, including analysing needs and business objectives, auditing knowledge sources, selecting suitable technologies and providers, configuring and integrating the system, training the team, and monitoring and optimising AI performance.
How does CogniVis ensure the accuracy of AI responses?
+
CogniVis AI uses advanced Retrieval-Augmented Generation (RAG), Chain-of-Thought and other techniques to ensure that AI-generated responses are accurate, up to date and based on designated knowledge sources.
AI in the channels you already use
Automate responses or display suggestions without changing your team’s everyday tools.
