Adversus includes optional AI-powered features designed to help you better understand, summarize, and work with conversations, and now also to assist in building KPIs through prompt-based suggestions. This article explains how these features work, what to expect, and how to interpret the outputs they generate.
What are AI features in Adversus?
Our AI features currently include:
Transcription of recorded calls (converting speech into text)
Summarization of the conversation based on what was said
KPI generation, where AI can help suggest relevant KPIs based on prompts you input
Quality Assistance based on transcribed conversations
Additional features are currently under development.
All AI results are generated based on the content of the conversation or the prompt you provide. These outputs are system-generated and should not be interpreted as final or verified by a human. For example, in some cases, the AI model may make a faulty judgment when evaluating requirement fulfillment using the quality assurance module.
We clearly label all AI-related content to avoid confusion.
How does the AI work?
When AI features are used, your data flows through the following steps:
Gladia’s speech-to-text engine handles transcription
Mistral’s language models perform analysis
Results are returned to Adversus and displayed inside the platform
KPI suggestions work slightly differently — you provide a short prompt or instruction, and the AI returns one or more KPI ideas based on that input.
Today, all AI features run after a call has ended or upon manual input.
How do I know when a result is AI-generated?
Every AI result in Adversus is marked with our Arthur the AI icon — our helpful AI robot.
If you see Arthur, you know the content was generated or assisted by AI. This includes:
Transcripts
Summaries
KPI suggestions
Quality Assistance
Any other upcoming AI-driven features
We use this icon consistently to ensure full transparency.
What are the limitations?
AI is a helpful tool, but it doesn’t always get everything right. Here’s what to keep in mind:
Transcription quality depends on the language spoken, noise levels, and accents
Summaries may simplify or overlook nuances in the conversation
KPI suggestions are based entirely on the prompt you enter — and might require refinement to match your goals
The system doesn’t “understand” context in the way a human would, and shouldn’t be used as the sole basis for critical decisions
In some cases, the AI model may make a faulty judgment when evaluating requirement fulfillment in the Quality Assistance module
We’re building in feedback tools so you can rate results and help us improve performance over time.
Who is responsible for AI outputs?
You decide how AI outputs are used in your workflows
We make sure the features are documented, optional, and clearly marked
Consent from end-users (your leads or customers) must be collected by you if required by law, before using features like transcription or summarization
Can I override or disable AI features?
Yes. All AI features are optional and can be enabled or disabled per account depending on your needs.
AI is offered as an add-on product and is not included in the standard subscription. Activating it will incur an additional cost, which depends on your usage and setup.
If you prefer not to use AI for any part of your setup, or only want to enable it for specific workflows, reach out to our support team or your Adversus contact.
Activation requirements
To activate any AI features, your organization must have a signed Data Processing Agreement (DPA) in place that includes Gladia and Mistral as subprocessors. These providers are used solely for AI-related processing and do not retain data after processing is complete.
If you’re unsure whether your DPA is up to date or would like to receive the latest version, please contact your Adversus representative or email support@adversus.io.