ChatHarmoni is an AI capability that quickly generates insights from your data. It supports conversations to guide AI research, allowing for deeper analysis and a clearer understanding of your data's context. ChatHarmoni uses the secure IpsosFacto API to keep your data safe from open model training.
Coming Soon
In this article
1. What is ChatHarmoni?
ChatHarmoni is Harmoni's conversational artificial intelligence (AI) chatbot designed to interact with your data. It can help to create and describe projects and generate analysis insights to get you to your insights faster.
After auto generating initial analysis insights, you can interrogate the data and ask follow-up questions or instruct ChatHarmoni to run additional analyses. Insights can be saved to a story and shared with others.
ChatHarmoni is in beta development and is not currently available to all.
2. How does ChatHarmoni work?
ChatHarmoni uses IPSOSFacto's API access for all generated results and takes benefit from all the legal protections and contractual agreements that IPSOS has in place around the LLM's available and in use. Further to this, ChatHarmoni also ensures contextually important information to the data like filters, weighting, measures, titles, etc., are also analyzed to allow the ChatHarmoni to answer with the greatest accuracy.
ChatHarmoni's answers all occur 'in session' and allow for onscreen display of replies through a chat style interface. ChatHarmoni can be used in the analysis area or in a saved story. You can copy answers to the story view comments area. Here you can adjust/edit as needed and then save to that story view. This can then allow the comments to be seen by other users with story owner access or be part of exports to PPT as the comments will flow to the notes section of the exported slide.
ChatHarmoni capabilities work for any standard data table, chart or respondent/record view analysis. You can use ChatHarmoni to help summarize and suggest themes when analyzing respondent level open-ended comments/verbatims.
Hallucinating AI
We should all be aware that ChatHarmoni is relying on the AI to ultimately supply answers and as we all know, there can be occasions where AI's are still 'learning' and may generate incorrect, fabricated, or misleading information. This is where you can provide important sanity checks on the data. Ensure that you read and understand the analysis. If you are uncertain, verify the results yourself.
3. Client Data Protections
Harmoni continues to provide a protected ecosystem of data delivery, exploration and access controls and it remains important to us to ensure any of our customer's data is protected. As we open the door to the AI, we continue to have this at the forefront of our actions. When we say 'The AI' we refer to the variety of LLMs (Large Language Models) that are publicly and commercially available for use via API/Token. ChatHarmoni uses IPSOSFacto for all generated results Through this relationship we have a broad range of LLM's available for use and we are also in a position to leverage the commercial and contractual protections to data that are in place or being negotiated. This currently includes, but is not limited to OpenAI, Google, Anthropic and Mistral. The data protections remain a critical success factor for us, so these agreements also dictate and prevent data from being passed back to train open models further.
4. ChatHarmoni Functionality
The functionality that is currently supported in ChatHarmoni is as follows:
- Asking questions of your data
- Harmoni function orchestration (many analysis operations)
- Interpreting data in an Analysis View
- In Stories
- In the Analysis Area using the Interpret Talent
- On-the-fly semantic/thematic exploration of open ends - In an analysis view saved to a Story
- Nesting in Analyses
- Orchestration of multiple analysis views or full stories - Save all Talent
- Current LLM: GPT 4.1
The following items are not yet supported in ChatHarmoni:
- Connected sources
- Verbatim/Open ends not yet supported in the analysis area (This is supported in Stories)
- Low sample warning/understanding
- Complex SIG testing application
- LLM preference
- Context Hierarchy
Where to from here?
Learn more about Harmoni's Smart Tools