November 2024
update on Nov-05-2024
Focus Group
LIVE on Test
TBD
Custom Meeting
LIVE on Test
TBD
Research Mode
LIVE on Test
TBD
Right Click menu in conversation
LIVE on Test
TBD
Brainstorming
LIVE on Test
TBD
Research Mode
Introduction
The concept of Research mode was developed to provide a distinct separation from Omnibox by utilizing a template-based format for insights. Omnibox was designed to offer users flexibility, and feedback indicated that it is highly valued by users; therefore, it remains an essential feature. Consequently, a research mode was introduced to integrate structured and pre-defined approaches.
To access research mode, navigate to the global navigation dropdown, also referred to as the organization dropdown, located at the top of the page and select "Research."
Please note that this mode is currently temporary. We are working on creating a more integrated flow that will allow both ad hoc and structured modes to be managed seamlessly without complicating the user interface. You are encouraged to test this feature at test.consumr.ai to evaluate its functionality. However, please refrain from announcing this feature to users, as changes are forthcoming.
Focus Groups
Introduction
Focus groups are part of our research. A few issues with AI twins we noticed was:
Users didn't know what questions to ask.
Irrelevant questions led to unimpressive answers.
Users who knew what to ask ran out of questions quickly.
The success of the AI twin hinged on users asking the right questions. We wrote articles to guide users on this, but saw no improvement.
During a recent product discussion that evolved into a debate, an intriguing suggestion was made: “Why don't we templatize the questions based on objectives?” This idea immediately captured our interest. We identified three key points that supported its potential value:
We can enhance user interaction by asking multiple questions at once and displaying the answers collectively, rather than requesting users to ask a question and waiting for an AI twin to respond gradually. This approach allows us to have objective-driven questions, and since an AI twin reflects real consumers, the entire interaction can be framed as a focus group.
Focus groups are a well-established method, having been utilized for several decades. Therefore, we can develop templates for questions that effectively elicit valuable information from the AI twin. This strategy will address initial issues and enhance the intelligence level of the interaction.
By focusing on the quality of the questions and the relevance of the answers, we can present findings using established models for representing data in focus groups. Such information can then serve as valuable assets within conversations.
Furthermore, this process can be conducted asynchronously. Users need only to select the objectives they wish to achieve, while the system handles all tasks and processes in the background, ultimately presenting the final output to the user. Initially, we considered posing questions to just one AI twin, representing thousands of consumers. However, we realized that incorporating multiple AI twins from different stages of conversion would allow us to gather more comprehensive data, encompassing perspectives from various demographics.
Demonstration Link
Custom Meetings
Introduction
While developing the concept for focus groups, we encountered two different approaches. One approach suggested using predefined templates with fixed questions posed to multiple AI twins. The other approach recommended allowing the engine to generate questions dynamically based on a given problem statement. During the focus group sessions, questions were determined based on specific objectives. After several days of discussion, it was decided to implement both approaches as separate features and evaluate their results. During prototyping, we realized that these prototypes offered significant benefits:
Focus groups were objective-driven, making the process of templatizing questions straightforward. Conversely, a problem statement differing from the objective enabled multiple AI twins to engage in discussion and collaboratively formulate solutions, leveraging their unique personalities and memories.
While focus group outputs followed predefined models, the outcomes of custom meetings could be detailed minutes that are supported by structured cards. The interactions among various AI twins may encompass not only factual information but also creative elements.
Custom meetings are now available in the test environment. This is not the final design but has been added so Intel users can use it and provide feedback and suggestions for improvement before it goes live. Work is currently underway towards a fully integrated black cell where custom meetings will be included as part of the system.
Demonstration Link
Brainstorming
Introduction
Brainstorming is another integral component of the research mode, conceptualised after the implementation of focus groups. The underlying idea is to utilise problem statements in a manner akin to custom meetings and engage AI twins to brainstorm based on a predefined model and method. For example, the Six Thinking Hats technique allows participants to consider a product, service, or problem statement from different perspectives, symbolised by various coloured hats. A red hat signifies emotion, prompting participants to respond with emotional viewpoints, while a black hat represents negative responses only. In contrast, a yellow hat encourages creative and positive responses.
Currently, we have three brainstorming techniques pre-programmed into this feature of the research mode. One limitation of brainstorming is that it relies less on data from memory and more on the personality-based creative responses, focusing on potential ideas rather than established facts.
Demonstration Link
Right Click menu in conversation
Demonstration Link
Last updated