The Challenge
Traditionally, enterprise applications are designed to drive the user through a pre-defined flow of forms with a narrow outcome in mind: check the status of an order, search over a customer list, or file a request for something. Functional but inflexible.
On the other hands, generic chatbots are flexible but without domain-specific design. The presentation of information lacks the structure we expect in enterprise applications, and we end up in long prompting sessions and cut-and-paste to our "actual" systems.
The Solution
To preserve the of both chat and structured UIs, we need to combine the flexibility of natural-language conversations with the organizing structure of enterprise applications.
This means allowing chat a various levels of granularity, operating on the enterprise data as input, and presented the output in structured ways (instead of only streams of text). How this is done depends on the domain, and is ideally tailored to balance flexibility with efficiency.
Business Value
Improved Flexibility: Conversational AI personalizes user interactions with flexibility beyond the constrains of traditional software where every use had to be design in advance.
Increased Productivity: With chat interfaces infused at both macro- and micro-level, we're able to both perform broader analysis of data, and automate many smaller tasks with precision. An example of this is Cube5 RFP, which has AI infused in over 10 places, augmenting the interface and user experience in both subtle and substantial ways.
Technical Capabilities
Structured Input and Output: When invoking Generative AI as a feature in a UI, structured data is used as input, and the output is (at least partially) intercepted and laid out in structured tables and charts.
Augmented Prompting: With AI infused for specific functions, prompting is either entirely hidden, or significantly augmented. Users need not write long prompt essays to perform work, as prompts are reformulated and augmented based on the context.
Multi-modal Input/Output: When engaging with enterprise data, presenting it in it's original form is essential, whether that's chart, tables, diagrams or images.