With agentic AI, companies are conducting enterprise extra dynamically. As an alternative of conventional pre-programmed bots and static guidelines, brokers can now “assume” and invent alternate paths when unseen circumstances come up. For example, utilizing a enterprise area ontology like FIBO (monetary trade enterprise ontology) may help preserve brokers inside guardrails and keep away from undesirable habits.
The bottleneck is now within the consumer expertise (UX) layer. Whereas brokers are dynamic and rework with the info drift guided by ontology, the consumer interface remains to be very a lot static. These experiences with fastened fields and configurations can hamper the inventive freedom given to brokers. Trendy requirements like AG-UI (agent Person interface) assist streamline communication between UX and brokers — however nonetheless the screens have to be pre-defined at design time.
A more recent know-how is taking this to the following stage, dynamically permitting brokers to render their desired consumer display screen primarily based on particular content material. One is A2UI – agent to consumer interface. With A2UI, we first outline a UX schema for a way parts ought to be rendered. This loosely coupled schema permits brokers to construct screens as per the info.
Brokers now talk with a A2Ui compliant “renderer” that dynamically renders screens primarily based on JSON content material that brokers produce dynamically. Screens are absolutely interactive and might talk again with respective brokers utilizing AG-UI. Firms like Copilotkit are actively constructing A2UI renderers that may dynamically construct the UI from JSON spec and wire it collectively again to the agent through AG-Ui.
Furthermore, utilizing newer compression requirements like token object notation (TOON) may help get hold of extremely environment friendly compression and embody schema like ontology and A2UI into context prompts. In fact, as fashions get smarter, they can even embody functionality to auto generate screens compliant with A2UI and AG-UI through pre-training.
The beneath schematic explains one view of this structure.
As proven, the A2UI specification is complementary to a enterprise ontology and centered on rendering logic for consumer interface parts. Taking an instance of mortgage approval, ontology will outline enterprise ideas like loans, events, curiosity phrases, covenants, or circumstances. This information is often in a number of supply methods in several varieties and a standard enterprise ontology helps unifies this into a standard “language." The A2UI specification will outline how consumer expertise parts can be rendered.
Sooner or later, solely the spec wants to alter, fairly than particular person screens, as a result of screens are generated with contemporary content material each time. Additionally, since A2UI makes use of AG-UI beneath the hood, screens keep connection to the unique agent that generated the content material. So, occasions like button clicks and kind submissions might be tracked and responded to. This whole expertise occurs inside a single pane of glass — as an example, a standard chatbot.
The enterprise deliverable is tying collectively ontology, brokers, A2UI JSON, dynamic content material screens and AG-UI message exchanges. Every thing is pushed by enterprise logic and relations outlined within the ontology, that means much less is left for interpretation by the UX designer and UI developer. We nonetheless want these roles on initiatives, however reusable parts are outlined and constructed simply as soon as. Rinse and repeat!
For instance, you possibly can outline that any communication message despatched to a consumer (error, data, warning) be rendered inside a panel together with your firm brand and be compliant with ISO 9241-110. With agentic AI and A2UI, a devoted agent can validate these messages and assemble them on display screen per requirements.
The chat interface nonetheless stays your major interface for customers, however A2UI parts are rendered the identical. Extra importantly, the present consumer screens might be reused as templates to dynamically generate newer screens. This makes your online business extremely strong to enterprise and regulatory modifications.
Patterns like A2UI reduce dependency on consumer interface and complement the dynamic nature of enterprise. Think about an organization undergoes an acquisition and should add new logos to 1000’s of varieties. Now, this logic might be configured within the A2UI spec and ontology and UI modifications can be propagated when customers entry varieties. This helps companies be dynamic and enhance worker productiveness.
Dattaraj Rao is innovation and R&D architect at Persistent Methods.
