Why Agentic Software is Different?¶
Future software will be very different to Conventional software created so far.
There are 3 areas of main differences:
1. Drive of Process¶
| Traditional Software | Agentic Software (Xians Platform) |
|---|---|
| Users drive the business processes forward (doer role) | AI agents self-drive the process with defined goals |
| Manual process execution and decision-making | Autonomous process execution and optimization |
| Human-centric workflow management | Users take on approver role when needed |
| Users responsible for each step of the process | AI involves users strategically rather than for every step |
| Goal-oriented automation with human oversight |
Autonomous Process Execution in Xians¶
The platform provides several technical capabilities that enable AI agents to drive processes autonomously:
- Enterprise-Grade Workflow Engine: Built on temporal.io, providing fault-tolerant, long-running process execution that can operate for months or years
- Multi-Agent Systems: Peer-to-peer agent collaboration where agents coordinate complex tasks through explicit workflows or emergent cooperation. Agents communicate through sophisticated event mechanisms enabling complex coordination without tight coupling
- Flow Types: Support for both deterministic business processes and non-deterministic conversational flows, enabling different levels of autonomy based on requirements
2. System Evolution¶
| Traditional Software | Agentic Software (Xians Platform) |
|---|---|
| Static intelligence systems | Dynamic learning systems |
| Rule-based engines with fixed logic | Agent-environment feedback loops |
| Intelligence remains unchanged unless software/data is manually upgraded | Continuous learning, adjustment, and improvement |
| Predetermined decision paths and responses | Adaptive intelligence that evolves through experience |
| State and reward-based optimization | |
| Self-improving capabilities over time |
Adaptive Intelligence in Xians¶
Technical features that support continuous learning and adaptation:
- Knowledge Management System: Dynamic knowledge bases that can be updated without code changes, allowing agents to evolve their behavior through the portal interface as well as through the API.
- Agent Memory Types: CAG Knowledge Prompts (Cache Augmented Generation), RAG (Retrieval Augmented Generation), and Cache-Memory are built in for Agents to use.
- Capabilities Framework: Extensible tool ecosystem where agents can acquire new abilities through MCPs, custom capabilities, third-party libraries, or other agents as tools
3. User Interaction¶
| Traditional Software | Agentic Software (Xians Platform) |
|---|---|
| Users often misused as data entry operators | Users engage through natural and adaptive means |
| Form-based interfaces and manual input requirements | AI reaches out through user's preferred channels of choice |
| Limited interaction paradigms | Multi-modal communication (Teams, Gmail, Phone, etc.) |
| Technology-centric user experience | Context-aware and personalized interactions |
| Human-centric communication patterns |
Architecture Requirements¶
To support these capabilities a new type of architecture is needed. This is where agentic platform architecture comes into play.
Natural User Interaction in Xians¶
Platform features that enable human-centric communication:
- Multi-Channel Communication: Native integration with WebSocket (Real time), REST APIs (Synchronous and Asynchronous), and webhooks (Listens to events and messages) for reaching users through their preferred channels
- Agent Handoffs: Seamless transfer of conversations between specialized agents when different expertise is required
- Agent Hierarchies: Support for agent hierarchies, where agents can be grouped into teams and sub-teams, and can be assigned to different levels of the hierarchy
- Sophisticated Message Passing: Structured conversation management with support for both text (
Chat) and structured data (Data) communication
These technical capabilities work together to enable the fundamental shift from traditional software patterns to autonomous, adaptive, and naturally interactive agentic systems.