Conversational interfaces are driving a new wav of innovation in user experience(UX) technologies. Just like many of its predecessors, conversational interfaces come with new requirements to integrate with back-office systems. This type of model of relying on natural language to integrate with line of business systems is a completely new paradigm in the integration space and, consequently, it is likely to drive a new series of patterns and technologies to address the new requirements.
Natural language integration leverages natural language constructs to access data or perform actions on back-office systems. this new integration paradigm will connect natural language based systems such as digital assistants or bots with many APIs or line of business systems. Needles to say that traditional integration platforms and patterns don’t apply when using natural language as the integration protocol. As a result, we need to rethink and adapt traditional integration models to the new natural language world.
Key Characteristics of Natural Language Integration
To understand the implications of conversational, natural language integrations, it might be useful to explore some of the characteristics of this new integration model.
Voice and Natural Language as Integration Commands
Natural language integration solutions should be able to process both text and voice command and translate them into call to APIs or line of business systems. This type of model entails that the integration solution should effectively derive the object and intent of natural language sentences and translate into a backend action.
In natural language integration solutions, at least of the endpoints will be a bot running on systems such as messaging platforms, wearable’s or digital assistants. This model contrasts with the traditional system-to system integration architecture which is the basics of traditional integration platforms.
Conversational APIs and Connectors
APIs are the detaul integration mole for most modern line of business systems. In the world of natural language integration, we are likely to see the emergence of new types of APIs optimized to process natural language, text and voice requests. Similarly, the concept of connectors in traditional integration systems will have to be extended to support natural language integrations.
Self-Service Integration Platforms will Drive in the Conversational world
Self-service integration platforms such as IFTTT or Microsoft Flow are uniquely positioned to power natural language integrations. The simple, one-to-one nature of self service integration platforms could address a large percentage of the natural language integration scenarios.
Stateless models became the architecture hallmark of traditional integration scenarios. In the conversational world, however, most integrations are stateful in nature as the context and state of the dialog are relevant on each step.
State-Machine Based Integrations
In the conversational integration paradigm, state machines will the default model to architect workflows and integrations . Conceptually, state machines are a natural model to abstract an natural language conversation. In that model, each state represents would represent an interaction of the dialog between the user/bot and the back-office integration system.
These are some of my initial ideas about natural language integrations. I plan to cover more about this subject in a future post.