Conversational interfaces require a clear strategy and implementation plan across service, technology, data, experience, and operations:

• Service: A conversational interface starts with utility. Why is a user interacting with brand by voice or text? What questions or requests will they send to a brand, and what responses do they expect? A clear definition of the service offering and how it translates into a conversation is the critical first step, whether that service is commerce, informational, analytical, or oriented around entertainment. In short: what problem does the brand solve that is so important that a user will add that brand to their phone address book for quick access?
• Technology: While conversational interfaces require a new technology stack, companies that establish the right technology foundation will find that conversational experiences can be quick to launch and evolve. It starts with having a service layer architecture for core brand functionality, and then selecting the right set of natural language cloud services that sit on top of the company’s APIs. In this manner, conversational interfaces can rapidly launch and evolve as new platforms become popular, without requiring massive change to the organization’s backend systems.
• Data: Natural language interfaces are a subset of broader machine learning and artificial intelligence services that are increasingly provided by brands to customers. None of this works without good data. Organizations must think about their data strategy, and how to build a single view of the customer that can learn over time to provide better service.
• Experience: Companies must think about the conversational experience they want to provide. Conversation flows, personality, tone of voice, use of language, and how the conversations are accessed is increasingly the new UI.
• Operations: Conversational interfaces require a rethinking of a company’s operational model. For example, in a customer service environment, where do intelligent agents stop and human agents start? Does the company have the right kind of data scientists, designers, and technologists who can make machine learning systems operate effectively and scale? And, how are these new disciplines organized within the enterprise?

Aaron Shapiro, Are you thinking about how customers talk to your brand?

This is pretty much the line of thinking that’s driven our last two and half years of design work at GDS.

• Service: the baseline and then the utility layer
• Technology: GOV.UK’s new technology architecture and the platform as a service
• Data: getting to authoritative sources of trustworthy core reference data
• Experience and operations: Service patterns and design patterns

---

Although our focus is less on conversational interfaces and more on conversation as an access issue – If you can’t explain what your service does to the people who need to use it, it is a failing service – but the work needed to fix it is the same.

You can pretend it's 2005 and subscribe to my RSS feed