The Conversational Stack

The Conversation stack

There is a term in IT called a 'technology stack'. Briefly, it refers to a suite of hardware and software components that comprise an application, server or network.

At Strategic Machines, we like applying this concept to conversations. While the process of two people conversing is easy to envision, it is difficult to mimic with even the most sophisticated technology stack. The complexity of a conversation quickly mounts when we factor in the elliptical nature of any dialogue.

For example, think about the dialogue between two people who are completing a restaurant order. In a simple case, the interchange can be mapped out as a sequence of questions and answers. While a chat- or voice-app for a simple case is easy to program, human interactions are rarely simple. Even a restaurant order can quickly diverge into related but unexpected topics like ingredients, delivery, substitutions, or pricing. The simple morphs quickly into the complex. A business owner would rarely deploy automation that could not deliver high customer satisfaction. A menu-driven kiosk works; a general-purpose chatbot does not.

The more sophisticated conversational platforms, like IBM, Google or Microsoft, include capabilities to detect language, decipher intent, identify parts-of-speech or translate written text to voice. However, in our view, the most challenging element of developing a 'dialogue app' is the content. To keep pace, the platform needs to ingest content relevant to the domain, like Healthcare, Finance, or Retail. Machines 'learn' by processing sample datasets, and with well-constructed algorithms, can infer the intent of a text based on the sample. The design, construction, and maintenance of these 'knowledge datasets' are expensive, especially as the context of the domain broadens. Costs mount exponentially with dynamic content.

Our design philosophy for conversational apps is to build for the long run: make it simple, engaging and affordable. We focus on well-defined context, and help clients develop a dialogue flow that embraces human intervention where required. We deliver capabilities at a compelling price point by building on top of large language models. The Mayo Brothers noted over a century ago that high performance was achieved 'not by working more complexly, but more simply'.

We hope to show you that our Machines can serve as your simple Conversation stack.