For any ambitious team working on a chatbot or voice experience, uncovering a design and development process that really works is critical. The tools and systems used in design, development, and maintenance can heavily influence your team’s ability to move fast, together. Any friction involved in the process can lead to discouragement and block progress, which will ultimately result in a poor user experience.
As with any technology, especially a relatively new one like conversational AI, there is no one-size-fits-all tool or process. Each team and use case have unique goals and requirements. Luckily, with the right mix of tools and processes, your team can create an environment where product improvements can easily be made at any point before and after launch. Here are some considerations to keep in mind when planning for conversational AI development.
Avoid the cliff between design and development
When a team sets out to create a product that will actually be useful for the end-user, they need the space to be creative, test out high-level ideas, throw away some ideas and keep others. In this phase you’re still determining the scope of your product and putting guardrails on your focus by saying, “Here’s what our product will do for the user…and here’s a list of things our product will NOT do for the user.”
This product ideation phase is crucial for validating your team’s hypotheses about what users want. It helps you avoid putting time and effort into high-fidelity prototypes that will end up not being used anyway. It helps the team feel confident about the capabilities they intend their product to deliver, and then rally around these decisions throughout the next phases: building the app, iterating on content, testing and fine-tuning the product, and launching it to production.
While the flexibility to test all kinds of ideas has its benefits, it is important to stay tethered to the practical constraints of the final implementation during the prototyping phase to avoid going down a path that may end up being infeasible. Because of this, it is helpful to keep engineers looped into these conversations before product and design decisions get finalized. Ultimately, the goal is to use the final prototyping artifacts directly in production.
Structure content to allow for scale
Content creation begins with scripting the most likely scenarios, or happy paths, and gradually expanding to the many possible permutations of dialogue that could take place, like misunderstandings or a special request. In the end, it’s not uncommon for the vast majority of the content to be outside of the happy path. Furthermore, each interaction can have many pieces of content working together – utterances, intents, entities, forms, stories, responses, variables, media assets, etc. Because of this, even a simple intent or function within the app can have a high volume of the associated content. Since these artifacts work in conjunction with one another, it’s critical for your team’s own organization to manage them purposefully.
To effectively manage chat or voice content at scale, content should live outside of the core business logic of the application. This allows team members to update the copy without having to go inside the application code, retrain or redeploy your bot. With a shared single source of truth for content, such as a CMS, team members can collaborate while still focusing on their individual craft. If the CMS connects live to your application, routine content changes can be made without the need to retrain your models or redeploy your app.
Beyond a separation of content from core business logic, applying necessary guardrails will allow your team to safely collaborate while preventing errors. This can include a system for tracking content versions and seeing a history of changes made. It can also include system validation checks along the way to ensure that changes published won’t break the application. Investing time in these protocols upfront will pay dividends well beyond the launch of the app.
Channel customer learnings
Design and development work does not stop once the app is released. There are endless ways to interact with a conversational AI app and it’s extremely likely that you’ve only anticipated a fraction of those interactions in the initial design. Humans can be quite creative with their words! Examining the real interactions between your users and your system will highlight the holes in your design and allow you to go back and build new content and logic.
To ensure a smooth experience for users, conversational AI apps need to be continuously improved. With the right toolset and process, your team can ensure a smooth transition from design to development and reduce the friction involved in making iterative changes. This will shorten your development phases and allow you to quickly go from fixing bugs to improving the overall quality and feel of the conversation. From there, adding the additional functionality you had originally planned for becomes possible.
Botmock is your team’s conversation design and dialogue management superpower. In Botmock you can drag and drop boxes of dialogue, easily showcase different use cases, and bring your team in to collaborate in real-time. And once you’re done designing, the developer handover is simple!
Jargon is the CMS for conversational AI. With Jargon, teams can author, deliver, and optimize chat and voice experiences. Technical or non-technical team members can publish content changes in a matter of seconds without needing to retrain and redeploy the application. Jargon allows teams to improve the quality of their app with tools built specifically for modeling and editing conversational AI.