On 7 October 2021 IBM launched what they call a new IBM Watson Assistant.
At first glance, it might seem like only a User Interface (UI) change. However, there are more to the the change than just the user interface.
A new bot structure is introduced; moving away from the model of one assistant with multiple skills assigned to it.
The new approach is one of a bot, with resolution methods linked it. Resolution methods respond to users with answers from these sources.
There are some questions and consideration on the new format, which I discuss on later.
For now, the new interface is focused on actions (and search). Dialogs will be added later. It will be interesting how dialogs might change, and fit into the the design.
The new interface is focused on: easy of use, quick bot development and deployment.
Currently there is no NLU (Intents & Entities) components.
The building, testing, publishing and analyzing steps are combined. As seen below defined as a workflow.
- The workflow approach, as seen here above.
- Each bot has a homepage. It is unclear if bots can be combined via an orchestration layer.
- Actions are the primary building block, for now.
- A new way to publish lets you review and debug your work in a draft environment before going live to your customers.
- Analysis is focused on completion rates of actions,
- and driving improvement making use of user conversations.
A few observations…
- It is a clear and crisp user interface.
- Managing versions and deployment scenarios are improved.
- Sequential steps guiding the user through all the steps and ensuring due diligence.
- Dialogs are not currently included; it will be interesting to see how that will look. Only Search and Actions
- Reporting is improved with Action completion rates and also a conversation driven journey improvement process.
- Moving away from the assistant with multiple skills model.
- The new model is constituted by one or more bots.
- A bot has different mediums by which it is surfaced.
- A bot also has various resolutions methods. Current resolution methods are Search and Actions.
- Dialogs are to follow.
- Version management is improved, deployment process with complete bot view.
- It is clear that web based chat interfaces are important and a focus point for IBM Watson Assistant. This is most probably part of IBM’s strategy to prepare and be ready to take full advantage of Web 3.0.
After building a few porotypes, some improvements came to mind which might be useful in future releases…
- Currently all actions created in the bot are included in the deployment version. I would like to be able, to select specific Actions, and only deploy selected Actions and not all actions in the bot. Thus excluding certain actions from certain deployment versions.
- An orchestration layer managing or combining different bots might also be helpful. Within a bot, there will be various actions. You will get to a situation where you do not want to duplicate actions across bots, and use multiple bots simultaneously in one implementation.
- Looking at the Resolution Methods, it will be helpful if you can add multiple Watson Discovery instances. Also, once dialogs are introduced, adding multiple dialogs will also be helpful. Thus there should be two layers of orchestration. The first being multiple dialogs, actions & Watson Discovery instances within the bot. The second being orchestration on multiple bots.
- Dialogs will have to be added for Watson Assistant for two reasons. One: The first being that the vast majority their enterprise implementations are Dialog based and not Actions based. Two: Actions is a low code approach, and as mentioned before, is only really good for small bot implementations like a survey bot, or customer feedback bot etc. Or, as an extension to Dialogs. Actions acting as an extension to Dialogs really makes sense and should the the primary implementation focus.
- In future, being able to add multiple Dialogs to your bot, as you would add multiple Actions will be a significant advantage. As with Search. This can improve the current scenario where developers cannot orchestrate multiple Dialogs, Actions and Search skills in their Assistant.
Part of the new IBM Watson Assistant is a Task Tracker. The task tracker takes the user through a sequence of events. With each event, there is an indication of time.
The tracker automatically updates as the user progresses through their process of bot creation.
- Choose your language (1 min)
- Name your assistant (1 min)
- Learn about Watson Assistant
Create a Conversation
- 4 steps left20 min / 0%
- Create your first action (5 min)
- Add your first example phrase (5 min)
- Create steps (5 min)
- Add 5 examples (5 min)
Share with your team
- 1 step left (2.5 min) / 50%
- Preview your assistant (2.5 min)
- Share your assistant!
With most chatbot development framework providers, there is a continuous quest to introduce more functionality & options.
Introducing functionality & user options, adds complexity. And the challenge is to surface this complexity to the user in the most simplistic and user friendly way.
Herein lies the true challenge.
As seen with IBM Watson Assistant Actions, at first the environment was extremely simplistic and minimalistic. But as usage grows, more functionality is introduced, and the interface growing in complexity… We have seen additions to Actions like:
- Debug Mode
- Variable Managing & Tracking
As a whole, the change in approach to conversational agent development by IBM is good, with focus on:
- Medium Integration
- Resolution Methods
- Completion and Conversation Driven Improvement
The section in this article called Considerations encapsulates my reservations.
As it stands now, most people will only use the new Watson Assistant environment for exploration and to porotype. The environment is not yet ready for a full-scale enterprise deployment; with the absence of Dialogs.
The biggest positive is the IBM’s intent with regards to Conversational AI in general, and Watson Assistant in specific.
Subscribe to my newsletter.
NLP/NLU, Chatbots, Voice, Conversational UI/UX, CX Designer, Developer, Ubiquitous User Interfaces, Ambient…
Cobus Greyling - Medium
Read writing from Cobus Greyling on Medium. NLP/NLU, Chatbots, Voice, Conversational UI/UX, CX Designer, Developer…
IBM Cloud Docs
Find documentation, API & SDK references, tutorials, FAQs, and more resources for IBM Cloud products and services.