Overall Satisfaction with IBM Watson Assistant
We identified a need for standardization in the information passed on to and by the technical team, as well as the loss of useful information when an employee leaves the company; as we do not have a structured knowledge base, training new employees and finding information with operational procedures was a very hard task. With the creation of our chatbot the persona SUSAN, we not only managed to maintain a single knowledge base shared with the entire team, but also facilitated the location of manuals, training videos, and key information about our customers in a simple and accurate way, which became available 24x7. Technical staff can ask questions specifying products, error codes, doubts or ask for help to obtain the desired information. In the next step, we intend to expand this service to our customers. Currently, SUSAN can be accessed by a web publication or by mobile devices through PWA technology. Through the use of webhook, we were able to expand the existing resources in our Watson Assistant bot, implementing material sent by email or forms requested by the user. An update that we are implementing in this tool is the automatic translation of texts into other languages and delivery of SUSAN as a Telegram Bot channel. We have a draft environment that can be accessed (with a very restricted access level) at this address: https://demodoka.ddns.net/en-susan
- It has a great NLU engine. The understanding of the sentence informed by the user and its association with the available intentions makes the dialogue more natural and fluid, avoiding linguistic or context confusion.
- Bot development is fast and has a short learning curve. For more advanced users, being able to write your formula directly into the JSON content is a big plus.
- Components such as slots and webhook are present in most virtual assistants, but Watson Assistant allows you to expand these features with combinations and flow rules, allowing you to cover much more of the possibilities and ramifications that a conversation can have. This was a differentiator for me to choose Watson Assistant among other options on the market.
- The possibility of using it in a free plan makes it possible to assess the extent to which the tool will meet the needs of the project.
- Visual formatting options are almost non-existent from the WA panel, using the Carbon Design System is cumbersome and the documentation is very poor.
- It doesn't have a native conversion of text to encoding with channels like Telegram and WhatsApp, as there is in google Dialogflow.
- The new iframe resource uses sandbox but without the option to be able to change its parameters. Could have the parameterization options in the text field itself or by the JSON editor
- Positive: Standardization and reliability in the information provided to users by WA
- Positive: Team engagement to provide content such as intents and responses to expand existing information on WA, we receive an average of 10 new suggestions per month of content and improvement that can be included in the bot
- Positive: Reduction of over 80% in requests for manuals and other documents
It would be easier if there was a native way to integrate with messaging apps like Telegram or Whatsapp, or a mobile version of the WA. You must use an external solution to obtain more complete analytical data. A native theme editor on the WA canvas with more formatting options will be most welcome.
AI and NLP are very assertive, the intentions get mixed up a bit when they have the same or very similar sentence examples, the intention suggestions feature for users I had to disable, as it presented very random suggestions and out of the context that the user was looking for. I miss a native integration with Watson translator.
WA unifies several features of the products presented above, in a way that simplifies the creation of a bot that uses conversational features. Compared to competing products, WA presented more features compatible with my needs, which include variable responses depending on the intent and entities identified during data collection with the user.
Do you think IBM Watson Assistant delivers good value for the price?
Are you happy with IBM Watson Assistant's feature set?
Did IBM Watson Assistant live up to sales and marketing promises?
Did implementation of IBM Watson Assistant go as expected?
Would you buy IBM Watson Assistant again?
Recommended for scenarios where all response text can be entered in the dialog tree, or in Actions mode text boxes. Also suitable for environments with complex dialogs, with many deviations. The 8-sec limitation for getting a response from a Webhook is a deterrent to using in more usage scenarios.