Organizations are constantly integrating, replacing, and creating data used in internal applications. Individual teams also have different ways of accessing and using internal data. Seasoned employees may know where to go to access the internal information they need, but new hires often have difficulty.
Generative AI offers easy and intuitive ways to unlock organizational knowledge so teams can do their best work.
If you are new to an organization, finding the organizational knowledge you need can take time. First you need to find out which systems have the information you are looking for. Then, you’ll need to request access to those systems and learn how to navigate them. In some cases, data may be locked, in PDF files, or in other formats that internal search engines have trouble indexing.
Three of the most common roadblocks to employee onboarding and effectiveness are:
- Access to internal applications
- Knowledge of how to use internal applications
- Time and effort needed to navigate internal applications
With the advent of Generative AI, we now have the ability to get precise and concise information from known data sources to drive employee efficiency higher.
Introducing Knowledge Workbot Accelerator
Workato’s Knowledge Workbot Accelerator solves the issue of getting correct information quickly even without prior knowledge of applications.
Because the world of Generative AI is constantly evolving, we created a modular accelerator that adapts to any external product, including LLM, Vector Database, and Datastore.
How It Works
Knowledge Workbot Accelerator works by using Prompt Engineering with a Vector Embedding Database Pattern.
This pattern has two parts:
1.) The Data Ingestion Module ingests organizational data into Vector database as embeddings.
2.) The Interaction Module gets questions from users and generates responses from Vector search & LLM.
Please note that the Data Ingestion Module relies on application identification and availability, and users must decide the frequency with which data for ingestion is refreshed.
Knowledge Workbot Accelerator provides out of box integration with Confluence, Github and on-prem files as data sources and can easily support other applications.
Addressing Data Security Concerns
Data security is essential when dealing with sensitive and confidential internal information. That’s why Knowledge Workbot Accelerator offers the following security features:
- Support for encrypting content and removing PII information from data before storing it in a Vector database.
- Workato’s zero-retention feature can be used to avoid data storage during processing.
- Access control for provisioning restrictions.
The Importance of Continuous Data Refresh
Keeping data ingested by the Vector database up to date is important to ensure optimal user outcomes.
Knowledge Workbot Accelerator will store user session details for each interaction including token consumption, prompt and answer, and feedback requests. This information will be reviewed periodically to improve response.
Challenges & Limitations
There are no granular user permissions for data stored in Vector databases. Any access control you may want to put in place is at the Workbot level.
Currently only text and PDF formats are supported in the accelerator. However, you can extend the accelerator to support other data formats.
Possible Use Cases
This accelerator is flexible in design and can be employed for a variety of uses, including:.
1.) Enabling product support teams for faster ticket resolution by providing AI access to knowledge base and historical ticket info.
2.) Making HR resources available to all employees in an easily ingestible format, reducing opened tickets.
3.) Supercharging internal IT processes and reducing licensing costs for third party internal search products.
The Workato Advantage
There are many solutions on the market in the internal search/knowledge base AI sector. Why use Workato?
1.) Workato has a wide range of connectors available to connect to knowledge base systems, LLM providers and Vector databases. Even if the connector is not available you can create a Custom SDK very easily.
2.) Workato Workbot for Slack and MS Teams can be used to build Workbot easily.
3.) Customers can start using this accelerator with Workato internal storage to store session and source docs metadata without getting an external datastore.
4.) Workato uses best practices for LLM interaction, including data security, conversation and usage tracking, and data chunking.
5.) Knowledge Workbot Accelerator decouples various modules to avoid vendor lock-in.
6.) The ability to store vectors in the Workato internal datastore gives customers time to decide on use case feasibility. Customers with smaller datasets may be able to use Workato’s internal datastore indefinitely.
7.) Workato provides platform connectors for OpenAI and Azure OpenAI. Connectors can be built with Custom SDK easily to integrate with any other LLM providers and Vector databases.
We’ll be releasing a training course on Knowledge Workbot Accelerator soon via our academy, but in the meantime, we’d welcome any feedback or suggestions. Also, if you’d like to request a demo, you can reach out to your CSM or email us at accelerators.feedback@workato.com.
