We recently held the Tray Megacast: An in-depth look at Merlin AI, where our CEO, Rich Waldron, and I unveiled Tray Merlin AI - our revolutionary natural language automation capability in the Tray platform. 

(If you missed the event, no worries! Watch it now on-demand). 

After seeing Merlin in action, many of you had interesting and insightful questions about how this new capability could enhance business processes and bring unprecedented efficiencies to your work. Since we didn’t have time to answer every question during the live event, we’re addressing them here, as well as expanding on some of the incredible capabilities Merlin has to offer.

In Part 1, we defined Merlin, discussed some of the possible use cases for conversational automation, and provided an update on Merlin’s availability. In this part of the series, we’ll dive deeper into Merlin’s capabilities. Let’s get started!

Merlin as your automation assistant

Q: Can Merlin generate code for us if there’s no straightforward way to automate something?

A: Absolutely! Merlin not only understands and acts upon natural language instructions but can also generate code suggestions for use within your workflows. When faced with complex automation scenarios that require a customized approach, Merlin can provide code snippets or recommend the best course of action to achieve your desired outcome. With this capability, you can harness the power of generative AI to overcome automation challenges and accelerate your workflow automation journey.

Q: Can Merlin generate code specific to a workflow we’re constructing?

A: Yes, indeed! Whether you are constructing a simple or complex workflow, Merlin can understand the context and provide code suggestions tailored to your specific automation needs. By working hand-in-hand with Tray’s builder interface, Merlin ensures that your workflows are efficient, effective, and aligned with your business requirements.

Data extraction and summarization with Merlin

Q: Can Merlin extract vital details from a large document and provide a summary report containing that information?

A: Extracting valuable information from large documents can be a time-consuming task. While Merlin does not currently analyze the content of unstructured data directly, it excels at understanding and processing natural language queries. Currently, we don’t send any customer data into the large language model (LLM), making it impossible to summarize unstructured data like text documents. However, we’re planning to give customers the option to send data if they explicitly want to, which would enable more advanced summaries. Stay tuned for these updates!

Q: Can Merlin perform classification or extraction on data coming from the connectors?

A: At present, we don’t pass sensitive customer data into the LLM for security and compliance reasons. We plan to offer customers the option to do this, which would enable more advanced summarization, classification, and extraction of unstructured data through the AI aspect of the LLM.

Workflow creation and connector configuration

Q: Can Merlin independently construct a complete workflow/process, including the selection and configuration of connectors?

A: Absolutely! Merlin is not just an automation assistant; it’s a comprehensive solution that can handle end-to-end workflow construction. From selecting and configuring connectors to defining authentication requirements, Merlin can guide you through the entire process. It will even prompt you for authentication when necessary, ensuring a seamless and secure workflow-building experience. 

Q: Can Merlin analyze existing workflows and suggest or implement improvements?

A: Merlin is designed to be an intelligent automation assistant that continuously learns and adapts to your workflows. While it can analyze logs and suggest error rectifications, its capabilities extend beyond basic optimizations. Currently, Merlin focuses on providing suggestions based on natural language statements and incorporating user input into existing workflows. We are actively exploring advanced optimization features and large-scale workflow architecture enhancements as part of our roadmap. 

Q: Can Merlin create and update custom API connectors? If we need a new connector that, for example, employs a scraping mode, would it be feasible? 

A: While Merlin’s current focus is on conversational automation and workflow building, we recognize the importance of custom API connectors and expanding our connector library. We are actively working on incorporating Tray’s connector-building capabilities with Merlin which will empower users to create and update their own custom connectors via natural language instructions. This includes the ability to build connectors that employ a scraping mode. Stay tuned for updates!

Q: Is Merlin compatible with all the connectors available on the Tray platform today, including customer connectors?

A: Yes, Merlin works seamlessly with all the connectors available on the Tray platform. Whether you’re using one of our 600+ connectors or a custom-built one via our Connector Builder, Merlin can access and leverage their functionality. Thanks to Tray’s standardized input/output schemas and Merlin’s flexibility, you can easily incorporate any connector into your workflows. Our commitment to providing a robust ecosystem of connectors ensures that you have the flexibility and power to automate diverse business processes. 

Handling data queries

Q: How can a user retrieve the specific data they’re searching for with connectors that offer native complex query options, like Salesforce or NetSuite? Can Merlin as follow-up questions?

A: With Tray’s standardized connector library and Merlin’s natural language processing capabilities, you can easily retrieve specific data from connectors that provide complex query options. Merlin can pair your advanced natural language statements with the relevant connector inputs, allowing you to interact conversationally and obtain precise data from systems like Salesforce or NetSuite. Furthermore, you can engage in iterative conversations with Merlin to refine your queries and ensure accurate results. 

Q: How do we map input with Merlin in the Tray builder, such as mapping “First_Name” to “First_Name__c” within an existing callable workflow? 

A: This is where the LLM shines as it is very good at understanding how language is related. When mapping data between two connectors/services, Merlin can determine that those fields are related and refer to the same semantic entity.

Look out for Part 3 in this series where we’ll delve into how Merlin keeps data secure, as well as a peak at the Merlin roadmap. Stay tuned!