Tiledesk will use a new semantic engine based on LLM (Embeddings + Vector indexing) for intent classification. This will allow using less phrases for training revamping the intent classification engine that is useful to address a lot of use cases that go beyond the new LLM approach. pre-release date: November

Developing **entity recognition** (dates, names, cities, products etc.) in training phrases

We would like to have multi-path connectors to better understand connections between blocks and reduce overlapping lines (typical of actual cubic-bezier connectors)

Subscribing to Whatsapp/Gmail/Facebook/Hubspot and other external events to trigger automations.

It'd be useful to be able to send WA templates from the app too to align the desktop chat to the app (ex prospect feedback)

Add to the designer the ability to insert and edit blocks that help the user to illustrate the flow. Textbox with HTML, images, lines, rectangles etc.

More users can simoultaneously edit/view in realtime on the same Automation graph.

Connect Tiledesk with Instagram to facilitate automation workflows for managing inbound requests and executing outbound tactics (i.e. send a message to people interacting with my posts and/or mentioning our profile in its stories)

Add a button to share the whole flow graph, so other colleagues can see the graph. The access can be anonymous, so the users can comment/review the graph in the anonyous chat.

With this new (highly requested) feature you will be able to only generate one or few training phrases for a block. A special button will auto-generate N "similar" training phrases, improving the Intent classification ability to execute an NLP trained block.

Having Project-globals attributes is something really useful. Thanks for you feedback. We're introducing a Project Globals that will allow cross chatbots sharing of project-wide attributes.

Actually, for security reasons, chatbot Globals cannot be exported. In this way you cannot accidentally export some sensible data saved in your Globals section. We will provide an option to allow to selected Globals to be exportable along with the chatbot. It's up to the flow designer to chose what can be exportable.

Manage chatbot rules as Designer events

Adding a copilot module that will learn from agents successful conversations and will provide real-time suggestions to agents based on conversations history and Knowledge base modules. pre-release date: November

Adding a way to access, search and restore past flows' publications.

Development of the "Send SMS" Action for the Chatbot Development Studio to dynamically send SMSs to a generic phone number

Import/export contacts in/from a .csv/.json files including all basic and custom info.

Today is not possible to get a response from a webrequest and make multiple replies, add dynamic buttons etc. Add a way to reply with text/media from an array. As an example: https://www.youtube.com/watch?v=HFqGAjiNoCY

Use the Designer to automate legacy Triggers events

Add a tool to report all the Knowledge base failures. The tool will propose to add a content on-the-fly to immediatly "fix" the missing reply.

In an effort to enhance Agentic AI support, we are upgrading our entire platform to robustly accommodate a wide array of paradigms, delivering a fully featured, state-of-the-art Agentic AI design experience. We are introducing native support for OpenAI tools (“functions”), enabling you to seamlessly invoke both built-in tools and create your own. Key highlights include: - Self-Training Tool (Copilot-Style) Chatbots will automatically learn from human-to-human interactions, leveraging Tiledesk’s Human in the Loop features for continuous improvement. - Automated Suggestions for Humans Users will receive real-time suggestions from previously successful conversations or official documentation. - Custom Actions for Designers Designers can build and add their own actions to the Design Studio, providing greater flexibility in chatbot workflows. - Enhanced Code Action A new, more powerful Code Action with multi-language support is in development, offering a robust framework for custom actions.

Creating a Redis Action that you can connect to your cloud Redis account. It will allow you to save and retrieve your own data from a central Redis DB to implement persistency in your chatbots. This persistency will be cross-chatbots and cross-projects, going beyond the short-live persistency of your conversations attributes.

Some specific Actions will be developed to help users get the maximum flexibility in Smart Assignment (queues) management. Actions planned: Action "Assigned?" - [ ] Yes - [ ] No Action "Still in queue?" - [ ] Yes - [ ] No Action "Reset agent counter" Execute a specific "Flow + Block" when some agent is not available (event => Action)

Add support for Video to Whatsapp

Add support for native Location to Whatsapp

This will provide an easy way to automatically set the unavailabily of the agent when he closes all his chat instances.

Everyone will be able to build his own actions to enhance the designer capabilities based on their own business logic

A new multi LLMs action will be released in September that will allow support for: - Google Gemini - LLAMA / Mistral (Through Groq integration) - Anthropric (Claude models) - And more to come pre-release date: September (end)

Condition action will check also JSON objects. For example, on a "person" attribute, type JSON, you can check also single properties like "persone.age". Actually only scalar (string) attributes can be evaluated in a condition action. pre-release date: September (end)

Deploy your chatbots flows on a Twilio account and let your users talk with your chatbots. pre-release date: November

If the same user (or another user) opens the same automation/chabot on multiple pages/tabs, you must be allowed to continue the design only on the most recent instance. pre-release date: September (end)

Developing a scheduler for URLs/Sitemaps re-crawling. In this way you can always get an up-to-date Knowledge base when the resource is online and is subject to frequent updating

The Design Studio will integrate a console to read live debug logs during flows execution. This will greatly simplify debugging and development. A "Debug" Action will also be introduced, able to "log" in the console. Some Actions will have the opportunity to be disabled, so you can only activate those ones during the debug and they will automatically become disabled in production pre-release date: November

With this improvements the flow designer will be able to accurately configure what chatbots and blocks will be available for the human operators to be put back in the conversation. A human icon on the block will show blocks available to human operators. See images.

The "Replace chatbot" Action will be able to dynamically select with attributes both the chatbot and the action to execute. A “slug” property for the flows (aka chatbots) is introduced to uniquely identify each chatbot and simplify navigation, including import-export operations. This is necessary because relying on the chatbot’s name is risky, given that names can easily change over time.