Setting up AI Actions for your HelpMate (Beta)

With AI Actions, your HelpMate can go beyond just chatting - it can now do things. From checking order statuses to updating customer info, AI Actions let your HelpMate connect to your external systems and take action in real-time. ⚡️

Here’s how to set it up:

Where to find AI Actions

You can add AI Actions while setting up your HelpMate:

1. Go to the sidebar menu:
2. Trengo AI > HelpMates
3. Select an existing HelpMate, or click Create HelpMate
4. Once you’ve completed the Basic and Advanced setup steps, navigate to the new tab: AI Actions (Beta)

How to create an AI Action

Inside the AI Actions tab:

1. Click “New action”
2. Give your action a name - e.g. Check delivery status
3. Add an action description (AI Prompt)
   ‍This tells HelpMate when and how to use the action. Example:
   “Check the latest order status for the customer. Only respond with status, not full order details.”
   

‍‍

‍4. Choose an Endpoint Method

This tells the AI what kind of operation to perform with your system:

• GET – Use this to retrieve data (e.g. order status, user info)
• POST – Use this to add new information (e.g. create a ticket, book an appointment)
• PUT – Use this to update existing data (e.g. change an address)
• DELETE – Use this to remove data (e.g. cancel a booking)

💡 Not sure which one to pick? Start with GET to safely retrieve info before trying other methods.

5. Enter the Endpoint URL‍

This is the API URL your HelpMate will contact (e.g. https://your-system.com/api/orders)‍

6 Add parameters and headers if needed

These pass important details like order ID, customer email, etc.

You can define them dynamically — HelpMate will fill in the values during the conversation.

Tip: Only add what’s needed, and make sure to mask or validate any sensitive information.

Real examples to get you started

Not sure where to begin? Here are a few ways teams are already using AI Actions:

• Support teams use it to check refund statuses, delivery updates, or pull customer-specific warranty information.
• Sales reps fetch personalized pricing, stock levels, or customer contract details right inside the chat.
• Operations teams trigger actions like updating CRM fields, escalating tickets, or syncing external data.

Remember: You can define the rules - tell HelpMate when to use an action and how to handle the response. This keeps conversations safe, relevant, and helpful.

🚀 Try it out and start today!

Once your AI Action is set up, head to the Playground to see it in action.

• Make sure HelpMate triggers the action at the right time
• Check if the response is clear and helpful
• Fine-tune anything before going live

The best part? You don’t need to write code - just describe what the AI should do, and let it take care of the rest.

Start today and give your HelpMate the power to not just respond, but take action.

‍

AI HelpMate Dashboard explained

The HelpMate Dashboard is your window into HelpMate activity as you deploy them to assist in handling customer conversations. Currently, you can see up to 30 days of data in the Dashboard for all HelpMates or specific HelpMates.

Please note that only HelpMates published in AutoPilot mode will be tracked in the HelpMate dashboard.

HelpMate tickets

This element shows the total number of conversations containing a message from your HelpMates. This is inclusive of conversations that are escalated to human agents.

Assistance rate

This element shows the percentage of conversations that HelpMates have started handling and then passed along to a human agent. The percentage is taken from the total number of HelpMate conversations.

You can see the average number of messages a HelpMate exchanges before escalating the conversation, as well as how long on average the customer spent in conversation with the HelpMate before it reached a point of needing human intervention.

Resolution rate

This element shows the percentage of conversations that HelpMates have started and then closed on their own without passing them to a human agent. The percentage is taken from the total number of HelpMate conversations.

You can see the average number of messages a HelpMate exchanges before ultimately closing the conversation, as well as how long on average it took to reach a resolution.

Open ticket rate

This element shows the percentage of conversations that HelpMates are handling but have not closed yet due to lack of customer response

AI HelpMate handles 50%+ of your customer conversations, so you always keep support and sales moving! To get the most out of your AI HelpMate and start saving your team valuable time, check off each of these steps and build the AI digital teammate that suits your customers best.

1. Activate HelpMate. Make sure to accept the terms and conditions to get started.
2. Build your knowledge base. Collect documents on the subject you want HelpMate to know and communicate about with your customers. These can be files from your device, a connection to a help centre you've already built on Trengo, or webpages that HelpMate will read to get information.
3. Configure your settings. Supply your HelpMate with the tone of voice, service role, and business details that represent your company. This information will inform how your HelpMate generates responses to customer queries.
4. Test your HelpMate. Visit the Playground, ask your HelpMate the questions your customers often ask you, and fine-tune its settings to ensure high quality responses.
5. Connect and go live. Assign your HelpMate to a messaging or email channel, and watch it handle incoming conversations. Any conversations that it hands over to your human team will appear in your inbox like normal!

With Trengo AI Journeys, you can now create even better conversations and routing for your customers. Its no-code, no-certification setup allows you to start building your own AI Journeys right away. If you need an estimate of how much average time saved when you use AI Journeys, take a look below:

‍

AI Routing - On average 40 seconds saved per trigger

‍AI Routing identifies and routes the conversation to the right path.
Without AI routing, an agent needs to review each conversation, quickly assess the message, and determine whether they can respond based on their expertise or if it should be assigned to another agent or team. This process applies to reopened conversations, ongoing conversations with a history, and entirely new incoming conversations.
‍

AI Message - On average 20 seconds saved per message

AI converts your prompts into natural-sounding messages in the most relevant language for the conversation.

Without AI message, an agent needs to draft a response to each incoming message. In some cases. They may also need to translate the message into other languages before crafting and sending an accurate response back.

‍

AI summary - On average 60 seconds

AI breaks down your full conversation into a summary.

Without AI summary, an agent needs to reads through all the messages or emails and summarize it based on all the context.

‍

Per AI HelpMate Message - On average 60 seconds saved per message

AI HelpMate helps you to answer questions from customers on autopilot based on added knowledge.

Without AI HelpMate message, an agent needs to identify the question, search relevant documents for the correct information, check potentially the history, and sometimes consult with colleagues to gather additional knowledge. They may also need to translate the message into other languages before crafting and sending an accurate response back.

‍

Per AI Input Collection - On average 15 seconds saved per question

AI input collection gathers information from users through a natural conversation.

Without AI input collection, an agent needs to discover which missing information should be retrieved and should be asked for, or to collect more information to identify the lead. They may also need to translate the message into other languages before crafting and sending an accurate response back.

‍

Condition - On average 15 seconds

Conditions are criteria's that needs to match in order to direct the chatter towards the right path in the journey.

Without conditions, an agent needs to check if the conditions from the chat are true or not.


Label - On average 10 seconds

Labels help you to categorise and filter conversations.

Without labels, an agent needs to click on the labels, identify and select the label that applies to the conversion between a bigger list of multiple labels.

‍

Assign to team - On average 10 seconds

It assigns the conversation to the right team.

Without this, an agent needs to click on teams, identify and select the team that needs to take care of the right team.

‍

Assign to user - On average 10 seconds

It assigns the conversation to the right agent (user).

Without this, an agent needs to click on teams, identify and select the team that needs to take care of the right team.

‍

Assign Round Robin - On average 15 seconds

Round Robin evenly distributes incoming conversations, to ensure a balanced workload across all team members.

Without Round Robin, an agent needs to monitor every team member's capacity.  Checking if they are online/offline at all times and decide to who will be the new owner of handling the incoming conversation.

‍

Close ticket - On average 5 seconds

It closes the ticket.

Without having this automated, an agent needs to check if the conversation has been wrapped up and can be closed.

‍

Jump to step - On average 5 seconds

This action allows you to jump to a specific step in the journey.

Without it, an agent needs to understand that the incoming question should be forwarded to the right team or agent based of the question or new information which arrived in the conversation.


HTTP Action - On average 2 minutes

Use HTTP actions to send data to external services or APIs. The response from the HTTP action can be used in the next steps of the journey.

Without it, an agent needs check the collected data in an external system. Often it is already logged in, uses the search of the right order/booking information and check's the details of the order/booking and based on the gathered information, writes a message back to the customer based of the status.

AI CSAT - On average 20 seconds

CSAT allows you to ask how the customer has perceived the help during the conversation.

Without AI CSAT, an agent needs to paste in a pre-defined message into the editor and submits it. After the response got back in, they need to manually store it in a sheet or list to analyse it and keep the overview.

‍

Contact to group - On average 10 seconds

Add or remove contacts from designated groups based on specific goals, keeping lists targeted and up-to-date.
Agents need to assess each contact's eligibility before manually adding or removing them from groups.

‍

WhatsApp template - On average 40 seconds

Send targeted WhatsApp messages using predefined templates tailored to customer details, ensuring timely and personalised communication.

Agents must search for the appropriate template, add specific customer information, and send it back to the customer. For system-triggered actions (e.g., booking or order confirmations), agents manually locate and personalise the correct template for each interaction.

‍

Wait - no real-time saving

Introduce a controlled delay within the Journey, pausing for a specified time before proceeding to the next step.

Agents naturally experience pauses as they manage multiple customer interactions simultaneously, resulting in “waiting” between tasks.

‍

Media - On average 30 seconds

Send media back to the sender in the ticket.

Agents naturally find the correct media on their computer and then selecting or dragging the right media to the conversation.

‍

Log contact moment in CRM - On average 60 seconds

Log contact moment in the CRM sytems like Close.io, HubSpot, Pipedrive, Microsoft Dynamics, Salesforce, Zoho and push ticket information and ticket URLs into the CRM system.

Without this action, agents must search themselves in an external CRM system based on names, emails or phone numbers and log the right information afterwards.

‍

‍

‍

Journeys is Trengo's no-code customer journey builder that creates full-funnel conversation flows in 70+ languages in minutes, so your team can stop wasting time on admin work and start providing personalised, instant support.

Here are some benefits on using Journeys:

• Personalisation. Its built multilingual in over 70 languages, it will be able to offer personalisation across languages without needing additional agents.
• A proactive approach. With the assistance of AI, the conversation sounds more natural, the tone of voice and style of the conversation can be chosen during the set up.
• Smart routing. Unlike other bots, Journeys is able to do smart routing combination in minutes. Some of the capabilities are: assign to a specific team, route to a human agent when there's no available answer, automatically end the conversation and many more.
• Instant support on various channel. This feature will allow you to have 24/7 support for the customers. Long gone are the days when customers have to wait until the next opening hours to get an answer. On top of that, workflows can be embedded in various channel, such as: email, messaging (social media) and website widget.
• Round robin feature. You can also assign the ticket using the round robin feature to ensure that the team has to balance the team's workload.

Easy set up

Here are the steps to start using Journeys once you have them available in your account:

1. Check out the Trengo left side bar and click the Journeys icon

2. You will see this screen as the homepage of the Journeys and you may use some of the sample to start the templates but you can also create from scratch. Click 'Create new AI journey'

3. Choose the trigger. Note: once it's chosen and a journey is created, you cannot change the trigger anymore.

4. Once a trigger is chosen, then you can start creating a journey in the canvas

5. Once you choose 'Select channels', you can now start building the journey

6. Start adding the action as you go along, you can choose the actions provided in the menu:

7. Each and every action has a different function. The action is self-explanatory, for example, 'Send AI message', it will mean that we will send a greeting that's AI-powered. Let's use that as an example:

Once you choose the action, it will trigger a dialogue box on the right side of the journeys where you can fill your criterial for this particular AI message. We also have some examples for your inspiration. Once you're good to go with the instruction, don't forget to click 'save.

8. Click the + button to add more action in your journey. In this example, we'll use 'AI input collection' and you fill the right side column as well.

Once you click 'Save', you need to think of the next action. What should happen after the user input the post code? Do you have a specific team to assign or would you ask more question? In this example, after asking the postcode we'll assign it to the team who handles logistics. So, let's add one more action.

9. Click 'Assign to team'. Choose the assigned team on the dialogue box and to end the journey, click Save. You will see the 'End' at the bottom as per the picture below

Testing the journey

You can test the journey by clicking the 'Play' button on the top right

Once you clicked them, you will see a dialog box appear on the right side of your screen. In this box, you will be able to test out your own journey before publishing it. It's placed on the side of the journeys so you can easily adjust the action as needed.

Once you're good to go with the playground, you can click 'Publish changes'

‍

There you have it! Your first journey has been created.

If you need a video guidance, you can watch our CPO, Patrick, explained Journeys in details here.

‍

AI Labeling in Journeys is a new feature that automatically categorizes conversations, allowing you to streamline your support process while enhancing lead generation efforts. It provides analytics on the types of products most frequently requested during chats (interested prospects/leads), enabling you to gain a comprehensive analysis. By analyzing closed conversations and tagging them with relevant labels, this helps you organize topics, prioritize responses, and gain valuable insights into recurring issues. With AI Labeling, you can save time, improve efficiency, and make better decisions based on data-driven insights.

How to use AI Labeling

Step 1: Go to Journeys and select the “Conversation closed” trigger:

Step 2: Add a Label Group with the following details:

Group Name: Define the topic or category.

Group Description: Provide the AI model with a short explanation as a prompt about what the labels in this group represent, such as 'Analyze the sentiment' or 'Identify which products or services the chatter is interested in.

Action: Add the labels which are matching the label group.

Step 3: Decide whether only one label from the group can be tagged or if multiple labels from the group can be applied.

‍Step 4: Save the Journey with AI labeling as action and let the system analyze all closed conversations to identify matching topics.

‍AI Labeling Dashboard Insights

‍You can access your AI Labeling data in Journeys side dashboard, which provides an overview of the most frequently analyzed topics. This dashboard makes it easy to spot trends, such as an increase in questions about specific products or recurring issues with services. These insights allow you to take meaningful action, whether it’s improving your support content, adjusting internal processes, or proactively addressing customer pain points.

AI Labeling is an essential tool for any team looking to enhance their support and sales operations. By automating the labeling process and delivering clear insights, it helps you stay organized, make informed decisions, and better understand your customers’ needs.