We recently added a survey feature that appears after every AI response; some of our users liked it, and some wanted to make it optional and localizable.

Ability to disable AI response survey

We hear you. With the latest update, you can now disable AI surveys by following the Chatbot Manage View -> AI settings -> AI response surveys path.

AI response surveys are enabled as default, if you don't want it, you should disable it manually.

Ability to hide AI indicator

Also, one popular feedback suggestion was to hide the AI indicator below AI messages. We made this optional, too.

You can again disable AI indicator settings in the same view above by following Chatbot Manage View -> AI settings -> AI badge with AI messages path.

Localization support for the AI response survey

You can easily add localization for the AI response survey in your preferred language or change the micro copies for these actions.

Follow Chatbot Manage View -> Localization path to set these values as you wish.

Thanks for sharing your feedback with us. We'll keep improving Mevo together.

Hello,

After a while, Mevo's first major release is here, and it's just a beginning. We've addressed lots of reported bugs and added new and exciting features that make your life easier, and we'll keep doing it.

We don't want to bore you, so here is a summary.

A single user can have multiple conversations

Now, all conversations are standalone entities. A single user can have multiple conversation items in the inbox. Users can also see their recent conversations when they revisit your webpage.

Conversation statuses

All conversations have a status property, you can filter conversations by status and the default filter value is open.

Also, you can easily mark as closed, and re-open conversations from the session header.

New user events

After the first AI response, users will see a feedback question in the chatbot like below.

According to the user response, you’ll see user events in session detail, such as user finding the AI answer helpful.

Or User demanded human agent support.

Improved human communication flow

We introduced the email replies feature a while ago. It was a small and good step, but we received some problems and complaints. The biggest one was that when you reply to a conversation, if a user sends a new message to the chatbot, AI gets involved in the conversation again, which makes things complicated.

Now, when a user demands human agent support, you’ll receive an email like the one below. Once you are involved in the conversation and reply via email, AI won’t involve you in the conversation again.

And, when you answer the user's question from the dashboard, Mevo emails your answer, like below, to the user with a link to the chatbot.

The user can continue the conversation with you by clicking on the Reply button. They'll see the chatbot in full-size webpage mode with your answer in the conversation.

And, if a user sends you a message, this flow will be repeated again and again until you mark the conversation closed.

A few more

You will notice some minor things when you start using Mevo today, if message sends from AI, the sender avatar and message bubble color is different than the messages sent from human agents.

You'll see a friendly warning if you close the conversation before replying to the user who demanded human support.

What's next?

Feel free to share your feedback with us on mevo.frill.co and hi@usemevo.com address. We've a long way to go and exciting features await. Thanks for being a Mevo user.

In this article, as a founder, I’ll mention our new vision about the product, our new roadmap, how we will keep you in the loop while prioritizing things, and how it will affect you as a user.

A short history of Mevo

In October 2022, Mevo was a live chat solution, planned as a Crisp alternative, even the domain was “mevo.chat” instead of “usemevo.com”. We had a working alpha version, then something big happened, and we decided to jump on to the AI train. We turned the platform into a dedicated chatbot training solution, and somehow we decided to add a rule-based chatbot because we wanted Mevo to be a chatbot generator. So whether you need AI or a traditional rule-based chatbot, if you use Mevo, you should have everything you need in one platform.

We released our first version in April 2023, and after a few months, in September 2023, we released our AppSumo deal. After 18 months, AppSumo is still a vital customer acquisition channel for us and will continue to be so for a short while.

What’s the current situation?

We recently released our new interface, which makes things easier. We introduced the New Inbox, Notification Receivers, Tags, Email replies when AI responses are insufficient, and more.

But we know there is still a long way to go. There are some known problems, and there are some processes to be improved, (we know our customer support flow should be improved, it’s ironic while we’re providing a solution to make customer support flows better, but it’s also an opportunity), white labeling for agencies, live-chat support, critical integrations like Whatsapp, Discord, Telegram, Shopify, the option to have other LLMs, and more.

You should have noticed that our latest release is December 2024. Of course, we shipped many minor fixes after that date, but there were no significant features. Even some people start to think, isn’t it Mevo maintained yet?

The reason was planning. We were planning and identifying the product pain points. We read the Discord messages, Frill feature requests, AppSumo questions, reviews, and emails from our users. It wasn’t easy to understand and decide which should be part of our new vision and which would not.

We know the pain points now, the things we should improve, the things we don’t even need in the product, the strengths of our product, and the weaknesses. Of course, we also know this is a dynamic process, and we should keep you in the loop, and we will.

What’s next?

As I mentioned, customer support is one of our heaviest tasks, too. Some of you have already noticed late replies to your emails, lack of information about the status of your requests, and sync problems that occur because of having multiple support channels (emails, Discord server, app Sumo questions, and even social media accounts).

We’ve decided those are the pain points we want to focus on. If we can’t solve our customer support issues, how can we claim to do this for our customers?

Even if it’s impossible to quickly handle all the customer queries due to the increased number of questions, we can at least provide an elegant way to follow that and create a transparent process. We can define support tiers, distribute the workload between agents, and try to find better ways to use AI.

And that’s precisely what we’re going to do; We’re going to turn Mevo into an end-to-end customer service platform.

Your customer asks a question. Your AI agent first attempts to answer using your internal or public knowledge base. If the customer isn’t satisfied, they can seamlessly transition to live chat. You can then convert this conversation into a customer support ticket, assign it to the relevant team, and your customer will receive notifications about each update. You will have an internal ticket management system and CRM module in minimal form.

How will it affect you as an existing Mevo customer?

We know some customers use Mevo to create general-purpose chatbots so that these changes won’t affect them too much. We’ll keep rule-based chatbots available because we’re still convinced of their value for customer support flows. We’ll find a way to fit them into the single chatbot flow with AI and live chat features.

However, there may be some features we’ll remove because they don’t fit with our new vision anymore;

Custom domains for standalone chatbot web pages. Offering a standalone page won’t fit with our new product vision, even if it’ll be technically possible to provide chatbots by using it as a link like bot.usemevo.com/bot-id, linking custom domains to them won’t be an option.

Ability to provide OpenAI key support. Therefore, Assistants API support will also be deprecated. We want to make the platform LLM agnostic because different models perform better for various kinds of problems, and this isn’t possible when relying directly on OpenAI.

Web scraping features. We don’t believe this offers as good a value as it sounds. Instead, we’ll introduce public knowledge bases that you can add to your content as support articles, and your chatbot will rely on that information specifically designed for support purposes.

Actions you should take

All of those features will sunset on 1 April 2025, so if you use one of those, you should take some actions, like updating the places you use a custom domain for your chatbot with our internal bot URLs, removing your OpenAI key and Assistant connection, and training your chatbot again after you remove the connection. You can continue using your chatbots using your Mevo credits in your plan. Your chatbot will work smoothly (probably better than before) if you provide all the same training content.

We understand that it is unpleasant for you to lose these rights that you have, and we think it is fair that you gain something in return. To compensate for the removal of these features:

For each lost custom domain feature, your training character limit will increase by 5M.

Losing the OpenAI key feature will grant you an additional 5M character limit.

Losing the web scraping feature adds yet another 5M characters.

Additionally, you’ll receive three extra user seats as compensation for these changes.

We will also offer optional credit plans that you can choose to use once you run out of token credits. These plans will be significantly more accessible and discounted for our existing customers.

We’ll keep you in the loop

You already know we’ve been using a feature request platform, Frill, for a while. We’ll keep using it, and when you check it, you’ll see our updated roadmap. You can still create feature requests, vote on the ideas you like, and leave comments on the ideas.

We’ll also announce a webinar soon about the upcoming changes, and it’ll include a Q&A session. If you have any comments or questions, you can raise them directly with the founder.

We’ll also announce a meeting day, that you can book a 15-minute call slot with 1-on-1 with the founder, this will be a survey-like conversation about the product, what are your favorite features, what features are you missing in Mevo, etc. And if you join these calls, you’ll gain rewards like extra tokens, credits and more.

Final

Thanks for reading this long post.

We believe this vision shift positions Mevo as a powerful alternative among customer service platforms, ultimately simplifying and enhancing your experience. If you have any comments, feel free to share with us by using hi@usemevo.com

Furkan Başaran,

Founder

Our users who have multiple chatbots or create chatbots for their customers need a better filtering mechanism to manage their chatbots.

This new and simple feature allows you to filter chatbots according to the tags you add quickly.

All you have to do is go to the Tags screen from the menu on the chatbot management screen and add a few tags to your chatbot.

Then, you can see these tags on your chatbot on the homepage and filter your chatbots according to these tags.

You can edit the existing training resources that are supported now. The supported training resource types are products, questions and answers, and texts.

Please note that after editing, syncing may take up to 2 minutes.

Hello 👋

This was our feature request board's second most anticipated feature, and it's here! If the user cannot get enough information from AI, you can manually email the user a correction or manually answer within Mevo.

In Inbox view, if the lead is captured for conversation, you will see a text input at the bottom of the screen, as shown below.

You need email credits to send email replies to users via Mevo. Email credits are added to all plans available in Mevo; you can see how many email credits you have by visiting the Settings -> Subscription view.

The email will look like the one below when you reply to a conversation via Mevo.

As you can see, we're improving our product based on your feedback and comments. Keep suggesting features you're missing in Mevo, and we'll try to implement them.

Thanks for being part of the Mevo community.

As you already know, we used to calculate chatbot usage limits using tokens. However, since this is a quite complicated metric, it wasn't easy for most non-technical people to understand. Mevo is not for technical people; it is for everyone who wants to build a chatbot, so we decided to replace tokens with message credits.

How was the conversion done?

After many complicated calculations, including the average token size per message in Mevo's database and the average input and output token spending per conversation, we found the perfect value for the equivalent of 1M tokens and 1500 messages.

It will count user and AI responses as single message credits; some messages longer than the average word count can spend multiple credits. We'll also work on some features so that you can easily see all the credit spending in a conversation.

New default model, GPT-4o-mini

Another important topic is the default model. We were using GPT-3.5-turbo as a default model from the beginning. However, there are more powerful and accessible models now. They have more input and output token limits and better benchmark values compared to GPT-3.5-turbo, which is a relatively old model.

Source: (https://context.ai/compare/gpt-4o-mini/gpt-3-5-turbo)

Newly created AI chatbots will use GPT-4o-mini starting today. In the upcoming days, we'll also automatically migrate all chatbots to GPT-4o-mini from GPT-3.5-turbo. If you use another model, you should not do anything; this change will only affect chatbots using the default model.

New pricing plans

We're also introducing three new subscription plans. You can check and see the details from our upgrade window, which you can access within the Mevo dashboard. Our landing page also will be updated soon.

If you have any questions or suggestions about those changes, please contact us at hi@usemevo.com.

No matter how much training data you provide, it is inevitable that a question will come up on a topic that you do not have training data.

Continuously providing new training data is the best approach to increasing the accuracy of your AI assistant's answers. With this new feature, you can quickly create a question-and-answer pair for the question your assistant could not answer while reviewing the conversation.

How to use it?

Go to your inbox, click on any conversation from the conversation list, and choose the message you want to create a question-and-answer pair by right-clicking it.

You can only right-click the messages the user sends, not the AI.

Hello 👋

OpenAI and Assistants API integration is one of our most used features. We recently received feedback from many users to improve these features, especially after our new interface update. Based on this feedback, we implemented some changes, and now OpenAI and Mevo syncing is more powerful and accessible than ever.

The best part about binding your OpenAI key to Mevo is the freedom to use the latest language models without being stuck with token limits. However, this integration process could have been confusing in many cases. Let's quickly summarize what we did to improve things.

The organization-level OpenAI key binding feature is no longer available

The only way to bind your OpenAI API keys is to bind a private key to the chatbot. Suppose your organization already has an OpenAI API key bound. In that case, our system will automatically add it to all your existing AI chatbots that do not have a dedicated OpenAI API key.

If your current Mevo plan cannot bind a chatbot-specific OpenAI key, don't worry; the chatbot-level key binding feature is now included in every package with an organization-level key binding feature. So, there is no loss of rights for anyone.

Better Assistants API synchronization

Another change relates to Assistants API; previously, if you connected the assistant you created in OpenAI to Mevo, your training data in Mevo did not affect your assistant; chats bypassed Mevo and interacted directly with your assistant. Any training data you added after connecting your assistant will be synced with your OpenAI assistant.

If you need help with these changes, you can reach us at hi@usemevo.com.

Thanks for reading.