From AI Complaints to 67% Deflection: How Innovorder Scaled Support Across 20+ Product Modules with Pluno

Aurélien Wei

I'm Aurélien. I'm 26. I'm currently working at Innovorder. So, Innovorder is an 11-year-old startup that sells and is building a SaaS product which helps restaurants to digitize their processes. So, we have a complete suite of solutions, like POS systems and kitchen display systems.

We also have AI scan products for big companies, big restaurant companies. And this year we had 12 million of turnover, and we have a growth of 30% per year. So we have to find solutions to help our customers after the sale.

Aurélien Wei

I'm the team leader of the customer care support at Innovorder. So we are 12 people in the customer care service. We have nine N1, level one agents, and we have two level two agents. Recently we recruited three people to have more senior people in the team.

And we are handling like 3,000 tickets per month. It depends on the seasonality, because we have more demands in January and September for some customers. For others it's more for summer, or when they are reopening the restaurant.

Actually, we have only French customers, but this year we will have some Spanish and Italian customers. It's the year of internationalization for us, and we contracted with a big company in cruise ships. So it will be our first big international customer.

Syed Ali

And these tickets that you get, are they complex tickets or are they mostly repeat type of questions?

Aurélien Wei

Basically, we have three types of tickets. We have 33% of issues and bugs of the solution, not only on the software but also on the hardware, because we are also selling hardware. It's a big, big type of ticket. After that, we have the rest, 66%, of questions or setting tasks for the customer.

And the restaurant companies, they have a high turnover of people. It's not jobs where you stay 10 years or 15 years. So the people that are working in restaurants are changing often. So we have some repetitive questions, repetitive trainings, reminders of how to create a product, how to connect the printer, some basic questions. So we have some repetitive demands by phone or by chat, WhatsApp or email.

Syed Ali

And what was the main problem that you were facing, that you thought, I should look for a solution now, or maybe AI can help me with this?

Aurélien Wei

The main thing that we are looking to AI solutions for is that our suite is really complex. We have maybe 20 different solutions. So we have a lot and a lot of functionalities. We have a lot and a lot of possibilities to change the functionalities.

So the main thing is to find the right documentation, or to find answers to the question of how to do this, how to set this up, how can I find a solution for this bug or this hardware issue, because we have different hardware brands. So we have a lot, a lot, a lot of information and need of knowledge, and that was the main issue: to find it.

So first, we were searching for an AI solution just for the knowledge, like a Slack bot or a knowledge manager, but connected with our solutions on Notion and Zendesk, to find the right answer. And after that, we saw that it's quite repetitive to go to another tool to find a solution, and after to copy that and come back to Zendesk. So for the agents, we were focusing on finding a solution that is integrated into Zendesk.

Syed Ali

Were you experiencing any customer complaints before implementing the AI solution? Was it taking you longer than expected to reply to customers, or can you talk about any problem?

Aurélien Wei

Yeah. Yes, we were taking more time to answer, and the answers were not as good as today with the AI. And also, we tested a lot of AI solutions that were not quite good, with hallucinations or not finding the right quotation.

Syed Ali

Yeah. Can you tell me about these solutions that you tried initially, the first time when you went to the market to look for a solution? What did you try, and how was your experience?

Aurélien Wei

Yeah. So we tried maybe five or six solutions before. Each of them was focusing on one issue or one specific solution, and not on all the things that we were looking for. So maybe we had one that was only on the copilot solution, but we also wanted to have an autopilot for auto-reply. So this one was not good. Another one was only on the auto-reply and not on the copilot.

The other one was not connected to our internal tools. It was only PDF and text, and we had to copy and paste all the documentation into the tool. The other one was only a chatbot that we had to add to our website. So it was mainly little solutions, but there was not a big one that answered all our issues.

Or the big one that we found, the budget was stratospheric. I had an American one that told me that it's 1 million a year to have this solution. But we can't. We are a little startup, a little company, and we are not like big companies. We have a limited budget, so we want a tool that is effective, cost-effective and adapted to us. Yes.

Syed Ali

Okay. And then how did you find us?

Aurélien Wei

Good question. If I remember, maybe it's Alex that contacted me on LinkedIn to present Pluno to me, only on the escalation to Jira. It was the first objective. And after, in the first meetings, I shared my experience with AI solutions on the copilot and auto-reply, and we worked together to create...

Syed Ali

And how was it for you to start, the process from start to finish?

Aurélien Wei

From the start, it was a real pleasure, because the team was really available to help us to integrate. All the demands that I made were included. I asked for some specific features, and by the next day, the week after, we had the feature live. So it was quite impressive to have such an effective team.

And today we are less in contact, because the tool is running, but at the beginning it was quite a pleasure to share feedback and have an effective solution.

Syed Ali

And what are you using it the most for? Which part of your support is Pluno used the most for?

Aurélien Wei

The most used is the AI bot, the AI auto-reply for customers. For the moment, we don't have any complaints from the customers about the AI, and it's one of the most impressive things for me. Because with all the solutions that we used before, yes, it was AI, but it was not good AI, so customers saw them and complained about this.

So this is one of the main KPIs for me: it's whether we have complaints about the AI. If not, it's a good AI.

Syed Ali

And how is the experience with the team when you need a new feature? Can you request it, or...?

Aurélien Wei

Today we are more taking what Pluno is giving us, and we are asking less. But if we need a feature, the team is answering very fast to get it, in the month maybe. Or if it's in the roadmap, we will see it later. But if it's not, and it's a little thing, we have the feature really fast.

Like the last one we asked for: it's a button to fill in the ticket, so to rename the title, fill the fields on the left, and it generates a close-and-resolve email. So we asked for it, and maybe one week after we had it, and I'm using it every day.

Syed Ali

That's very good. And metrics-wise, can you draw some comparisons?

Aurélien Wei

In one year, we have plus 15 points of CSAT, the customer satisfaction. I don't think the 15 comes from the AI, but I think it's part of one of the solutions that we provide to our customers to have faster answers. So in a way, it's not. Yes, so the customers have the first answer faster. So yes, in a way it's better.

Syed Ali

How long did it take before to answer a question when it came in, versus now?

Aurélien Wei

Before, on the direct channels, it was maybe from 20 minutes to maybe 1 hour. Now the AI is answering in one minute to give an answer. So yes, it's a better result. And after, if it needs escalation to a human, we are now at five, six minutes for a human to come in.

And with Pluno, when it escalates, we have better visibility, because it provides context, some first answers, and asks questions to the customer. So the human comes in and has all the context to give a solution.

More impressive is the onboarding of the new agents. I recruited three people last month, and in one week, two weeks, three weeks, they are already resolving tickets and answering the customer. Before, they were waiting maybe six weeks, eight weeks before knowing the solution, finding the solution in the documentation, writing good answers with no mistakes in the French language and spelling. So on this, the AI is giving us faster onboarding, and so it's better.

Syed Ali

Yeah. As a recommendation, if someone hasn't tried it yet and they're in a very similar situation to yourself, what would you say to them?

Aurélien Wei

Test it and you will see the results directly. You make it in less than one hour. You can use your Zendesk, give some prompts and some guidelines, and directly you have results. You have auto-reply. If you test it for one day, you have results.

Syed Ali

You're saying the solution speaks for itself. So all they have to do is give it a try and they will see.

Aurélien Wei

Yes. And it doesn't need days and days of settings and testing and coming back. It's directly good. It was one of the good things that we had. When I shared it with the CEO of the company, he just tested it one time, one question, and he saw that the AI was impressive and good.

Syed Ali

Is your team also happy with it?

Aurélien Wei

Yes. Yeah, sure. The support team is happy, and also the other teams, because they can share the auto-reply. They can share it with their customers, with all the customers, and they are happy to give instant replies to customers.

And they also use it. Other people in the company use it as a training bot: if they have a question, they ask Pluno directly, and it gives them the answers on how to deal with this, how I reconnect the printer, how I create a product in the back office, and all these things.

Syed Ali

And you mentioned your knowledge is scattered in a lot of places. Is Pluno able to accurately get this information from different places?

Aurélien Wei

Yes, it was. Yes, it's a good thing that it can have a good range of the documentation. So it can go to Notion, to Jira, to past tickets, to our internal documentation. And so it takes what it needs, and after, it makes a mix of it to give a proper solution.

And also, it gives the link of the documentation. So if we see that the documentation is no longer good, we can change it, or if we see that it didn't find anything, we have to write the documentation.

But the good thing is that it's trained on past tickets. We don't have time to make the documentation every day, to add it, remove it, update it. But 99% of the time, we are giving the solution on the tickets, with the emails or the phone. So Pluno can find this in past tickets to give to new tickets. So we don't have to write the documentation like before.

Syed Ali

And because of this, do you think it's also capable of handling complex support tickets?

Aurélien Wei

Yes. Because for complex support tickets, it can find similar past tickets and give some guidelines for the human agent to check, or to diagnose something. And maybe we don't have one documentation for this specific type, but we have documentation for similar problems. But the human will not check each documentation each time to find this. So Pluno is doing this lookup.

I'm quite impatient to see the new features of Pluno, and to see in the next months, next years, what it will be like. I was among the first customers, when it was just one page, and now we have a full back office, in a really short time. So we started in September, we are in February, so six months after, and we have a complete back office.

If I compare with Intercom, which we tried, it's just really impressive to have a similar and better result with Pluno, which is a six-month company, compared to Intercom, which is a big one and was not working at the price we wanted.

Syed Ali

Deflection-wise, how many tickets are now automatically handled by Pluno?

Aurélien Wei

So in the last month, from 10 January to 10 February, we had 380 AI-resolved tickets. So it is 67% of the tickets that the AI resolved. But today we have only the auto-reply for the chatbot and WhatsApp. Next month we want to add it to the email tickets, so we will see after that. Before the end of the year, I want to have it also on the phone, because we have a lot of repetitive questions on the phone.

In another metric, it's not 100% Pluno, but it's helping us. So we have one metric that is the number of tickets we have each year in comparison with the use of the Innovorder solution. And how do we calculate this? We calculate with the orders, because we have customers that are making orders on our systems.

And if we compare 2024 versus 2025: in 2025, we had 5% fewer tickets than in 2024, but we had 25% more orders. So we have a lot more customers, but we have fewer tickets. So it's a good thing to have fewer tickets in support, because a good support or a good company is zero tickets, zero support.

So it's one of the main metrics that we are seeing now, and it's helping us with this through auto-reply, because the auto-reply tickets are not counted in the tickets on the desk. And so the customers are more autonomous, because the AI is giving instructions and training to our customers. So it also helps us to make the customers more autonomous.