From AI Hallucinations to Zero Complaints: How Innovorder Deflected from 0% to 67% of Support Tickets with Pluno
A conversation with
Aurélien Wel
About
Innovorder provides a complete SaaS suite to help restaurants digitize their operations: POS, kitchen display, care system, and AI-powered tools. Learn More
Industry
Restaurant Tech / SaaS
Company Size
51-200
Support Tool
Zendesk
Key Features
Deflection AI in Web chat & Whatsapp support
AI Copilot
AI Tagging & Field Filling
🚧 Core Challenges
12-person team overwhelmed by 3,000+ monthly tickets across 20+ product modules
Knowledge was scattered across Notion, Zendesk, Jira, and past tickets, and agents spent significant time just finding answers
No effective deflection. Previous AI tools hallucinated and generated customer complaints
Response times on chat and WhatsApp ranged from 20 minutes to over an hour
New agent onboarding took 6–8 weeks before they could work independently
✅ Solutions
Deflection Agent auto-resolves repeat questions on chat and WhatsApp before they reach the queue
AI Copilot helps agents diagnose complex issues and draft accurate responses across 20+ modules
Multi-source knowledge sync pulls from Notion, Jira, past tickets, and documentation. No separate knowledge base needed
One-click ticket closing auto-fills fields, renames tickets, and generates closing emails
Under 1-hour setup: connect Zendesk, add guidelines, and auto-replies start immediately
Innovorder is an 11-year-old French SaaS company helping restaurants digitize their entire operation: from point-of-sale and kitchen display systems to AI-powered tools for large chains. With €12 million in revenue, 30% year-on-year growth, and expansion into Spain, Italy, and its first major cruise ship contract, the company is scaling fast. At the center of their customer experience is a 12-person support team led by Aurélien Wel, managing over 3,000 tickets per month for a complex, multi-module product. Before Pluno, their AI experiments had only made things worse.
THE CHALLENGE
Complex Product, Scattered Knowledge, Failed AI Attempts
Innovorder's product suite spans 20+ modules with thousands of configuration options. About a third of their 3,000 monthly tickets are hardware and software bugs across multiple product areas and device brands. The remaining two-thirds are how-to questions and configuration requests, made relentless by high staff turnover in the restaurant industry, which means customers are constantly onboarding new employees who need to learn the system from scratch.
But volume alone wasn't the core problem. The team faced compounding bottlenecks:
Knowledge everywhere, answers nowhere: The information to resolve any given ticket might live in Notion, Zendesk, Jira, or buried in a past ticket. Agents spent significant time tracking down the right answer before they could even start responding, and often sent back partial information.
Slow response times: On direct channels like chat and WhatsApp, response times ranged from 20 minutes to over an hour.
Months-long onboarding: New agents needed 6–8 weeks before they had enough product knowledge to independently resolve tickets and write accurate responses.
Every AI tool hallucinated: Aurélien evaluated five or six AI solutions before Pluno. Some offered only a copilot with no auto-reply. Others only supported PDFs. Most had no real Zendesk integration. And one American vendor quoted $1 million per year. The common thread: customers could tell it was AI, and they complained.
THE SOLUTION
One AI Layer That Connects Everything, Without a Separate Knowledge Base
Pluno first reached out to Aurélien on LinkedIn about Jira escalation support. After a few conversations, the scope expanded into a full integration: AI auto-reply for customer-facing channels, an agent copilot, and deep connection to the knowledge Innovorder already had scattered across its tools, no migration required.
Deflection Agent: Complaint-Free Auto-Reply on Chat & WhatsApp
Pluno's Deflection AI handles incoming tickets on chat and WhatsApp, the two channels where customers expect instant answers. Unlike every previous tool the team tried, Pluno's responses don't trigger complaints. For Aurélien, this is the headline metric: zero customer complaints about the AI since launch.
Multi-Source Knowledge: No Separate Knowledge Base Required
Previous tools required Innovorder to manually copy documentation into a new platform. Pluno solved this by connecting directly to what already existed: Notion, Jira, past tickets, and internal documentation. It synthesizes answers from across sources and links agents to the original material so they can verify and update documentation as needed. Since 99% of solutions were already captured in ticket history, Pluno could surface institutional knowledge that was never formally documented.
AI Copilot: Diagnosing Complexity, Accelerating Agents
For tickets that need a human, the Copilot provides context, suggests diagnostic paths based on similar past tickets, and hands off cleanly. New agents see suggested responses that help them ramp faster, while experienced agents use it to explore troubleshooting paths before escalating to engineering.
One-Click Ticket Closing: A Feature Built on Request
Aurélien requested a button that auto-fills ticket fields, renames the title, and generates a closing email — all in a single click. It was live within a week and is now part of his daily workflow. This kind of responsiveness from the Pluno team shaped the product to fit Innovorder's specific workflows throughout the partnership.
THE RESULTS
A Team That Scaled Without Scaling Headcount
Innovorder has been using Pluno since September 2024. The impact across the following six months has been immediate and measurable — and the team is still only using auto-reply on two channels.
⚡ 0% → 67% Ticket Deflection Rate
On chat and WhatsApp alone, two-thirds of incoming tickets are now resolved by AI without reaching an agent. Email and phone are not yet on auto-reply — once added, deflection is expected to climb further.
🚀 Response Time: From an Hour to Under a Minute
Chat and WhatsApp responses that previously took 20 minutes to over an hour now happen in under a minute. When tickets do reach a human, Pluno hands off the full conversation context, a summary, and any clarifying questions already asked — so agents resolve, not re-read.
📈 New Agents Productive in 1–3 Weeks Instead of 6–8
Three recently hired agents were resolving tickets independently within their first few weeks — a process that previously took six to eight weeks. The Copilot's suggested responses and knowledge surfacing compressed the learning curve dramatically.
💬 Zero Customer Complaints About AI
Every previous AI tool generated customer complaints. Pluno has produced none. For Aurélien, this is the number one KPI — if customers complain about the AI, it's a bad AI.
📊 +15 Points on CSAT
Customer satisfaction scores rose 15 points over the past year. Aurélien credits faster, more complete responses as a meaningful contributor alongside other improvements.
📉 25% More Customers, 5% Fewer Tickets
Innovorder tracks tickets per order processed as their north star metric. In 2025 vs. 2024: 25% more orders, 5% fewer tickets. More customers generating less support load — the definition of a great support operation scaling right.
Usage has even spread beyond the support team. Other departments now use Pluno as an internal training tool — asking it how to set up products, reconnect printers, or navigate the back office. Questions that would have flooded the support queue are now self-served.
"My CEO tested it once, asked one question, and immediately saw that the AI was impressive. That was it."
THE TAKEAWAYS
"Test It. You'll See Results on Day One."
When asked what he'd tell another support leader considering AI, Aurélien's answer is direct: skip the weeks of configuration, separate knowledge base maintenance, and inconsistent results. Pluno's setup took under an hour — connect Zendesk, add some prompts and guidelines, and auto-replies start immediately. If you test it for one day, you'll see results.
What sets Pluno apart, in his view, is the combination: auto-reply that doesn't hallucinate, a copilot that agents actually use, and a team that ships features fast. When Aurélien requested new functionality, it was often live within a week. The early collaboration shaped the product to fit Innovorder's exact workflows.
For a company about to serve international customers — including a major cruise ship operator — maintaining a high-quality, complaint-free AI experience isn't optional. It's how Innovorder scales without proportionally scaling headcount.