Webinar Recap: How Nana Uses AI Agents to Transform Customer Experience

Most customer support teams grow by adding headcount. The math eventually stops working. Ticket volumes rise, quality drops, and agents burn out doing the same repetitive tasks on rotating shifts.

Nana, one of Saudi Arabia's largest super apps, found a different path. In Behind the Agent: How Nana Uses AI Agents to Transform Customer Experience, Karim Mustafa, Head of Customer Experience at Nana, shared how the company deployed AI agents across its support operation and what happened to their team, their costs, and their quality scores as a result.

Why AI Powered Customer Support Matters

Customer support is a quality problem before it is a cost problem.

The hardest part of managing a support team was not hiring people. It was training them to show empathy consistently, respond with warmth, and maintain quality across thousands of conversations daily. That is nearly impossible to scale with humans alone.

Nana started in 2016 as a grocery delivery pioneer in Saudi Arabia. Over a decade, the company expanded into a full super app covering laundry, car washing, food delivery, and more. As the product grew, so did customer contact volume. The team was running 54 support agents across rotating shifts just to keep up with demand.

The turning point came when Nana moved from FAQ bots and basic chat flows to a true front-end AI agent built on ElevenLabs. The agent handles customer conversations directly inside the Nana app. It resolves issues, answers questions, tracks orders, and escalates when needed. The quality was immediate and consistent.

The business impact was significant:

1. Nana reduced its customer support costs by 85%
2. Response times dropped to seconds

The agents were moved out of front-line support and into higher-value roles focused on improving the AI workflows themselves.

Several former support agents are now embedded in other departments, using their knowledge of customer pain points to improve operations across the business.

Behind the Agent: Nana

Before ElevenLabs, the team tried multiple CRM and AI providers. Each one promised automation and delivered disappointment.

Some vendors were charging a premium for solutions that were simply white-labeled ElevenLabs technology with added costs on top.

After switching to ElevenLabs directly, the results changed. The chat agent went live in the Nana app and is running in production today. A second use case using AI voice agents for outbound calls is also live and being expanded. One major internal project, which would have taken weeks to build through traditional workflows, was built and connected to all backend systems in a single day.

Karim described the shift in how the team now thinks about problems. When a new operational challenge comes up, the default question is no longer how many people they need. It is how quickly they can build an agent to handle it.

The remaining support team members were given a choice between fixed shifts or flexible hours tied to AI-focused KPIs. Every single one chose the flexible model.

Best Practices from Nana

1. Be multimodal. Nana found that their customers also wanted to type, not just call. Matching the channel to customer behavior made adoption easier and results faster.
2. Focus on prompt quality, not just technology. The AI will reflect how well you instruct it. Karim's team invested time in writing prompts that taught the agent how to show empathy and respond in a way that matched the brand.
3. For Arabic language deployments, use tashkeel. Adding vowel markings to Arabic text in the agent prompt significantly improves pronunciation accuracy and naturalness.
4. Move human agents upstream, not out. The best use of your remaining team is improving the AI, not replacing it. Give them ownership of workflow design and quality review.
5. Build for scale from day one. Once the first agent is live, the same platform can support outbound calls, internal workflows, ads, and cross-department use cases. Nana is now the department other teams look to when a new project needs to launch quickly.
6. Go directly to the source. Several vendors Nana evaluated were reselling ElevenLabs at a markup. Building on ElevenLabs directly gave Nana access to the latest models, lower costs, and direct engineering support.

Watch the full session

Watch the full session here.