Hostinger's Kodee saves over €9M a year, resolves 75% of support chats and replies in 9 seconds

Hostinger's AI Agent Kodee Cuts Support Costs by €9M/Year and Fully Resolves 75% of Chats

Hostinger upgraded its AI chat assistant, Kodee, from a quick-answer bot to an action-taking agent. It now completes real tasks in client accounts and fully resolves 75% of conversations, saving over €750,000 per month in operational expenses.

For support leaders, this is the benchmark: faster responses, higher resolution rates, fewer repetitive tickets, and a leaner team focused on complex work.

What changed

Kodee doesn't just explain how to fix an issue - it fixes it. It can perform 350+ actions across hosting, WordPress, domains, payments, and security. Clients describe their goal; Kodee executes.

When Kodee takes action instead of only giving information, resolution rates and customer satisfaction climb by about 4%.

Numbers that matter

• 750,000 conversations in August; 75% fully resolved without human help (up from 50% at the start of 2025).
• Average first response time: 9 seconds (down from 28 seconds a year ago).
• 93% of all support requests answered in under 2 minutes (up from 77% last year).
• €750,000+ saved in August alone; projecting €9M+ in annual savings.
• Team composition: Customer Success headcount down 20% to 312, while a team of 14 supports and improves Kodee.
• Coverage: 50+ languages; over 4 million customers across 150+ countries.
• Without automation, the company estimates it would need 700-800 additional specialists to meet demand.

Why this works for support teams

• Speed: AI responds in seconds and handles unlimited concurrent chats.
• Scalability: Ticket volume has more than doubled in four years; automation absorbs the load.
• Quality: Action-first flows beat info-only flows on resolution and CSAT.
• Cost: A small enablement team replaces hundreds of incremental hires.
• Localization: Multilingual AI reduces the need for separate language queues.

What Kodee can do (model these categories)

• Web hosting: Site migration, performance optimization, backups.
• WordPress: Create posts, manage WooCommerce products, pull store performance reports.
• Website Builder: Edit products, update SEO, change images.
• VPS: Check server health, secure access, run malware scans.
• Domains: Register domains, manage DNS, configure settings.
• Payments: Purchase services, upgrade plans, review invoices and refunds.

Under the hood, Kodee connects to infrastructure via the open-source Model Context Protocol (MCP), which enables controlled, auditable actions across tools and data. Learn more about MCP here: Model Context Protocol.

Team impact without layoffs

Hostinger reports the Customer Success team is smaller this year, but roles shifted rather than being cut: repetitive tasks went to automation, while specialists moved toward complex, higher-value work. A compact team maintains the AI system, expands the action library, and monitors quality.

Playbook you can use now

• Define an action library: List the top 50 repeat tasks (e.g., password resets, DNS edits, refund checks, plan upgrades). Convert each into a one-click or AI-driven action with clear preconditions.
• Add lightweight approvals: Require client confirmation for sensitive changes (DNS edits, plan purchases). Keep the path to "yes" one tap away.
• Route by intent: Send "can-take-action" intents to the AI first; send complex or high-risk intents directly to humans.
• Measure the right metrics: Track Action Rate, First Contact Resolution (FCR), CSAT by action vs info-only, cost per resolution, and seconds-to-first-response.
• Build multilingual coverage: Use AI translation to respond in a client's language and reduce specialized queues.
• Keep humans close: Offer instant handoff for preference or complexity. Make it obvious and easy.
• Operate with guardrails: Role-based permissions, audit logs, rate limits, and spend caps for purchases or upgrades.
• Upskill your team: Move specialists into QA, automation design, and complex case resolution. Train everyone on writing action-ready prompts and testing flows.

What this means for your roadmap

• Shift from "answering tickets" to "completing tasks."
• Prioritize actions that reduce escalations and refunds first.
• Design for seconds, not minutes: response time targets under 10 seconds are achievable.
• Budget for a small enablement team to maintain the AI and action catalog.
• Report savings and quality together to get stakeholder buy-in.

Upskill your support org

If you're building an action-first support model, consider structured learning paths for support roles and automation. Explore options here: AI courses by job.

Kodee proves that AI that acts - not just answers - changes what support can deliver: faster help, fewer handoffs, and a cost structure that scales without constant hiring.