Orange Rolls Out Algolia-Built AI Agents Across 26-Plus Country Markets

Orange Group has deployed AI agents to handle billing and account queries across consumer and business operations in more than 26 countries, using technology from US-based search vendor Algolia.

The agents are built on Algolia's Agent Studio platform, with further expansion planned. Orange will use Algolia's infrastructure to bring search, recommendations and AI agents under a single governed layer. That gives the operator one central point to control, audit and measure relevance as customer interactions move between channels.

For practitioners, the design choice matters more than the rollout count. The technology handles customer conversations using information drawn specifically from Orange's billing records, tariffs, product data and internal policies. It does not rely on the general knowledge of large language models — a deliberate constraint for a regulated customer support environment where wrong answers carry compliance risk.

The agents work across both sales and support enquiries. Customers can input different types of questions without navigating separate systems, which removes a common source of drop-off in operator self-service flows.

What controls does Orange put on each agent?

The deployment includes two distinct control mechanisms:

• Individual guardrails define which subjects each agent can discuss, limiting the scope of any single agent to approved topics.
• Cost controls set limits on how much each agent can spend, capping compute and inference expenditure per interaction.

Together, these give Orange oversight of AI interactions in a channel where operators face regulatory scrutiny of how customer conversations are handled.

What results has the Algolia partnership already delivered?

Orange first adopted Algolia's search technology in 2018. The platform now runs in 24 markets, spanning online shops, support websites and TV applications. The partners report three measurable outcomes from that foundation:

• A 63% increase in order rates through search.
• A seven-point improvement in customer support NPS.
• A 53% reduction in searches returning no results.

These figures come from the vendor-operator partnership and describe the search layer, not the new agents. The agent deployment has no published performance numbers yet. Treat the 63% and NPS gains as evidence of search maturity, not proof of agent outcomes.

What do the executives say?

Nicolas Coudert, digital factory director at Orange Group, said an effective search engine should do more than return pages and "provide a direct and relevant answer". That framing positions the agents as an extension of search — answering directly from Orange data rather than serving links.

Algolia CEO Stephen Lynch said the deployment showed how agents could operate across different customer interactions. "What makes that safe in a regulated industry is governance with every agent grounded in Orange's own data, bounded by explicit rules, capped on cost," Lynch said. "Accountable agents are the ones that get deployed."

The second sentence is effectively Algolia's pitch to every regulated operator: agents without governance controls do not pass procurement. Expect rivals in the search and AI infrastructure market to make the same argument.

What comes next?

Orange will explore extending AI across its wider digital ecosystem. The stated candidates are:

• TV services.
• Personalised recommendation engines.
• Shopping assistants.

Further experiments could allow agents to add products to customers' carts and eventually complete orders. That would move the technology from support automation into transaction execution, a step with clearer revenue implications but also higher risk exposure when an agent acts on a purchase.

No timeline accompanies these plans in the announcement. The 26-country agent rollout is live now; the cart and order-completion capability remains at the experiment stage.