Add to cart rose nearly tenfold among engaged shoppers, measured against a control group.
Orange Conversational Assistant
One assistant replaced search and support across three business units, raising click-through 14%.
A full-page assistant replaces the support and search page across ecommerce, care, and media, live in under 2 weeks.
What the agent does
Answers billing, contract, device, and store questions in one conversation, choosing which knowledge source to draw on.
01
The problem
Why this exists
A telecom support page is a superset of several unrelated jobs. A customer arrives wanting an FAQ answer, a contract comparison, a device, a shop address, or a troubleshooting article, and a single search box has to serve all of them from content that lives in different systems.
Read more
The result is a page nobody uses well. And the internal version of the same problem is worse, because a call center agent needs contract, pass, and store data from internal systems that a search index does not hold.
02
Outcomes
What changes when this is in place
- Orange reports a 14% increase in click-through rate, which it attributes to putting the right content in front of the customer at the right time.
- The assistant covers FAQ and assistance queries, contract and pass offers, product and device discovery, store locator, and support content in one place.
- It was deployed across three business units: ecommerce, customer support, and media.
Read more
- It went live in under 2 weeks from feature activation on orange.ci, which is the deployment figure Orange emphasizes.
- Better content targeting reduces customer service inquiries as well as improving conversion.
03
How it works
Architecture
Orange built the agent on Algolia Agent Studio with advisory support from Algolia rather than delivery. It uses a multi-agent architecture: a central orchestrator routes queries to specialized sub-agents for FAQs, product discovery, contract offers, store locator, and support content. The agent answers billing questions itself and hands more complex requests, such as invoice modifications, to a specialized agent.
Read more
Three things distinguish the build. First, Orange plugged in its own Google Gemini 2.5 Flash API key. Algolia provides no model infrastructure, so Orange controls its own model spend, vendor relationship, and data residency. Second, native MCP server integration lets the agent reach Orange's internal systems for contracts, passes, and store data, which is what makes the internal call center use case possible without custom glue code. Third, cost efficiency was an explicit engineering constraint, with model routing tuned to keep per-conversation cost low at scale.
The assistant replaces the traditional support and search page entirely when triggered, rather than sitting beside it.
Open standards used
Model Context Protocol (MCP)REST APIs
Tech stack
Algolia Agent StudioGoogle Gemini 2.5 Flash
Resources
Next step
Request an intro
Algolia is behind this solution. Ask about consolidating search and support, and the MACH Alliance will put you in touch with the person who can answer.
Not ready to talk to anyone? MACH Insider is the Alliance mailing list. New solutions, program openings, and the research behind them, once a month.
Everyone credited on this solution
- AlgoliaAgent Studio
Every organization named here has a page listing the rest of its work on the Hub.
Related solutions
Call to action clicks rose 274% and lead conversions 190%, with more configurations completed.
Lexus Europe AI Concierge
A shopping assistant that lifted conversion 25%, taken from a split test to all traffic.
Ask Frasers Shopping Assistant
Verification
Published case study
The detail on this page is drawn from a case study published by the contributor or the MACH Alliance. Any figures on this page are the ones that case study states.
Where this came from
Solution detail is published from the contributor's 2026 Agent Ready submission to the MACH Alliance. Submitter names and contact details are not published.
Status: Live in productionAdded Jul 8, 2026A listing is an editorial decision, not a MACH Alliance certification.
Built something worth showing?
Put your agentic solution in front of the enterprise leaders who use the hub to decide what to build next. Submissions are open to MACH Alliance Members, Collaborators, or Ambassadors.