Asiacell's deployment of Druid AI's Laila agent represents a scaled, operational validation of what agentic automation can achieve in high-complexity environments. The telecommunications provider has automated 79% of digital customer interactions end-to-end whilst simultaneously shifting channel adoption from 12% to 54% and reducing monthly voice calls from one million to 500,000. This is not a pilot or proof of concept—it is a live system serving 20 million customers across multiple channels in one of the world's most demanding operational contexts. The transformation rested on three foundational decisions: a human-centred design process where leadership worked as call centre agents before building features, six months of granular customer inquiry analysis to identify highest-impact use cases, and progressive API integrations that evolved the agent from FAQ tool to transactional executor handling balance checks, service activations, and account management.
The implications for CX teams are material. First, the 79% figure establishes a new baseline expectation for what agentic systems should deliver at scale, raising questions about whether teams currently running older chatbot architectures or even first-generation AI agents are leaving significant automation on the table. Second, Asiacell's zero-press IVR strategy—redirecting customers from voice to WhatsApp without friction—demonstrates that channel migration is not a customer adoption problem but a design problem; teams treating digital adoption as a separate initiative from automation are likely misaligned. Third, the emphasis on continuous agile development to sustain 80% automation in a complex environment signals that deployment is not the end state. For Zendesk administrators and support leads, this raises a critical tension: how do you balance the operational stability required by large-scale deployments with the iterative refinement that agentic systems demand?
The webinar's focus on "what failed first" and "what the AI still hands off to humans and why" suggests Asiacell encountered real constraints in automation scope. This matters because it positions agentic systems not as replacement technology but as orchestration technology—one that requires explicit decisions about interaction boundaries and escalation logic. For CX consultants advising on agent selection, the question becomes whether your platform vendor has the operational maturity and transparency to support this kind of sustained, scaled deployment, or whether you are being sold a capability that works in controlled environments but fractures under real-world complexity.
How Asiacell automated 79% of customer interactions with Druid AI Agents | Free webinar — RTIH Retail Technology Innovation Hub