30% Rider Reactivation in 7 Days: How Deliveroo Scaled Operations Across 176,000 Sites With ElevenLabs Voice Agents
Deliveroo's Challenge
Deliveroo operates a food delivery marketplace across the UK and Europe, connecting 176,000+ restaurant and partner sites with a large distributed rider network. Three persistent operational problems eroded network efficiency, each requiring outbound contact with a different population at scale. First, rider applicants were dropping out of the onboarding funnel before activation: incomplete applications went stale as candidates lost momentum without timely re-engagement. Second, restaurant operating hours on the platform were frequently inaccurate — listings showing restaurants as open when they were closed generated failed orders and degraded customer experience, but verifying hours across 176,000 sites was beyond the capacity of any practical human outreach programme. Third, partner tag activation — the process by which restaurant partners unlock promotional and feature eligibility — was proceeding too slowly, leaving value unrealised on both sides of the marketplace. Each problem shared the same structural constraint: the outreach volume required to resolve it at meaningful scale exceeded what a human operations team could execute efficiently and consistently across a distributed, geographically dispersed network.
How ElevenLabs Solved It
Deliveroo deployed three distinct ElevenLabs voice agents, each targeting one of the operational problems. The first agent conducted outbound calls to inactive rider applicants — re-engaging candidates who had stalled mid-onboarding, answering questions, and guiding them toward activation. The second agent called restaurant sites to verify operating hours in real time, confirming accurate listings and flagging discrepancies for correction. The third agent contacted partner sites to guide them through tag activation — the steps required to unlock promotional eligibility. All three agents operated under ElevenLabs Agents, enabling Deliveroo's operations team to deploy them without building custom voice infrastructure. The system gave the operations team real-time visibility into outreach outcomes — confirmation rates, activation completions, and reactivation conversions — across a network of a scale that manual outreach could not have reached consistently.
The Outcome for Deliveroo
The rider reactivation agent reached 81% of inactive applicants and converted 30% of those contacted back into active onboarding within seven days — compressing a re-engagement cycle that would otherwise extend across weeks of sporadic manual follow-up. The restaurant hours agent confirmed accurate operating information for 75% of targeted sites, reducing the volume of failed orders attributable to stale listing data. The partner tag activation agent achieved successful activation at 86% of contacted sites, unlocking promotional eligibility that had been deferred due to the friction of manual guidance. Amy Marangon, Senior Manager of Operations Strategy at Deliveroo, summarised the result: "ElevenLabs Agents transformed our operations — giving us real-time visibility across our network and enabling faster, more consistent engagement with riders and restaurants."
What Deliveroo Learned
Deliveroo's three-agent deployment illustrates a pattern that applies broadly to any platform business with a large distributed partner or supplier network: when outreach volume exceeds human operational capacity, the choice is not between voice AI and human outreach — it is between voice AI and nothing. No operations team can make consistent, timely outbound calls to 176,000 sites to verify hours or reactivate onboarding funnels. The 30% rider reactivation rate within seven days is not a comparison against human performance — it is value that was previously simply not being captured. The architectural insight in Deliveroo's approach is deploying separate agents for separate workflows rather than attempting a single general-purpose agent. Rider reactivation, hours verification, and tag activation are distinct tasks with different conversation flows, different success criteria, and different escalation paths. Specialised agents for each produce higher success rates and cleaner attribution of outcomes. For logistics platforms, gig economy operators, and any business managing a distributed network of independent partners or contractors, the economic case for voice AI in operations is not primarily about cost reduction — it is about coverage. At 176,000 sites, the constraint is reach, not labour cost, and voice AI resolves the reach problem.
- For platform businesses with distributed partner networks, the economic case for voice AI in operations is a coverage argument, not a cost argument — the alternative is not cheaper human outreach, it is no outreach at all at the required scale.
- Separate agents for separate workflows outperform a single general-purpose agent when tasks have distinct conversation flows and success criteria. Rider reactivation and hours verification have nothing in common operationally — conflating them into one agent would degrade both.
- 30% reactivation within seven days quantifies the value of timing in re-engagement. The rider onboarding funnel decays as candidates pursue other options; a voice agent that contacts within days of drop-off reaches a meaningfully different — and more recoverable — population than outreach that arrives weeks later.
See how ElevenLabs can help your business address similar challenges.
Affiliate link — we may earn a commission at no cost to you.