Direct-to-consumer meal delivery, health and wellnessLive site
Outbound Chef Voice Agents
A family of outbound AI voice agents, each giving voice to one of the brand's chef personas, that call customers at the lifecycle moments that decide whether they stay.
Before
Email and SMSLifecycle moments handled mechanically, flattening a brand built on chef personalities into a utility
After
Chef callsThe right chef on the phone at the right moment, triggered automatically by order, delivery and cancellation events
Overview
The right chef, the right customer, the right moment.
A US direct-to-consumer meal delivery brand had built its identity around a cast of distinct chef personalities, each with their own backstory, food opinions and signature phrases, and that cast was the reason customers stayed. The moments that mattered most for retention were exactly the moments a human chef could never be on the phone.
NFI built a family of outbound AI voice agents, each voicing a distinct chef from the client's roster, triggered by lifecycle events in the client's commerce and CRM systems and running entirely on the brand's own accounts.
The problem
Retention moments a human could never staff.
The first order after checkout, a completed delivery, a pending cancellation: email and SMS did the job mechanically, and a human calling team at that cadence was not economically viable.
01
Flattened brand
Email and SMS handled the moments but turned a personality-led brand into a utility.
02
Uneconomic staffing
A human outbound team at the cadence lifecycle events demand did not add up.
03
Compliance and logging
Every call had to respect US state-level calling windows and land back in the CRM with its outcome.
The solution
Persona, trigger, compliance, and a full record of every call.
Each agent voices a distinct chef with their personality, tone and signature phrases: a warm post-order welcome from one chef, a delivery confirmation from another, a retention conversation from a third.
Calls are triggered automatically by events in the client's commerce and CRM systems (a paid order, a completed delivery, a subscription cancellation tag) and routed to the right persona for the moment.
A compliance layer enforces US state-level TCPA calling windows and reschedules out-of-hours calls rather than skipping them. Every call lands back in the CRM with its transcript, outcome tags and follow-up actions, an error-monitoring workflow flags misfires and dropped calls, and telephony, CRM and model access all stay on the client's own accounts.
Outbound AI voice agents with branded personasLifecycle triggers from commerce and CRM eventsTCPA calling-window compliance layerCRM call logging with transcripts and outcome tagsError-monitoring workflowClient-owned telephony, CRM and model accounts
Every lifecycle moment covered, every call on the record.
3Lifecycle triggers live: order placed, delivery completed, cancellation requested, each with its own chef
TCPAUS state-level calling windows enforced automatically, with out-of-hours calls rescheduled rather than skipped
100%Of calls logged end to end in the CRM with transcript, outcome tag and customer context
OwnedTelephony, CRM and model access all on the brand's own accounts, with error monitoring on every trigger
Questions
About this project.
Calls are triggered automatically by lifecycle events in the client's commerce and CRM systems: a first paid order, a completed delivery and a pending cancellation, each routed to the chef persona that fits the moment.
A compliance layer enforces US state-level TCPA calling windows and reschedules out-of-hours calls rather than skipping them.
Every call lands in the CRM with a transcript, outcome tags and follow-up actions, an error-monitoring workflow flags misfires and dropped calls, and the telephony, CRM and model accounts all stay on the client's own tenancy.
