Customer support · Production

Blackhawk Network automatically resolves half of customer inquiries with Ada generative AI

The problem

Tango Card's scripted Ada chatbot required every new automation use case to be manually built as an individual answer path, creating growing maintenance overhead. Knowledge management was reactive and ad hoc, and AI performance insights stayed siloed within the support team.

First attempt

The scripted chatbot model's flexibility came at the cost of substantial manual upkeep — every new capability required building and maintaining individual answer paths, accumulating hours of overhead.

Workflow diagram · grounded in source
1
Customer inquiry arrives
trigger
“all incoming customer inquiries across all channels”
2
Ada AI autonomous resolution
ai_action
“Automatically resolving half of all incoming customer inquiries across all channels”
3
User lookup and order retrieval
ai_action
“User Lookups: Greet logged-in customers by name and streamline handoffs by identifying who they are upfront. Order Info Retrieval: Provide product-specific responses to recipients—without extra back-and-forth or guesswork.”
4
Smart case routing
routing
“Smart Routing: Automatically route cases based on use”
5
Agent review of AI conversations
human_review
“agents now review and critique AI conversations—a skill once reserved for QA roles”
6
AI Certified Agents train the model
feedback_loop
“These agents will continuously test use cases in self-service channels, helping train the AI model while reinforcing skill-building within the team”
7
Knowledge base optimization
feedback_loop
“Content is now created with the AI agent in mind. That means carefully crafted, tested, and monitored articles that not only answer questions, but do so in a way that's optimized for generative AI.”
Reported outcome

Blackhawk Network now automatically resolves half of all incoming customer inquiries across all channels, has upskilled frontline agents into AI collaborators, and expanded automation across all brands and channels.

Reported metrics
containment rate (scripted Ada era)over 70%
Agent cognitive loaddramatically reduced cognitive load
Repetitive admin workcutting down dramatically on repetitive admin work
Scripted-era maintenance overheadhours of upkeep, repetition, and overhead
Reported stack
AdaUdemy
Source
https://www.ada.cx/case-study/blackhawk-network
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Blackhawk Network now automatically resolves half of all incoming customer inquiries across all channels, has upskilled frontline agents into AI collaborators, and expanded automation across all brands and channels.

What tools did this team use?

Ada, Udemy.

What results were reported?

containment rate (scripted Ada era): over 70%; Agent cognitive load: dramatically reduced cognitive load; Repetitive admin work: cutting down dramatically on repetitive admin work; Scripted-era maintenance overhead: hours of upkeep, repetition, and overhead (source-reported, not independently verified).

What failed first in this deployment?

The scripted chatbot model's flexibility came at the cost of substantial manual upkeep — every new capability required building and maintaining individual answer paths, accumulating hours of overhead.

How is this customer support AI workflow structured?

Customer inquiry arrives → Ada AI autonomous resolution → User lookup and order retrieval → Smart case routing → Agent review of AI conversations → AI Certified Agents train the model → Knowledge base optimization.

Related customer support cases