Customer support · Production

Forethought AI enables 72% self-serve rate and 96% sentiment accuracy for Kickfin customer support

The problem

Kickfin needed to provide 24/7 customer support for restaurant and hospitality workers accessing the platform at 2–4 AM, but relied on overnight human staffing that was hard to fill and impossible to cover on short notice. Training new reps was also a significant burden on support leadership.

First attempt

Before Forethought, Kickfin had no automation tools for customer self-service, forcing the team to staff overnight human shifts that were chronically difficult to fill and cover.

Workflow diagram · grounded in source
1
Customer submits inquiry via widget
trigger
“Kickfin's widget is both on its public website and password-protected customer support portal. Depending on where a customer interacts with the widget, it will serve up the appropriate level of information from Kickfin's knowledge base.”
2
Solve AI detects intent and responds
ai_action
“Forethought's AI Agent trains on Large Language Models, as well as data from Kickfin's own support team to provide more accurate responses to its customers' inquiries. The chat widget automatically picks up on customer intent and respond…”
3
Triage detects inquiry sentiment
ai_action
“Triage determines the sentiment of the support inquiries received by Kickfin, empowering agents with personalized classifiers to understand these inquiries at scale. When reviewing case history, Kickfin's support leadership team easily u…”
4
Assist surfaces knowledge to agents
ai_action
“agents are empowered to search any knowledge base or website right within Salesforce (Kickfin's help desk), and get relevant, AI-powered answers. Agents resolve cases faster and more accurately with relevant macros, past tickets, and kno…”
5
Analytics adjust workflow performance
feedback_loop
“Analytics within Solve have allowed the support team to track the effectiveness of individual workflows and easily adjust them based on performance.”
Reported outcome

Kickfin achieved a 72% self-serve rate and over 2,000 chat deflections with Forethought Solve, eliminating the need for overnight staffing.
Triage predicts customer sentiment with 96% accuracy.

Reported metrics
Self-serve rate72%
Chat deflections+2,000
Sentiment prediction accuracy96%
Agent time savingsincredibly time-saving
Reported stack
ForethoughtSolveAssistTriageLarge Language ModelsSalesforce
Source
https://forethought.ai/case-studies/kickfin
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Kickfin achieved a 72% self-serve rate and over 2,000 chat deflections with Forethought Solve, eliminating the need for overnight staffing.

What tools did this team use?

Forethought, Solve, Assist, Triage, Large Language Models, Salesforce.

What results were reported?

Self-serve rate: 72%; Chat deflections: +2,000; Sentiment prediction accuracy: 96%; Agent time savings: incredibly time-saving (source-reported, not independently verified).

What failed first in this deployment?

Before Forethought, Kickfin had no automation tools for customer self-service, forcing the team to staff overnight human shifts that were chronically difficult to fill and cover.

How is this customer support AI workflow structured?

Customer submits inquiry via widget → Solve AI detects intent and responds → Triage detects inquiry sentiment → Assist surfaces knowledge to agents → Analytics adjust workflow performance.

Related customer support cases