The advertising business of a major social platform
Advertiser support resolution down from 8.9 minutes to 1.4, an 84% improvement.
A social platform’s small-business advertisers needed help with campaign set-up, billing, performance, compliance reviews and ads stuck in review. The earlier chatbot had to be told what to do and could not follow varied phrasing. An Agentforce agent, grounded on the knowledge base and service records through Data 360, resolves most of it in conversation.
At a glance
- Client
- The advertising business of a major social platform
- Engagement
- Replacing a rules-based chatbot with a grounded agent
- Sector
- Technology
- Based in
- California, United States
The stack
01
The situation
What they were working against.
Advertisers frequently needed help with campaign set-up, billing, ad performance, compliance reviews and ads stuck in review. The rules-based chatbot they had before had to be told explicitly what to do, could not handle varied phrasing or a multi-step conversation, and left human reps answering the same questions repeatedly.
02
What was built
The build, in the order it happened.
An agent in the help centre
Advertisers start a chat with the agent directly from the ads help centre rather than filing a ticket.
Grounded on knowledge and cases
The agent searches the knowledge base in Salesforce and the service records in Service Cloud, harmonised by Data 360, to diagnose why an ad was not approved, a pixel is misconfigured or a login fails.
Step-by-step, in plain language
Guidance delivered conversationally, with the multistep cases the old bot could not follow now handled end to end.
03
The results
From a scripted bot to an agent that understands the question.
Average resolution time fell from 8.9 minutes to 1.4, an 84% improvement. Advertiser satisfaction scores rose 20%, case deflection increased, and live reps save 760 hours a year.
- Average resolution time, down from 8.9 minutes
- 1.4 min
- Higher advertiser satisfaction scores
- 20%
- A year given back to live reps
- 760 h
Client name withheld. Figures are as reported for this implementation; where one was a stated target rather than a measurement, the label says so.
04
Our read
How we would build this at your scale.
What we take from this build, and what the same outcome looks like for a team of fifty to five hundred people rather than a global enterprise.
Replacing a rules bot is the cleanest agent business case
You already have the baseline, the volume and the failure log. For a SaaS company we start by mining the old bot’s transcripts for the topics that failed, and scope the agent on those.
Agentforce & AI 02Harmonise knowledge and cases first
The diagnosis works because articles and case history are searchable together. Data 360, or a disciplined knowledge and case model in Service Cloud, is what makes that true.
Data 360 Services 03Resolution time is a weekly number
An 84% improvement is held, not achieved. Topic coverage, escalation rates and satisfaction belong in a monthly review under a managed retainer.
Managed ServicesMore case studies
A national flag-carrier airline
Refund turnaround cut from about 14 days to about four hours.
ReadA North American paperboard and folding-carton manufacturer
A customer portal where packaging buyers track orders, shipments and build requests themselves.
ReadA regulated consumer-products manufacturer selling wholesale to retailers
Salesforce and QuickBooks Desktop kept in step every hour, with nothing re-keyed between them.
ReadWant the same result, without the enterprise bill?
Thirty minutes with an architect. Bring the part of this story that looks like your problem and we will tell you what it takes to get there: which products, how many weeks, and what we would leave out.