In 30 days, AI-accelerated delivery with Revecast Orchestrate produced more than $1 million in delivery value for a university’s online learning center, and gave its staff an AI assistant for case management, while the project finished a month early and under budget.
Mission:
To better support its students and the staff who serve them, the center set out to move off its legacy Education Data Architecture onto Education Cloud, bring case management and email onto one platform, give staff sourced answers to the questions they field every day, and build a data foundation for future phases, all on a fixed budget and timeline.
Salesforce Products:
- Education Cloud
- Agentforce
- Data360
- Salesforce Knowledge
- Revecast Orchestrate
CHALLENGES:
A Large, Multi-Part Scope
Moving off a legacy Education Data Architecture (EDA) to Education Cloud, while also standing up case management, email-to-case, security and single sign-on, reporting, and a data foundation for later phases.
Staff Buried in Repetitive Questions
Staff answer the same questions about courses, programs, and scholarships every day, pulling answers from knowledge articles, the website, and records spread across the org.
AI That Has to Be Trusted
An assistant that touches student and constituent data can’t act on its own. Every change needs review, every answer needs a source, and access has to follow each user’s existing permissions.
A Fixed Budget and Timeline
The core migration had to land on schedule and on budget, leaving little room to add AI capabilities without pushing everything else back.
SOLUTION:
Shared Context from Day One
The SOW, discovery notes, and backlog were distilled into one shared project context. Every working session started by reading it and ended by updating it, with decisions and lessons logged for the next session.
Built in Short, Validated Loops
Using Revecast Orchestrate, the team built, tested, and published one change at a time. The agent went from first design to first working version in 7 days, and through 88 published versions in a full build of about 3 weeks.
An Agentforce Staff Assistant
A Gen-2 Agentforce employee agent with five specialist subagents works across cases, contacts, and opportunities, answers from Knowledge and the website with sources, and drafts scholarship email replies for staff review. More on how it works below.
Nothing Shipped Unvalidated
Every change is previewed and approved before it saves, and the agent runs under each user’s own permissions. The final run passed 429 of 429 Apex tests with 90%+ coverage per class, and 114 prompt-injection tests changed no data.
THE AGENTFORCE STAFF ASSISTANT:
Answers with Sources
Staff ask questions in plain English. Answers come from published Knowledge articles first, then approved university web pages, and every answer links to its source. If nothing is found, the assistant says so instead of guessing.
Finds and Summarizes Records
It looks up constituents, cases, and opportunities, and asks which record is meant when a name matches more than one. It can also roll up a person’s cases and recent activity, or show how an opportunity connects to its account and campaign.
Changes Only After a “Yes”
Before it creates or updates a case, contact, or opportunity, logs a call, or sends an email reply drafted from a Knowledge article, it shows the exact change and waits for a plain yes. If staff don’t confirm, nothing is saved. It checks for duplicate cases first and runs under each staff member’s own permissions.
Monitored from Day One
Every action is logged without storing chat text, every finished conversation is scored for quality, and any problem becomes an owned finding with a due date on an admin dashboard. The team can see whether the assistant is being used, working, and answering well.
RESULTS:
$1M+ in delivery value in 30 days, with 5,817 hours of work saved
Delivered beyond the original plan: $1M+ in extra scope at no added timeline
Project finishing a month early and about 15% under budget
