My final project for a Kellogg School of Management Executive Ed course on Generative AI: Executive Strategies to Unlock Enterprise Value.
This is a proposal with ideas for enhancing student success and retention at USF and other universities using AI tools and data integration.
Note - the AI Canvas used is a proprietary Kellogg School of Management model.
2. TASKS TO BE FULFILLED BY AI
Reach Out to Students At
Risk of Attrition
Help Students Set &
Achieve ROI Goals
Log Interaction Notes &
Send Follow-Up
Messages
Respond to Student
Inquiries & Resolve
Problems
Increase
Student
Retention
[AI-ASSISTED CASA/STUDENT RETENTION AT USF]
3. CASA DEPARTMENT AUTOMATION
[AI-ASSISTED CASA / STUDENT RETENTION AT USF]
Meeting
Summaries &
Email Logs
AI-Assisted Messages
(Responses to Inquiries &
Meeting Follow-Ups)
Personalized Playbooks & Content
(for ROI Goal Achievement & Proactive
Outreach)
Student Insights
(Unique ROI Goals & Real-Time Health Score)
USF LLM (Private Instance)
SUPERNORMAL
CHATGPTWRITE
R
STORYD
SALESFORCE
EINSTEIN NEXT
BEST ACTION
AZURE
INVOLV
E.AI SEEK
SYNTHE
SIA
4. QUESTIONS & CONCERNS
[AI-ASSISTED CASA / STUDENT RETENTION AT USF]
• The more cross-functionally obtained data available we are able to incorporate into for
these models and initiatives, the more robust they will be.
• For instance, students’ ROI goals could be captured from their admissions applications,
and a real-time student health/attrition risk dashboard would benefit from shared data on
campus ID swipes, Canvas course engagement, etc.
• Data access will require university-wide buy-in (with top-down leadership support) from the
beginning, trust that data shared will be kept private and secure, and benefits developed
to convince partners they will gain from sharing their data.
Access to University-Wide Data
• USF needs to find an existing or new vendor to power a private, fine-tuned LLM instance
that will keep student data private, secure, and in line with legal regulations like FERPA.
Data Security & Privacy
• CASA team members have a bias towards personal, human-powered service. Education
on the benefits of AI-assisted task management and the power of Human + AI vs. a
Human working alone will be needed to motivate adoption.
Change Management
5. [AI-ASSISTED CASA/STUDENT RETENTION AT USF]
Business & Customer Problem
CASA is the USF team responsible for retaining
student enrollment/tuition revenue. Staff strive to
help students achieve ROI, but need more time
and insights to reduce attrition effectively. Staff are
bogged down handling support concerns, meeting
with students, sending emails, and logging notes.
USF has tons of useful data on students, but it is
scattered in systems that CASA staff can’t access.
Business & Customer Value
Increasing the portion of students retained would
lead to more tuition revenue through increased
enrollment. Retention increases would lead to
improvements in graduation rates, which would
improve USF’s national rankings (lowering
customer acquisition costs). It would also increase
alumni donations (growing USF’s endowment and
lowering its dependence on tuition revenue).
Jobs to be Done
Identify students’ unique ROI goals and facilitate
students’ achievement of those goals. Increase
efficiency of student interactions, problem-solving,
and follow-up tasks to free up more time for
proactive outreach. Identify which students are
most at risk of leaving at any point in time and
employ personalized playbooks to convince them
to stay and help them achieve their ROI goals.
Data & Model Management
Student data from across USF must be
consolidated while maintaining data privacy and
security. A private instance, highly secure USF-
trained (vendor) LLM should be used to power Gen
AI initiatives. Creating prescriptive dashboards and
playbooks will require a human in the loop and
collaboration with CIPE (data scientists who
maintain USF’s unique retention algorithm.
Rapid Solutioning & Prototyping
Leverage IT department to determine vendor(s) to
to hold consolidated data and host/power private
instance LLM. Use 12-week sprint to create and
train MVP of USF LLM. Then, use 12-week sprints
to create MVPs for ROI insights, real-time health
score dashboards, personalized playbooks, AI-
assisted messages, and automated admin tasks
like logging notes and emails into Salesforce CRM.
Prompt Design & Engineering
Develop initial prompts using CASA & USF
staff/faculty correspondence to identify the most
common use cases. Determine secure ways to
record student meetings (in-person and virtual) and
feed them into LLM for summarization. Develop
initial prompts for engaging playbooks and ROI
plans/student touchpoints using broader Higher Ed
and Marketing/Customer Success experts.
Test & Scale
Test USF LLM with staff super-users and smaller
student groups, ensuring accurate insights,
messaging, and notes. As the LLM and use cases
are validated, consider scaling and expanding to
other USF staff and faculty that interact with
students (e.g. faculty advisers) to obtain even more
student data for real-time health scores and more
consistently employ playbooks across campus.
Risk Mitigation
Ensure compliance with FERPA and other student
privacy laws. Test frequently to avoid drift and
keep humans in the loop to ensure accurate and
bias-free messaging. As new users (e.g., faculty
advisers) interact with LLM, dashboards, and
playbooks, make sure access by role ensures data
privacy and security.
Change Management
Have top university leadership share widely why
these projects are a top priority to obtain critical
buy-in. Determine how to make data sharing
worthwhile for cross-functional partners like
Admissions and faculty. Offer workshops for staff
and faculty to explain the benefits and security of
private instance LLM and AI-assisted outreach to
build trust and increase adoption.
DEFINE
“Find it”
DESIGN
“Bottle it”
DEPLOY
“Scale it”