Every Answer, Right When It’s Needed
Reimagining Student Access Through Conversational AI
ROLE
UX Writer / Content Designer
CONTEXT
DePaul University — HCI 590
PLATFORM
WhatsApp Business + Voiceflow
TYPE
Solo research & prototype project
DEFINING THE PROBLEM
International students couldn't get answers when they needed them most.
DePaul University serves a diverse student body that includes a significant international population — students who commonly rely on WhatsApp for communication and collaboration. These students face a compounded challenge: navigating a new country, a new institution, and a university website whose information architecture makes it genuinely difficult to find answers quickly.
University offices operate on fixed hours. The website buries critical information behind multiple clicks. And when a prospective student in a different time zone has a question about the application deadline at 2am, there's no one to ask. The result is unanswered questions, application drop-off, and students feeling unsupported before they've even enrolled.
HIGH- LEVEL TIMELINE
Researched, designed, and prototyped as part of HCI 590 — with implementation scoped to be deployable before the next application cycle to immediately benefit prospective and current students.
MAIN GOAL
Implement a WhatsApp chatbot that answers the most common university-related questions in real time, 24/7 — using conversational, trust-building language on a platform students already use and trust.
THE BIGGER PICTURE
Hypothesis
If students can access accurate, conversational answers to their most common university questions through WhatsApp — a platform they already use for daily communication — they will experience less friction during the inquiry and application process, feel more supported, and require fewer follow-up contacts with university staff.
Retention
Reduce application drop-off and student disengagement caused by an inability to get timely answers to basic questions about deadlines, campus, and services.
Logic
Demonstrate that meeting students on the platforms they already use — rather than redirecting them to institutional tools — dramatically lowers the barrier to getting support.
Key goal
Provide 24/7 access to streamlined, accurate university information through a chatbot that feels like talking to a helpful person — not navigating a bureaucratic website.
Why WhatsApp specifically: With over 2 billion users worldwide, WhatsApp is the primary communication platform for many international student populations. It's already where they coordinate with classmates and family. Placing the university's support presence there — rather than behind a website login or an email form — meets students exactly where they are, in a context they associate with trusted, personal communication.
WhatsApp's architecture also supports a foundation of trust that other platforms lack — it's end-to-end encrypted, Meta-secured, and interpersonal by design. For students nervous about sharing personal information with a new institution, that context matters.
2B+
WhatsApp users worldwide
2B+
Support availability via chatbot
24/7
Core inquiry types prototyped
~50%
Lower failure rate vs. desktop chatbot
UNDERSTANDING THE MARKET & USERS
Who we're designing for
The chatbot needed to serve a range of student types at different points in the academic journey — from prospective international students exploring DePaul before applying, to current students mid-semester with urgent logistical questions. The content had to be accurate, brief, and warm — never bureaucratic, never dismissive.
International students
Prospective applicants
Current undergrads
Graduate students
CONSTRAINT
University websites are built for comprehensiveness, not speed. A student asking "where is the campus?" shouldn't have to navigate three levels of a site map. The chatbot needed to surface direct, factual answers in two sentences or fewer — no links, no "please visit our website for more information."
FRUSTRATIONS
A key friction point identified in research: students who engage with an institutional chatbot often worry about spam or data sharing afterward. The chatbot's opening message and Meta's verified business badge were both critical trust signals — and had to be reflected in the copy from the first interaction onward.
BRAINSTORMING
Breaking down the process
The design process moved from research to knowledge base architecture to conversation design — with each stage directly informing the next. The chatbot's value depended entirely on the quality and structure of the content behind it, not just the technology running it.
1
Identify the six most common inquiry types
Analyzed university support contact patterns and website navigation data to identify the six questions students most frequently needed answered: campus location, class start dates, available advisors, application process, contact numbers, and faculty information. These became the chatbot's core content architecture.
2
Build the knowledge base from university sources
Uploaded university website content — faculty pages, academic calendar, admissions, student life, about pages — as structured data into Voiceflow's knowledge base. Used direct page URLs alongside PDF documents to ensure the chatbot's answers were grounded in accurate, up-to-date institutional information rather than approximations.
3
Write conversation flows with trust-first copy
Designed every response to prioritize brevity, accuracy, and warmth — following a clear system prompt: always be brief, always summarize, keep responses under two sentences wherever possible. The welcome message established the institutional context ("Nationally Respected, Personally Focused") while immediately signaling helpfulness.
4
Test on WhatsApp vs. desktop and iterate on failures
Ran both platforms through the same six test questions and documented where each failed. The desktop chatbot exhibited a significantly higher rate of failed recognition — particularly for conversational phrasing like "When do classes start?" The WhatsApp version's NLU model handled natural language more reliably, validating the platform choice and informing further prompt tuning.
MINI STYLE GUIDE
Conversation design principles applied
Brief, always
Every chatbot response was written to be fewer than two sentences. On WhatsApp, users read in a messaging context — not a reading context. Long responses break the conversational rhythm and signal that the bot doesn't understand how the medium works.
Conversational, not institutional
University websites sound like universities. The chatbot needed to sound like a knowledgeable, approachable person. Passive constructions, hedging language, and formal phrasing were replaced with direct, first-person-adjacent responses that felt like getting an answer from a well-informed friend.
Trust before information
The first thing the chatbot communicates isn't facts — it's safety. "This business uses a secure service from Meta" is the first message students see. The welcome copy builds on that foundation, establishing helpfulness and institutional warmth before any question is asked.
Always offer a next step
Every response ended with an invitation — "Are there any other questions?" — rather than a period. The goal wasn't to answer one question and close the conversation. It was to make students feel they had a resource they could keep coming back to throughout their entire inquiry journey.
BREAKING DOWN THE PROCESS
The conversation transformation
Below are representative examples of how university information was transformed from web-native copy into chatbot-ready conversation — shifting from formal, link-heavy institutional language to brief, direct, warmly confident responses.
BEFORE - WEBSITE COPY
Campus Locations
DePaul University maintains two campuses in the Chicago metropolitan area. The Lincoln Park Campus serves as the primary residential campus, while the Loop Campus is located in Chicago's central business district. Please visit the Campus Maps page for detailed directions and parking information.
BEFORE - FAILED CHATBOT RESPONSE
User asked: "What advisors are available?"
Bot response: "Sorry, I didn't get that. Please try again."
The desktop chatbot failed to recognize this common, naturally-phrased question — a pattern that repeated across multiple inquiry types and significantly undermined user confidence.
AFTER - CHATBOT RESPONSE
TBD
DePaul has two campuses in Chicago — one in the Loop (business district) and one in Lincoln Park.
Are there any other questions?
AFTER - WHATSAPP CHATBOT RESPONSE
TBD
DePaul offers academic advising through each college, plus dedicated advisors for international students, graduate programs, and financial aid.
You can find your assigned advisor through the student portal, or ask me a more specific question about advising.
Are there any other questions?
Before
After
KEY MOMENTS
Where the content work had the most impact
A chatbot's content design lives and dies in three moments: the opening, when the bot doesn't understand, and when the answer is sensitive. Each required a distinct approach.
The first message — establishing trust
The welcome message sets the entire tone of the interaction. "Nationally Respected, Personally Focused — I'm your helpful AI assistant" immediately anchors the bot in DePaul's institutional identity while positioning it as personal and approachable. This wasn't accidental — it was the single highest-stakes piece of copy in the entire flow.
01
When the bot doesn't understand
The desktop chatbot's failure to recognize common questions — and its generic "Sorry, I didn't get that" fallback — was the most damaging content decision in the original flow. The WhatsApp version's NLU handled natural phrasing significantly better, but the fallback copy itself was also rewritten to acknowledge the gap without making the student feel at fault.
02
The privacy moment — before the first question
Before any student types a single word, WhatsApp displays the Meta secure service notice. Rather than treating this as a liability to work around, the conversation design leaned into it — ensuring the chatbot's opening copy built on that security signal rather than undercutting it with impersonal or over-formal language.
03
CONCLUSION
Deliverables & Outcomes
Access
24/7 availability for international students across all time zones — no office hours required
Friction
Streamlined answers to the six most common university inquiries, reducing the need to navigate the university website
Trust
Conversational, secure, warmly branded experience that met students on the platform they already trusted
SUMMARY
Always on. Always helpful. Always human.
-
Designed and wrote the full conversation architecture for a WhatsApp chatbot serving DePaul University's student population
-
From knowledge base curation and system prompt design through welcome copy, response writing, and failure-state handling.
-
Built and tested a working prototype via Voiceflow, integrated with WhatsApp Business through Meta's verified channel.
-
Comparative testing between the WhatsApp and desktop versions documented a measurably lower failure rate on WhatsApp
-
Validating both the platform choice and the NLU configuration approach.
Future Outlook
The next phase of this work is performance metrics — defining what chatbot success looks like post-implementation and building the measurement framework to track it.
Beyond DePaul
The conversation design model developed here offers a replicable template for any university looking to extend institutional support into the communication channels their international students actually use.
As chatbot technology continues to mature, the content framework, trust-first, brief, conversational, will only become more valuable.
THANK YOU!
NIKITA KING
