It was 11 p.m. on a Sunday when Jasreen Bawa opened her laptop and typed her first message into TerrierGPT. A pitch deck was due in 48 hours for a mock airline’s crisis communications strategy, and she was stuck. In seconds, the tool pulled up case studies from BU’s library, suggested some recent academic articles on crisis PR, and sketched out a structural outline that actually matched her course rubric.
“It didn’t just give me answers,” said Bawa, a senior majoring in public relations at the College of Communication. “It understood what I was trying to do for class. It felt like a research partner who also happened to go to BU.”
That kind of reaction is becoming more common across campus. Since BU launched TerrierGPT last fall, the tool has become a part of how many students work. Additionally, it is a test case for whether a university can actually get this AI integration thing right, rather than just slapping a policy PDF on a website and calling it a day.
What Is TerrierGPT?
TerrierGPT is BU’s custom multi-LLM AI assistant, built in partnership with a third-party provider and integrated into the university’s own internal systems. Rather than replacing tools like ChatGPT or Google Gemini, it expands on them, drawing not only on those models but also on BU’s library databases, course management platform, and vetted academic resources, rather than the open internet in all its chaotic glory. It knows what a COM syllabus looks like. It can find a peer-reviewed source and actually tell you where it came from.
Any enrolled student, faculty member, or staff can access it through the MyBU portal with their Kerberos login. No download, no subscription, no extra steps.
The New AI Policies
The tool launched alongside a revised set of AI guidelines that took nearly a year of faculty debate to produce. The bottom line is that AI is fine as a research and editing aid, but you still have to write your own work, and you have to say when you used it. (Naturally, you should follow the AI policy from each of your instructors.)
Syllabuses now include a dedicated AI section. Citation formats for AI-assisted work have been standardized. There’s a clear list of what crosses the line: submitting AI-generated prose as your own, using outside tools when TerrierGPT is specified, and completing assessments without AI disclosure.
Compared to some peer schools, BU is ahead. Northeastern and Boston College have broadly similar policies, but few universities have built and deployed their own tool rather than just issuing guidelines about the public ones. Harvard has largely left it to its individual schools to sort out.
What Do the Students Think?
For Bawa, TerrierGPT has become a regular part of her workflow, and she’s pragmatic about why. “Every agency I’ve interviewed with this year has asked me what I know about AI in communications,” she said. “Being able to say I’ve actually used a real tool has made a difference.”
Not everyone’s sold, though. Yuvna Kumar, a senior biology major at CAS, has poked around with it and keeps running into the same problem.
“I end up fact-checking everything anyway,” Kumar said. “In my field, a misrepresented study result matters. So sometimes it just feels like extra steps.” She’s open to using it for writing help and formatting, but for actual science research, she’d rather go straight to the source.
That friction is something faculty are navigating, too. Some COM professors have redesigned assignments around TerrierGPT, asking students to document their AI-assisted process as part of the grade. Others have opted out entirely, making the case that good writing requires a certain amount of struggle to develop and shortcuts short-circuit that.
Not every student’s hesitation is about accuracy. For some, the issue is simpler: they don’t trust it. A recurring concern among students who avoid TerrierGPT is the feeling that using a university-built tool means the university can see everything they’re doing. “It’s built by BU, run by BU. I just assume someone can read my conversations,” said one anonymous CAS junior. BU says that’s not how it works. According to the Digital Learning and Innovation office, conversations are not monitored by faculty or administrators, data is not shared with third parties, and prompts are not tied to individual academic records.
The Bigger Picture
None of this is happening in isolation. Across the industries that COM graduates enter (PR, journalism, advertising, and film), AI is already transforming jobs. Agencies use it for media monitoring and drafting. News organizations use it for transcription and data work. Ad shops use generative tools to iterate creative work at speeds that weren’t possible two years ago.
The question BU seems to be asking is whether graduates knowing how to use these tools thoughtfully will be better off than those who avoid AI. The university is betting the answer is yes. Harvard’s decentralized approach leaves room for inconsistency. BU is trying to make a consistent bet across the whole institution.
What’s Next
Updates to TerrierGPT are planned each semester, including more discipline-specific data integrations and multilingual support. There’s a feedback portal at bu.edu/terriergpt where students and faculty can flag problems or suggestions. The Digital Learning and Innovation office is running free workshops through March on prompt writing and discipline-specific use cases, with a parallel series for faculty.
The Bottom Line
On any given afternoon in the GSU or the COM lounge, TerrierGPT is open in a tab somewhere between a half-finished paper and a Spotify playlist. For most students, it’s functioning in the background, in the same way that Grammarly became a subconscious habit. You reflexively use it without thinking about it.
Bawa, for her part, thinks that’s fine. “I still wrote my pitch deck,” she said. “TerrierGPT helped me start. The argument, the strategy – that was mine.”
That’s what BU is going for. AI should be aiding our lives, not becoming it.
Leave a Reply