TL;DR
Dreamforce showed what a higher ed AI campus can look like, but most universities are bogged down by a disconnected “city of systems” and cautious committees. The path forward is starting small. By setting narrow trial guardrails and building data integrations workflow by workflow, campuses can prove ROI quickly and build a scalable foundation for AI.
At Dreamforce, Salesforce’s AI pitch for higher ed was ambitious: autonomous AI systems that can handle complex administrative workflows, answer student enrollment questions, initiate donor touchpoints, and track academic progress.
To many, the vision seems unrealistic for campuses across the country.
That’s because universities are still trying to wrangle a “city of systems,” the complex, fragmented network of software and hardware used to manage daily campus operations. Core student records live in legacy SIS platforms like Banner or PeopleSoft, financial records live in separate databases, donor histories live in advancement systems, and daily communications happen inside CRMs. On top of that, running a campus’s healthcare centers, athletic facilities, radio stations, and parking systems requires connected technology to support every member of the community, not just learners.
AI automation relies on the quality, accuracy, and organization of the data powering it. Agents can’t manage complex, multi-step processes when information remains trapped in isolated systems.
So while the vision is compelling and the technology is ready, bringing AI onto your campus requires closing a dual readiness gap: making approval decisions fast enough to evaluate AI, and building the clean data connections it needs to act. Both gaps close one workflow at a time, and campuses can start now.
Matching governance to the speed of AI
Dreamforce showed AI agents in ideal conditions, but campus conditions are messier. One university that came to us already had two AI agents running in a pilot, and the project still stalled in committee.
That’s the main bottleneck in higher ed. AI requires rapid development, testing, and iteration, while campus committees are built to look at the big picture. Their job is to protect the university’s mission, values, and stability over decades, taking a deep and careful approach so policies can last for years.
Committee members naturally feel they need to figure out every future risk and ethical question right away. But when a team tries to write a permanent, campus-wide policy in the middle of a short software trial, the small pilot gets weighed down trying to solve giant policy questions it was never meant to handle, and the project stops moving.
The fix is to narrow what the committee has to decide. Ask it to set the guardrails for a 90-day trial in one high-friction workflow, like application processing during peak enrollment. The review then centers on questions like these:
- Access: Which records can the agent read? For example, application records and document checklists, with financial aid data off-limits.
- Actions: What can it do? Flag incomplete applications, draft reminders about missing documents, and route complete files to counselors.
- Oversight: Where must staff sign off? On every outbound message for the first 30 days, and on every admission decision.
- Success: What results would justify expanding? Faster time to a complete file, a smaller peak backlog, and counselor hours saved.
Each question covers one office’s work over a set timeframe so that the committee can answer it quickly. The trial then shows whether the agent helps the admissions team and stays within its limits, and leadership gets an approval process it can reuse for other departments.
Tackling the “city of systems” workflow by workflow
You don’t fix a “city of systems” by demolishing every building at once. That’s the flaw in multi-year data warehousing projects and total ERP overhauls: they try to solve all data fragmentation at once, and campus sees nothing until the end. In higher ed, we call that “phase nothing,” months of discovery that produce nothing campus can use.
Preparing for AI means giving agents secure connections to the specific records each task requires, while the SIS and its records stay in place. MuleSoft handles those connections between Salesforce and legacy systems, and if a campus already keeps data in a warehouse like Snowflake, Data 360 can use it there without making a copy.
IT teams can build targeted, reusable connections one workflow at a time:
- Transfer Credit Reviews: Connect SIS course histories directly to the evaluation workspace in Salesforce.
- Donor Outreach: Connect donor histories in the advancement system to gift officers’ records.
- Advising Workflows: Bring degree audit progress into a single advisor dashboard.
By the fifth workflow, many of the connections a new pilot needs already exist, so each one has less plumbing to build. Together, those connections become the campus-wide data foundation AI needs without a multi-year overhaul.
The path forward: Start small, then speed up
Most campuses are setting their 2027 budgets right now. Flashy AI software licenses can be tempting in those discussions, but buying massive enterprise platforms before your campus is ready is a trap.
Instead of pouring money into big platform promises, here’s how to shape your 2027 plan:
Budget for data governance and integration groundwork first
Agent licenses get the headlines, but they fail without clean data behind them. 2027 budgets must prioritize the unsexy baseline work: cleaning student data and integrating legacy SIS and CRM systems so AI can actually access accurate information.
Hold off on institution-wide rollouts
Don’t try to flip a switch on campus-wide AI or get trapped waiting for committees to solve enterprise policy. Trying to go big right away leads straight to administrative gridlock.
Launch one tightly scoped pilot to build internal muscle
Go slow to go fast. Pick a single, high-friction workflow — like one specific task in admissions or advancement — and test it. A small pilot tests your governance and change management in the real world, giving you the playbook and evidence you need before scaling.
What’s next
This moment reminds me of the early 2000s, when online education got off to a slow start in higher ed. I was part of that shift, and plenty of people told us it couldn’t work at traditional universities. It did, and we beat our targets.
AI puts higher ed in a similar spot. Each campus will move at its own pace, and the ones that get ahead will put working prototypes in front of staff and shorten the time it takes to approve them.
Online education started with a handful of courses. The campus Salesforce showed at Dreamforce can start with one pilot.
For the full set of Dreamforce 2026 announcements, read our recap. If you’re deciding which workflow to pilot first, Coastal’s AI Pathfinder is a no-cost session that reviews your Salesforce org and the use cases you’re weighing, then gives you a prioritized plan and a short readout to bring to your leadership or evaluation committee.
