The HBCU AI Moment

An institution can make a hundred reasonable decisions about AI and still end up somewhere it never chose to go. Purpose is what lets it know whether those decisions are adding up to the university it actually intends to become.

Two rooms

Over the past several weeks, I have sat in rooms at North Carolina Central University and Winston-Salem State University, listening to faculty, technology leaders, consultants, students, and administrators work through this moment in real time.

At Central, the conversation carried the energy of possibility. OpenAI instructors led most of the breakout sessions, and by afternoon I was building an app myself, part of a competition to turn Codex loose on whatever problem we chose. Around me, people built tools to track student progress, match students with mentors, map food deserts, and find a pickup soccer game. Within hours, they had working websites. I left with a new term in my vocabulary: MVP, the minimum viable product. The least a thing must do in order to function.

At Winston-Salem State, that excitement had come closer to the ground. Professors described what they had actually changed in their classrooms. Faculty developers talked about who shows up for training and who does not. Consultants talked implementation. The question had moved from what AI could do to how to do it well.

Neither institution is improvising. Central has its Institute for Artificial Intelligence and Emerging Research. Winston-Salem State has its Center for Innovative and Transformative Instruction. What I encountered was not a set of institutions waiting for someone else to define the future. It was HBCUs building, experimenting, training faculty, writing policy, and redesigning assignments faster than any committee cycle can comfortably manage.

Some professors were already asking the harder question. They described changing assignments so students would still do the intellectual work that mattered. They did not call it formation, and they did not need to.

If AI can perform more of the task, what does the student still need to learn to do?

Leaders named the aspiration, as you will see. What I did not hear in either room was an institutional frame capable of carrying that question beyond one professor's classroom and into the decisions themselves. Governance was one conversation, ethics another, implementation a third. Purpose sat somewhere behind all three. When those four never meet in the same decision, efficiency tends to settle the argument, because it is usually the easiest thing to measure.

I wrote this to learn from the people doing this work, not to grade it. At the end, I ask for one thing: a single AI decision your campus is wrestling with.

What follows is built the way a lawyer builds a case. First the facts: where HBCUs stand. Then the rules those facts establish. Then what follows from them.

I. HereAI has already arrived

What the rooms showed

At Central, I spoke with the director of information technology for the School of Law. We talked about the arc of legal technology: Lexis and Westlaw on CD-ROM, then the web, and now artificial intelligence. Law is one of the places where the formation question cannot remain abstract. A lawyer who can no longer independently find, read, and weigh authority has not merely gained a tool. She has outsourced a core professional capacity.

Later that day, one of Central's AI leaders interviewed two state senators on stage. They spoke about the breadth of North Carolina's engagement with AI and the state's intent to bring HBCUs into it, while she pressed on the different verticals any serious AI strategy has to account for. Her own framing went well beyond access. The task, she said, is to “prepare people for futures they cannot see,” and to prepare students to lead in a world where AI is ubiquitous. “AI literacy is not enough.” And HBCUs, in her words, “should help build the future.” AI at HBCUs is no longer only a campus matter. It is becoming state policy, with public money, workforce development, and political attention attached.

In another hallway, I met people from organizations whose mission is to prepare the Black community to use AI well as the tools arrive. Their concern is access and readiness. It is a real concern and a necessary one, especially when technological transitions have so often rewarded those who entered early and widened the distance for those who did not.

At Winston-Salem State, CITI helped create a gathering that brought students, faculty, technologists, consultants, and the private sector into the same conversation about actual use. Faculty developers from North Carolina A&T described their training work candidly, and one of them spoke directly to ethics. They had created substantial material, but broad AI training does not automatically draw faculty. Faculty engage when the conversation speaks to their discipline. The historian wants to know what AI means for sources. The business professor wants to know what it means for analysis. The writing professor wants to know what remains writing. The nurse wants to know what judgment must remain clinical. The lawyer wants to know what cannot be delegated without weakening professional competence.

They also marked the pace of change. Three years ago, the conversation was Grammarly. Now it is systems capable of drafting, analyzing, coding, researching, and producing substantial portions of academic work. That did not feel like gradual curricular development. It felt like the ground underneath the curriculum moved.

I also spoke at length with a senior technology leader at a South Carolina HBCU. He has watched several generations of technology move through institutions, and he described campuses where some faculty adopt quickly while others hold more carefully to what has worked. Many senior faculty have lived through several technological revolutions already. They remember the promises, which is one reason their caution should not automatically be treated as obstruction. It may contain institutional memory. But he was clear on one point: the technology is no longer a choice. The only questions left are how it will be used and what will govern it.

The webmaster stage

That conversation connected with several others I had with professionals whose careers began, like mine, in the late 1990s and early 2000s. We remember another transition. There was a time when an organization that wanted to be on the World Wide Web needed a webmaster, someone whose job was to understand the strange new combination of servers, HTML, links, browsers, and infrastructure required to put an institution online. Then the web disappeared into everything. Nobody announces anymore that their organization uses the web. It became infrastructure.

At Winston-Salem, I joked that we may be somewhere around the webmaster stage of AI. The line got a laugh, in part because several people in the room remembered the webmaster. But the analogy matters. Right now, AI is unusually visible. We hold AI summits, appoint AI committees, hire AI specialists, run prompt workshops, and debate whether a particular assignment allows ChatGPT. That visibility will not last.

AI will increasingly be embedded in word processors, legal research, learning-management systems, analytics, clinical systems, search, and technologies that will no longer introduce themselves as artificial intelligence at all. That is precisely why institutions need something more durable than today's list of approved tools. They need a purpose capable of surviving the tools.

What the national data shows

The rooms are consistent with the numbers. In June 2026, the Thurgood Marshall College Fund's Payne Research Center released Ready to Scale, Poised to Lead, a readiness study produced with Tyton Partners. It surveyed leaders at HBCUs, historically Black community and technical colleges, and predominantly Black institutions, so its findings are broader than HBCUs alone.

Seventy-five percent of those leaders reported using AI daily or weekly. Yet only 40 percent were confident AI was delivering meaningful value at their institution today, while 90 percent believed it could support their mission. Leaders named funding, staff capacity, and infrastructure among their barriers. A majority expressed at least some concern about the ethical risks of AI decision-making. Only 21 percent currently measure AI literacy as a student learning outcome; 81 percent expect to within three years.

That gap interests me more than the adoption number. AI is already present, but institutions are still working out what meaningful use should mean. A technology can be widely available, enthusiastically used, and strategically funded before an institution has a common account of what that use is supposed to accomplish.

A 2025 survey by Huston-Tillotson University, Ellucian, and UNCF found similarly widespread use: 96 percent of HBCU faculty and 98 percent of students surveyed had used AI tools, while many reported that their institutions had not formally implemented them. Winston-Salem State measured its own version of that gap. In a Fall 2025 survey of 100 WSSU faculty, 79 percent reported actively using generative AI, while only 17 percent said their department had clear guidelines.

WSSU published those numbers itself and built workshops and faculty learning communities around them. That is not evidence that Winston-Salem State is behind. An institution willing to measure the distance between practice and guidance in public, then build into the gap, is leading.

What the vanguard is building

The leading edge is substantial, and it is not one thing. An HBCU reader should be able to locate their own campus somewhere on this map.

The final column is my classification of each initiative's dominant function. Several of these institutions are doing substantial work across more than one category.

In this sample, the largest announced dollar commitments sit in capacity and research infrastructure, while few initiatives publicly describe their work in the language of formation. That is not a criticism. HBCUs are doing what institutions facing a technological shift should do: building expertise, creating infrastructure, attracting new resources, and making sure Black students are not spectators in the next economy. HBCUs have too often been asked to enter technological revolutions after the infrastructure, capital, and norms were already set elsewhere.

The argument here is not that they should build less. It is that the building has reached the point where another question deserves equal institutional weight: What is all this capacity for?

II. UnderSix working rules

None of what follows is a statute. These are working rules drawn from what I saw and what the data shows, and I want HBCU leaders to test them: where they hold at Jackson State, Prairie View, Howard, or Florida A&M, where they fail, and what they miss. Until someone shows me otherwise, here is how they read.

Rule 1: Adoption does not wait for governance.

Seventy-nine percent of WSSU faculty surveyed use generative AI; 17 percent report department guidelines. Three quarters of surveyed leaders nationally use AI weekly or more. Faculty do not wait for a university-wide framework before deciding what students may do with ChatGPT. Students certainly do not wait. Administrators find workflows to accelerate, and researchers find models that synthesize faster than they can.

By the time a committee settles a policy, hundreds of smaller decisions have already been made in classrooms, advising offices, research projects, and administrative workflows. Governance still matters. But policy often arrives after practice has already begun writing the rule.

Rule 2: Capability is not coherence.

A building, grant, research center, course-redesign initiative, or new degree can dramatically expand what an institution can do. It does not answer what the institution is trying to accomplish. Greater capability makes the question more urgent, because more can be done before anybody has to stop and ask whether it should be.

That is why the distance between 90 percent optimism about AI's promise and 40 percent confidence in its present value matters. It is not a technology gap. It is a purpose gap: institutions gaining the ability to do more before they have a shared language for deciding which uses actually advance their mission.

Rule 3: Training travels through the discipline.

The A&T faculty developers named the mechanism. Faculty come to training that speaks to their field. They are less likely to come to AI training in the abstract. That is not apathy; it is how academic expertise works.

But it creates a second problem. The formation question gets answered one course at a time, by individual professors, with no shared language that lets what one discovers travel to the next. The professors I heard adjusting assignments were working on different versions of the same problem. The institution needs a way to learn from them, rather than requiring each discipline to rediscover the issue alone.

Rule 4: Borrowed tools arrive with borrowed purposes.

MVP is a sound engineering discipline: ship the least that works, then iterate. It is also not how you form a person. A student is not a minimum viable product iterated toward competence, and a graduate is not shipped.

The consultants I spoke with at Winston-Salem State work at the intersection of education and the professional world. What I took from those conversations, though none of them put it this way, is that the two sectors want different things from the same tools. A business must answer to economic performance while preserving enough human proficiency, judgment, and accountability to function safely. A university exists to form the student. Efficiency may matter operationally, but it is not the university's educational purpose.

That difference changes how we read friction. A student taking forty minutes to wrestle with a paragraph may be doing exactly what the course requires. An aspiring lawyer struggling through a difficult argument may be developing legal judgment. A nursing student working through an ambiguous clinical scenario may be building the pattern recognition that will matter when a patient's presentation fails to match the textbook. A seminar that takes ninety minutes to reach an answer a model could produce in nine seconds may still be working perfectly. The outcome is not only the work produced. It is the capacity developed in the person who produced it.

When educational institutions adopt corporate AI playbooks without translating them through educational purpose, they import a definition of success designed for another kind of institution. The technology can be the same while the purpose of using it is not.

Rule 5: Caution is evidence too.

It is tempting to divide a campus into innovators and resisters. Sometimes that is accurate. Often it is too simple. Faculty who have taught through personal computing, the internet, learning-management systems, online education, MOOCs, analytics, and several generations of educational software have earned the right to ask what is actually different this time.

The person slowing adoption may not be standing in the way of innovation. That person may be carrying memory the institution has not yet learned how to use. Some resistance is still resistance, and some is anxiety or exhaustion. But institutions should not train the question out of people merely because it slows implementation. The most important question in any technology transition remains the simplest: What is this thing for?

Rule 6: What gets measured begins to steer the purpose.

Eighty-one percent of institutions in the TMCF study expect to measure AI literacy as a learning outcome within three years. That is good. Students should understand these tools, their strengths and weaknesses, and how to use them responsibly.

One of Central's AI leaders put it plainly on stage: AI literacy is not enough. Literacy measures whether a student can use AI. It does not measure whether the student can still exercise the capacity AI now performs for them. If institutions measure literacy and efficiency while leaving human capacity unmeasured, those outcomes will quietly steer institutional purpose. Metrics rarely stay neutral. Over time, what we repeatedly measure teaches us what to value.

What accumulates

Each individual decision can make sense. A professor redesigns an assignment because AI can generate the old one in seconds. A department introduces AI tools because graduates will meet them at work. An administrator automates a process that consumes staff time. A mentoring app matches students with advisors faster than a person could.

The harder question appears when we ask what accumulates across all of them. Institutions rarely surrender human capacity in one dramatic decision. They lose it incrementally, through hundreds of reasonable ones.

What happens to judgment when recommendation becomes automation? What happens to responsibility when the human being remains formally in charge but increasingly depends on a system to tell them what to do? What happens to learning when production becomes effortless? What happens to relationship when an interaction that once required another person can be completed more efficiently without one?

Those are not questions about whether AI works. They are questions about what institutions are for.

III. ThereforeWhat must remain human?

The question will be answered by someone.

The state senators at Central made one thing plain: public investment in AI at HBCUs is coming with its own priorities attached. Vendors are already in the room. Funders write outcomes into grants, employers bring workforce demands, legislatures bring policy priorities, and students bring expectations formed by consumer technology. None of those parties is necessarily wrong.

But if an institution does not decide what its AI adoption is for, that question will not stay open. Someone else's purpose will fill the blank. In 2024, a coalition that included the Thurgood Marshall College Fund, NAFEO, WCET, the Online Learning Consortium, and Complete College America argued that HBCUs should not merely adapt to artificial intelligence but help shape it. They are right, and the same case was made from the stage at Central. Shaping requires something adaptation does not: an answer to what the institution is unwilling to give up.

I am not the only one asking.

The last session I attended was led by an educator from Atlanta who began as an elementary school teacher. She is openly pro-technology, and she is also deeply concerned with the human capacities that still have to be nurtured as technology use expands. That she reached the formation question by a different road matters. It means this is not one person's framework projected onto a technological moment.

The question surfaces wherever educators take both sides seriously: the power of the technology and the development of the person. The entry point may be pedagogy, ethics, workforce preparation, professional competence, or student development. The problem underneath is increasingly recognizable.

Human first. Then bend the technology toward the institution.

Those questions are what led me to develop ReHuman. Its starting claim is simple: human capability is the driver. Technology should be bent toward the institution's purpose, rather than the institution bent toward the technology.

That does not mean requiring people to perform meaningless work because humans used to perform it. Machines should do what machines do well, and institutions should gladly release work that consumes time without forming, protecting, or expressing anything essential. The concern is different: do not casually surrender the activities through which human beings develop capacities they still need.

In practice, ReHuman asks four questions of any practice AI touches, in order.

1. Personhood. Does the practice encounter the person, or merely a representation of the person? The representation is not the person.
2. Answerability. When something consequential happens, is there a human being with meaningful authority who can explain, justify, reconsider, and answer for the outcome? An answer by itself is not enough.
3. Formation. What capacity was the practice meant to develop in the person doing it? The artifact is not the formation.
4. Presence. Does the activity lead toward human relationship where relationship matters, or does efficiency quietly replace the relationship itself?

These are the same four questions from the end of Part II. Judgment, responsibility, learning, and relationship now have names.

For each question, AI can strengthen the human capacity, bypass its exercise, or begin to replace it. That assessment yields three institutional responses: preserve the practice where the human exercise is essential, redesign it where AI adds value while the important capacity remains, or release it where machines can do the work and nothing essential is lost. This is Formation Before Permission: before asking whether AI may be used, ask what the activity is for.

Consider the mentoring app

Take one of the applications built in the room at Central: a tool to connect students with mentors. It may be an excellent idea. The point is not to keep AI away from mentoring. The point is to understand what mentoring is before deciding which parts of it can safely be optimized.

On Personhood, does the system encounter a student, or only a profile assembled from variables? On Answerability, who can be reached when the match is poor, the assumptions are wrong, or a student who especially needs support falls through the cracks? On Formation, does the system merely make a match, or does the process also teach the student to seek counsel, ask for help, and someday mentor someone else? On Presence, does the app lead toward a relationship, or slowly become a substitute for one?

The right answer may still be to build the app, build it well, and scale it. The difference is that the institution now knows what must survive the scaling. None of this replaces governance, ethics review, implementation, privacy, or AI literacy. It is what lets those conversations meet, because purpose travels into the same decision.

What follows for institutions

If these rules hold, several conclusions follow.

1. Name the purpose before the policy. State what human capacity a consequential practice exists to develop, protect, or make possible before writing the rule that governs AI use.
2. Put the four conversations at one table. Governance, ethics, implementation, and purpose should meet in the same decision, not live permanently in separate committees.
3. Carry formation through the disciplines. Faculty encounter AI through their own fields. Put the formation question there, in the discipline's terms.
4. Measure formation alongside literacy. Ask what evidence shows students can still exercise the capacities their credential claims they possess.
5. Treat caution as evidence. Bring experienced faculty into design conversations rather than merely targeting them for adoption.
6. Build structures that outlive today's tools. Today's AI committees and workshops exist because the technology is still visible. It will not stay visible. Purpose is what has to remain when the label disappears.

HBCUs should learn across campuses

One of the most hopeful things about this moment is the variety. NCCU is building infrastructure. Winston-Salem State and A&T are working deep in faculty practice. Bowie State, Howard, and Jackson State are redesigning teaching. Savannah State is credentialing educators in ethical AI. Spelman is building governance with a working group devoted to justice, and Morehouse has committed to AI that enhances rather than replaces rigor. Shaw has put moral agency into the title of a doctorate. Fayetteville State and Prairie View are moving AI into consequential defense research.

No institution is doing everything, and none needs to. What HBCUs have instead is the possibility of a network that learns from one another's decisions rather than recreating the same argument a hundred times. What did you preserve, redesign, or release? Where did automation create unexpected loss? Where did AI make the education genuinely better? Where was the old assignment never serving its purpose in the first place?

A shared casebook of real AI decisions would be more useful than another universal list of best practices. Best practices travel badly when purpose does not travel with them.

Why HBCUs matter in this conversation

There are two mistakes to avoid in writing about HBCUs and technology. The first is the deficit story: HBCUs are behind and trying to catch up. The second is romance: because of their histories, HBCUs will somehow make better technology decisions than everyone else. Neither does these institutions justice.

HBCUs have silos, resource constraints, uneven adoption, extraordinary professors, exhausted staff, ambitious leaders, young innovators, veteran skeptics, and all the ordinary complexity of higher education. Their history does not guarantee wise AI decisions. It does give them particular reasons to care about the questions AI is forcing.

HBCUs were built in a society that denied Black people equal access to institutions designed to develop talent, confer authority, and open opportunity. Formation was never incidental to the HBCU mission. Institutions that have spent generations forming Black lawyers, nurses, engineers, teachers, ministers, scientists, officers, and civic leaders have reason to care not merely whether their graduates can use the newest tools, but whether those graduates still possess the capabilities their professions and communities will rely on when the tools fail, mislead, or cannot answer the moral question at hand.

Institutions that have done remarkable work with less should also pursue every technology capable of expanding access, multiplying faculty capacity, and opening doors that money and geography once kept closed. My argument is not that HBCUs should slow down. It is that this ambition deserves an equally ambitious account of the human beings all this technology is supposed to serve.

The question underneath

The next chapter of AI in higher education will not be decided by which university gives the most students access to the newest model. Access matters, as do literacy, infrastructure, and governance. But the institutions that matter most will be the ones willing to ask a harder question before efficiency settles it for them.

What are we unwilling to stop teaching human beings to do?

HBCUs have every reason, rooted in their history and mission, to take that question seriously. A future measured only by return on investment was never the point. The point was always people gaining the knowledge, judgment, agency, and relationships to become more fully capable of acting in the world. The HBCU AI opportunity is therefore larger than access, adoption, or even governance. It is agency: the capacity to bend technology toward an institution's human purpose, rather than letting the technology redefine that purpose.