Generative AI in Higher Education Constituent Relationship Management (CRM): Opportunities, Challenges, and Implementation Strategies

Abstract

This research explores opportunities for generative artificial intelligence (GenAI) in higher education constituent (customer) relationship management (CRM) to address the industry’s need for digital transformation driven by demographic shifts, economic challenges, and technological advancements. Using a qualitative research approach grounded in the principles of grounded theory, we conducted semi-structured interviews and an open-ended qualitative data collection instrument with technology vendors, implementation consultants, and HEI professionals that are actively exploring GenAI applications. Our findings highlight six primary types of GenAI—textual analysis and synthesis, data summarization, next-best action recommendations, speech synthesis and translation, code development, and image and video creation—each with applications across student recruitment, advising, alumni engagement, and administrative processes. We propose an evaluative framework with eight readiness criteria to assess institutional preparedness for GenAI adoption. While GenAI offers potential benefits, such as increased efficiency, reduced costs, and improved student engagement, its success depends on data readiness, ethical safeguards, and institutional leadership. By integrating GenAI as a co-intelligence alongside human expertise, HEIs can enhance CRM ecosystems and better support their constituents.

1. Introduction

Higher education in the United States faces a decline of traditional college-aged students, rising costs alongside decreasing tuition and endowments, and societal attitudes questioning the value of higher education [1,2,3,4]. The higher education industry has also experienced significant technological turbulence since the COVID-19 pandemic when it transitioned to entirely virtual operations. These challenges have taken their toll on university employees. Gallup indicates a 35% burnout level for university employees, second only to K-12 educators [5]. Given this backdrop, higher education institutions (HEIs) seek ways to optimize their operations so they can continue to attract and retain learners, alumni, and employees.

Customer (or more broadly, constituent) relationship management (CRM) processes and technologies are used to unify contact profiles and data in order to conduct and track workflows and communications. The industry term “Customer” in CRM is renamed and expanded to incorporate many “Constituents” in the higher education setting. CRM systems are used to orchestrate student recruiting processes, track admissions decisions and statuses, manage student advising and services, conduct university fundraising and alumni outreach, and many more engagement and communication processes. In this paper, we use the terms customer and constituent interchangeably.

Many higher education leaders view generative artificial intelligence (GenAI) as a way to increase the power of their CRM technology to address the sector’s urgent challenges. In private industry, research has already been undertaken on GenAI’s potential to aid in customer acquisition, retention, churn avoidance, and upgrading [6,7,8,9,10,11]. In higher education, these activities translate to recruiting and admissions, student success, and advancement (alumni relations and fundraising). While previous AIs have had the power to conduct big data analyses and predictive analytics, researchers point to the power of unlocking unstructured “dark data” as one of the greatest benefits of today’s GenAI capabilities [9,12]. Whereas past CRMs with AI enabled could track college application data, class lists, attendance, and grades and make sophisticated analyses and predictions, today’s GenAI can combine those data with image recognition of uploaded transcript images, can summarize advising notes, and can conduct sentiment analysis of discussion posts, all leading to automated care plans, approval processes, alerts, and communications. This data harmonization and activation is key to GenAI’s potential for CRM to strengthen constituent engagement, leading to lower student acquisition costs, higher tuition and fundraising revenue, and greater student retention and outcomes.

In addition to potential constituent benefits, GenAI’s use for CRM may also help address the industry’s staffing and burnout concerns through smart automation and increased productivity. As one manager suggested, “I see automation of mundane tasks as a staff retention tool” [13]. The workload that GenAI could assume would free up employees’ time to focus on value-added tasks where direct human contact with constituents is most essential.

CRM vendors including Salesforce, Microsoft, Slate, and Campus Solutions are quickly incorporating GenAI into their higher education CRM offerings as they see considerable potential to add both value and revenue. Other technology companies like Ivy.ai, Brado, and GPTfy, are marketing GenAI integration into the CRM ecosystem. With strategic investment into GenAI applied to CRM use cases, HEIs may make significant strides toward addressing industry challenges and providing greater benefit to their constituents.

This study explores the opportunities and challenges for GenAI in higher education CRM applications and identifies key readiness and implementation factors. It was guided by the following research questions related to use cases for GenAI in higher education, industry challenges and opportunities, technology type classification, and implementation considerations.

• RQ1. What are the opportunities and challenges in utilizing GenAI in higher education CRM?
• RQ2. What factors guide readiness and successful implementation?

Research into GenAI opportunities for higher education CRM applications is in its infancy. Of the approximately 4000 scholarly articles on CRM and AI published in the last ten years, 70% have been written in the last five. Only 200 of those have focused on the higher education industry, 80% of which were written in the last five years. As we could find only six articles on GenAI related to higher education CRM, we believe our study contributes to a promising new area of research. Of the current six articles, five focus on chatbots and one on predicting student retention based on GenAI image analysis from student identification cards. Our study extends the current research by exploring the potential for integrating additional types of GenAI throughout the student lifecycle and determining readiness criteria and recommendations for successful implementation. This contribution clearly connects GenAI and higher education CRM and opens new lines of inquiry into use cases, evaluation, implementation, and adoption.

Backgound

The roots of CRM systems can be traced back to the advent of database marketing in the 1980s, followed by the emergence of sales force automation (SFA) tools and contact management systems in the late 1980s and 1990s. Contact management software tools like ACT! (launched in 1987, https://www.act.org/) and GoldMine (1989, https://www.goldmine.com/) helped salespeople track customer interactions, appointments, and communications. However, CRM systems as we know them now were introduced in the early 2000s as a way for organizations to better integrate their customers’ data into a single database and maintain a complete record of interactions with them through multiple channels like web, email, text, call centers, and social media [14,15,16]. CRM systems evolved to integrate with enterprise resource planning (ERP) systems and into a cloud-based software-as-a-service (SaaS) model. After 2010, CRM systems began to integrate with social media, mobile, email, and in-app interactions, enabling seamless business–customer interactions across multiple channels. As AI has evolved, it has been incorporated into CRM systems to provide descriptive, predictive, and prescriptive modeling capabilities.

The seminal work related to transformer networks and GenAI technology can be found in Vaswani et al., Devlin et al., and Brown et al. [17,18,19]. Vaswani et al., produced a landmark paper that introduces the transformer architecture that is used in many generative models due to its efficiency and scalability in handling sequential data without recurrent networks [17]. The work in Devlin et al., introduced bidirectional encoder representations from transformers (BERT) and advanced understanding of the context of words in search queries, thus influencing subsequent generative models [18]. Brown et al., demonstrated the unprecedented capabilities in generative pre-trained transformer 3 (GPT-3) for natural language processing (NLP) tasks using a “few-shot” learning approach [19].

Between 2010 and 2015, HEIs began implementing CRM systems for discrete use cases such as executive education or business school enrollment [20,21]. Following these initial use cases, universities explored CRM at the enterprise level. Today, HEIs use CRM systems for degree and non-degree recruiting and admissions processes, student progress tracking, academic and career advising appointments and tracking, alumni engagement program management and communications, fundraising and donor management, corporate relations, and many other individual processes and workflows. HEIs may have a plethora of CRM implementations at varying levels ranging from large-scale, university-wide applications to individual college- and department-level programs.

Although higher education often lags other industries in technology innovation, with the rise of GenAI, higher education finds itself in the unusual position of incorporating new technology at the same time as other industries. Other industries are finding that AI-powered CRM systems can leverage artificial intelligence to automate tasks, analyze customer sentiment, predict purchasing behavior, and offer virtual assistants for instant support [10,22,23]. The transformer network foundations of GenAI technology enable CRMs to generate content, draft personalized messages, and dynamically respond to customer queries in a conversational manner, greatly enhancing customer engagement and satisfaction [6]. Such AI-driven CRMs automate data entry, lead scoring, and communication while providing hyper-personalized interactions, allowing businesses to cater to customer preferences at scale. CRMs now support sophisticated self-service options such as AI agents that empower customers to resolve many of their issues independently and quickly. In higher education, GenAI may enable significant CRM capabilities such as analyzing and summarizing complex student data, modeling enrollment predictions, recommending interventions to prevent student attrition, individualizing constituent communication, and offering virtual agent assistants to provide student and alumni support and task automation. HEIs are exploring these capabilities to see how they might further optimize their administrative technologies and address their industry challenges.

2. Methods

This study employed a qualitative research approach based on the principles of grounded theory [24]. Grounded theory is particularly suitable for exploring new and emerging phenomena, such as the implementation of GenAI in higher education CRM applications, as it allows for the development of theories based on data collected from participants directly involved in the field and development of theoretical insights directly from data [25].

Data were collected through semi-structured interviews and an open-ended qualitative instrument, both of which posed the same questions and were designed to elicit unrestricted participant responses without pre-defined answer constraints. Data were collected from 20 participants from three key groups: higher education CRM and GenAI software vendors (4), implementation consulting firms (5), and CRM professionals actively exploring or piloting GenAI in higher education CRM contexts (11). This purposive sampling strategy represents a cross-section of key players in this field, and it ensured that participants possessed relevant expertise and experience to address the research questions effectively [26]. The sample size was appropriate for allowing us to reach theoretical saturation [24,27,28]. The semi-structured interview format provided a flexible structure, allowing participants to share insights and experiences while also enabling the researchers to probe into specific areas as needed [29]. The questionnaires allowed additional exploration and supported an exploratory approach to identifying themes related to GenAI in CRM applications.

Data analysis followed an open-ended coding process guided by the principles of grounded theory [30]. Open-ended coding was used to identify and categorize key concepts and themes emerging from the data [25]. This initial coding phase was followed by thematic analysis, which involved identifying patterns and relationships among the coded data to develop a comprehensive understanding of the phenomena under study [31,32]. A co-researcher reviewed and validated the coding and themes to ensure consistency and reliability. As an innovative approach to enhance trustworthiness, we utilized the GenAI code assistant GPT feature known as the Delve qualitative analysis tool (https://app.delvetool.com). Once the code book was developed to the point of saturation by the researchers, the tool was used to recommend the coding application of a subset of data, followed by researcher review. The researchers retained final coding decisions to ensure methodological dependability and confirmability [33]. This approach ensured that AI-assisted insights were integrated thoughtfully, and that final coding was grounded in human interpretation and oversight, which is essential for maintaining rigor in qualitative research [34].

Ethical guidelines were strictly followed throughout the research process. Informed consent was obtained from all participants, and their anonymity and confidentiality were maintained by using participant codes (e.g., P1, P2). Appendix A provides an overview of the participant roles. The study adhered to the ethical guidelines set forth by the governing institutional review board (IRB). This methodology allowed for a rigorous exploration of GenAI in higher education CRM applications, grounding our findings in the perspectives and experiences of key stakeholders while leveraging innovative tools to enhance the analytical process.

3. Results

We utilized semi-structured interviews and open-ended data collection instruments with CRM and AI software vendors, implementation consulting firms, and HEIs who have already made the decision to invest in GenAI in higher education CRM applications. All participants were knowledgeable professionals evaluating, piloting, and utilizing GenAI for CRM applications themselves or with their clients. De-identified participants are listed in Appendix A. Our questions related to how they saw the current and future state of GenAI in higher education; what challenges and advantages the industry faces in GenAI adoption; potential cost/benefit and return on investment (ROI) considerations; and what skills or resources are needed to make progress in GenAI adoption. From the transcripts of the interview and questionnaire responses, particular topics emerged as salient themes. These themes, identified in

Table 1. Summary of major themes from our dataset.

Theme	Frequency	Description
Administrative automation	High	Recognized as a significant area for efficiency gains in non-core tasks.
Chatbots and virtual assistants	High	Frequently cited as an accessible, low-cost AI application.
Data quality and integration	High	Noted as a key operational challenge for AI implementation.
Efficiency gains	High	Expected benefit, especially in streamlining administrative processes.
Financial constraints	High	A recurring barrier, especially for high-cost AI tools.
Student engagement and support	High	Seen as a primary goal for improving CRM interactions and retention.
Ethics and compliance	Medium	Focus on ensuring ethical AI deployment and maintaining data privacy.
Leadership and vision	Medium	Highlighted as critical for successful AI implementation.
Scalability	Medium	Cited as important for long-term viability of AI solutions.

, focused on various characteristics of the GenAI space, from types of GenAI applications to challenges and critical success factors.

3.1. Types of GenAI in Higher Education CRM

To address RQ1 regarding GenAI opportunities for higher education CRM, it is helpful to understand the types of GenAI that appeared in our data (listed in order of reference frequency in interviews and research):

3.1.1. Textual Analysis and Synthesis

Use cases for textual analysis and synthesis include analyzing surveys and feedback, automating knowledge article creation, and creating personalized communications such as recruiting emails, individual application reminders, student messaging, and personalized alumni newsletters or donor updates. Analyzing and creating text is the core of evolving chatbots into their next evolution as virtual agents. Virtual agents combine several capabilities like textual analysis and synthesis, data and text summarization, and code development. This was by far the most highlighted use case for GenAI in CRM throughout the entire constituent lifecycle and across all participants and research, and it could be become the “killer app” for GenAI in higher ed CRM [7,35,36]. Whereas early chatbots relied on a search and return of standard responses based on significant training and pre-developed utterance responses, today’s GenAI-enabled agents allow the analysis of structured and unstructured data to create personalized, conversational responses and workflow automations. For example, a traditional admissions chatbot might ask a prospective student to enter their name and cell phone number and ask if they would like to attend an info session. The chatbot would have to send the user to a separate registration form. A virtual agent can use contextual cues to anticipate the user’s intentions, such as using prior indication of interest in online classes to suggest online information session registration. The agent’s higher automation level can fill and submit the form, send a confirmation, and even suggest additional services based on context, such as recommending and scheduling an American sign language (ASL) interpreter based on a prior question about ways for hearing-impaired applicants to participate.

3.1.2. Data and Text Summarization

Generating summaries of multiple interactions and data points has great potential for higher education CRM. This technology is already in use for corporate use cases [37,38,39]. Early pilots at several schools include applicant interview summaries, transcript summaries, scholarship guideline summaries, academic and career advising summaries, and advancement donor contact reports.

3.1.3. Next Best Action Recommendation

This is an evolution of existing AI predictive analytics paired with GenAI summarization and text synthesis [40,41]. For example, in an enrollment use case, the system might utilize existing prospect data for regression and enrollment likelihood, summarize interactions, then generate conversion action recommendations to drive enrollment based on past analysis and training. For externally facing student use cases, this feature could generate course or internship recommendations. For advancement, it could generate next steps for donor outreach or alumni engagement opportunities.

3.1.4. Speech Synthesis and Translation

Many call centers already use speech synthesis to route incoming calls, and this feature can assist in recruiting as well as for student call centers and internal support services [42,43]. Consider the potential for real-time language translation on an inquiry call from an applicant speaking their native language. A listening bot would translate the question to the university’s native language, find the answer, and translate it back to the caller’s language in real time.

3.1.5. Code Development

Sometimes overlooked, code development is a crucial part of GenAI capability. This capability not only allows faster CRM system code generation based on prompts, but its power is infinitely exponential [38,41]. For example, GenAI could be used to create repeatable code for adding constituent relationship types to contact profiles based on interactions. Vendors are introducing CRM virtual agents that can be developed and operated by having bots create other bots. The AI itself can create code based on its own self-generated AI prompts. This is also a common practice used in training large language models (LLMs). Through having one model evaluate the results of another model by comparing it to expected results, AI can train another AI. One study participant from a higher education technology vendor with significant GenAI model development experience stated that, in their most recent project, this method of GenAI validation increased model accuracy from 74% to 90% (P17).

3.1.6. Image and Video Creation

Image and video creation was the least explored GenAI use case as its capability and ease of use lags behind LLM text synthesis capability. However, it has significant potential in marketing and communication personalization use cases [7,38]. An example of a higher education CRM application is the creation of GenAI images or videos personalized to individual recipients in recruiting materials promoting new programs or alumni engagement or welcoming newly admitted students.

3.2. GenAI in CRM Across the Student Lifecycle

The student lifecycle encompasses the range of activities in which a student engages, from their application through their time as a university student and into their post-university alumni experience. HEIs anticipate that the true potential power of these GenAI use case types lies not only in their individual CRM applications, but also in their systematic combination across the constituent lifecycle. These applications showcase the versatility of GenAI in enhancing various aspects of higher education CRM systems, from improving recruiting and admissions processes to supporting student success and alumni and donor engagement. Each type of GenAI can be applied to CRM use cases across the student lifecycle, as shown in

Table 2. Examples of GenAI types in CRM applications across the higher education lifecycle.

	GenAI Type	Recruiting and Admissions	Student Success	Alumni/ Development	Other
	Textual analysis and synthesis	Bots/virtual agents for admissions, application, financial aid questions Automated essay scoring	Bots/virtual agents for student wayfinding and cases Automating case routing and resolution	Bots/virtual Agents for alumni engagement and donor services Crafting personalized messages for alumni engagement and fundraising campaigns	Automating creation of program materials Analyze survey responses for sentiment and improvements
	Data and text Summarization	Summarizing applicant profiles to streamline the admissions process	Summarizing student data for advising or to identify at-risk students	Summarizing donor outreach and history for targeted outreach	Generating knowledge articles from case resolution notes
	Next best action recommendation	Recommending next steps for applicants based on their profiles and interaction	Suggesting student personalized advising, interventions, care plans, mentor matching, and resources	Recommending alumni engagement strategies or donor proposals based on their interests and interaction patterns	Providing recommendations for service and program improvements Recommend inquiry responses based on past engagement and interactions
	Speech synthesis and translation	AI recruiters handling large prospect volumes and multi-lingual support	Real-time advising translation Text to speech for student requests	Personalized messages in donor-preferred languages	Enhancing accessibility with text-to-speech for visually impaired students
	Code development	Developing automated CRM workflows for processing applications or generating custom reports	Creating tools to track and improve student engagement and performance. Code small apps for advising and program recommendations based on students’ data and preferences	Create custom scripts to segment and engage alumni based on specific interests	Developing CRM extensions to automate repetitive tasks across departments, enhancing operational efficiency
	Image and video creation	Creating engaging visual content for recruitment campaigns Generating personalized welcome videos for admitted students using student data	Producing instructional university process videos and interactive content	Designing visually appealing materials for fundraising events Producing alumni spotlight videos using images and achievements	Generating dynamic visual content for campus announcements or event promotions based on CRM interaction data

.

3.2.1. Recruiting and Admissions

Figure 1 shows the process for enrollment starting with prospective applicant marketing and outreach and finishing with successful student enrollment and matriculation. Various GenAI technologies are employed at different steps of this process. For example, during the marketing and outreach phases, summarization, text synthesis, and image creation technologies may all be deployed to create a more personalized applicant experience and raise their chances of submitting an application. In the application decision phase, the focus is on data summarization, textual analysis, and recommended next best actions to evaluate and convert candidates to students.

3.2.2. Student Success

Figure 2 shows a simplified version of the student success process. Many of these activities will occur repeatedly and concurrently. Virtual agents (ideally connected) could play a crucial role at each stage of this lifecycle. Advising and support for academics and career planning can benefit significantly from virtual agents for 24/7 student service and automatic routing of student questions and cases to the correct support queue. Summarization from multiple data sources will be important for building holistic student profiles and tracking graduation requirements and status. The recommended next best action AI can be used for matching organizations, events, mentors, and even internships or jobs based on students’ interests, skills, and goals.

Participants suggested that GenAI may offer significant value in addressing the ongoing mental health crisis. According to Inside Higher Ed’s 2024 Student Voice survey, which included responses from over 5000 undergraduates, significant numbers of students report that poor mental health impacts their academic performance [44]. GenAI virtual agents could offer 24/7 confidential support for students at times and through a channel they may likely utilize most with less inhibition than with other channels. Additionally, by predicting student wellness needs and recommending wellness care plans, it could help prevent student mental health crises.

3.2.3. Advancement (Alumni Engagement and Fundraising)

Figure 3 shows engagement with students after they graduate and become alumni (a status they hold for the rest of their lives). GenAI could play a significant role in tracking and promoting alumni engagement through personalized lifelong learning activities and targeted potential giving opportunities. Several studies suggest that personalized communication needs to start early to keep alumni connected, and that trying to win alumni back after communication lapses is extremely difficult [45,46]. Therefore, the text synthesis and personalization afforded by GenAI may substantially benefit this use case.

3.3. Higher Education Industry Challenges and Strengths

In assessing these potential applications, participants and HEIs specifically discussed challenges and strengths of the higher education environment related to RQ1.

“You’ve got the innovators. You’ve got the people that want to be the first ones out there to change. And then you’ve got the people that are more comfortable and set in their ways, who have been doing this forever and ever and been hugely successful. So why would they want to change? And you have to somehow figure out how to bring these two groups together, because you don’t want to be too fast and put the university and students at risk. But you don’t want to be too slow and be left behind, because clearly this is moving forward, and this is having an impact on universities, on work, and on society as a whole.”

(P2)

Higher education is simultaneously an environment of stimulating innovation and inertial stagnation. It houses the most forward-thinking minds alongside the most archaic of systems and processes. The GenAI space is crowded and fast-moving, at odds with higher education’s capacity for analysis and change. There are so many types of GenAI and so many potential applications within and outside of CRM that one participant noted, “There’s just a lot of competing technologies that make the space noisy and confusing to Higher Ed” (P6). Study participants highlighted challenges they perceived as unique to the higher education environment, particularly its fragmented and siloed data environment, the high need for and risks involved in ensuring ultra-tight security and privacy; uncertain cost/benefit and ROI; and the challenge of promoting employee adoption. They also identified industry strengths born of the nature of working in higher education, such as a strong commitment to student development, significant knowledge sharing within the institution and with other HEIs, and their growing experience with technological change. Highlights of these factors can be seen in

Table 3. Strengths and challenges of the higher education industry related to GenAI in CRM.

Industry Factor	Description	Illustrative Quote
Strengths
Knowledge Sharing	Information and best practices shared inside the intuition as well as with other HEIs industry-wide	“The ability to share is a superpower. And then, maybe because AI comes more out of the academic world than other kinds of technology. Maybe this is the moment when we could bridge our kind of ‘researchy’ superpowers on one side with how we run our institutions.” (P4)
Commitment to students	Dedication to student success as a vocation	“We have a vested interest in providing for the students. What can we do to meet the students’ needs? What can we do to make this a better experience? That breeds a curiosity.” (P6)
Technology pivot experience	COVID-19 response developed technology change skills	“Who ever thought higher ed could change technology that fast? But we did it, and now we know how.” (P19)
Challenges
Organizational silos	Decentralized structures and budgets lead to disconnected technologies and conflicting goals	“Subject matter experts in the functional areas of the institution often have nothing to do with the content on the website.” (P6)
Slow processes and decisions	Committees and competing priorities create a slow pace for decisions and execution	“I would even say it’s slow to adopt. I think it hinders the progress of a lot of institutions. It’s the whole, you know, overanalyzing. And so you just are paralyzed by all of the different people involved in the conversation and the opinions and the concerns.” (P8)
Poor data environment	Data housed across units and in many disconnected systems with no collective ownership points	“Data’s a mess. And it’s not an isolated issue like, ‘Wow, we’ve really found that schools in the Northeast have it.’ It is a very wide issue.” (P6)
Highly regulated data environment	Security and privacy are highly critical and regulated	“We have to protect them. If students use the tools <unsecure GPTs and apps> on their own, they’re happily signing away their privacy.” (P7)
Difficult cost/benefit and ROI calculation	Structure and lack of metrics focus make measuring success challenging	“Oh, that’s great! The student affairs cost went down. But if you actually look, it might not be a positive ROI if you didn’t account for the IT costs. Cost goes down in one place, up in another. And are you comparing all together for ROI?” (P6)
Adoption	Overworked staff may resist what they perceive as more work and a job-threatening technology	“You’ve got the people that are kind of more comfortable and set in their ways, who have been doing this forever and ever, and been hugely successful. So why would they want to change?” (P2)

.

4. Discussion

The research findings suggest a wide array of possible GenAI use cases for CRM applications across the higher education constituent lifecycle. A gap in current research lies in the evaluation criteria by which to assess readiness for implementing GenAI CRM applications as well as implementation considerations once a project is selected. To address this gap and RQ2, we present a table of readiness evaluation factors and scoring to identify the most valuable and viable GenAI CRM opportunities for an HEI and offer a discussion of considerations for implementation strategy and adoption.

4.1. Evaluating Readiness for GenAI in CRM

Study participants indicated that GenAI CRM application ideas come from both business as well as from IT, and that the project selection process can be challenging, even for experimentation. One implementation consulting firm cautioned, “Be selective and start strategically…he wants to use it for teaching assistants, and they need it for their help desk, and you’re having an issue with onboarding. And so now we’re trying to solve everybody’s problem in a very complex institution where there’s different organizations within the institution. And all of a sudden nothing gets fixed. None of it happens” (P6).

The ideal GenAI use case becomes the killer app for driving greater CRM adoption and value. When AI teams at Google [47] consider project starting points, they look for “AI-shaped problems” with criteria that make them suitable for AI such as “huge combinatorial search spaces, large amounts of data, and a clear objective function to benchmark performance against,” and save for later projects those that seem viable but lack key criteria like connected data. To select potential starting points in higher education, it is helpful to consider a set of readiness criteria for each potential use case. Based on our findings and current scholarly literature,

Table 4. Readiness criteria for selecting GenAI use cases in higher education CRM.

Readiness Criterion	Description	Key Questions
Strategic alignment	The degree to which this application fits with the HEI’s current strategy and mission	How well does this application fit our strategy? Would this application make it to the top of our leaders’ priority lists?
ROI	Evaluation of the actual ROI or organizational benefits compared with the costs of implementation and operation	With what revenue-based ROI is this associated? How will it be measured? What are the actual costs and benefits across units, and who will accrue them? Will our budget structure allow us to measure this well enough?
Feasibility and implementation speed	Analysis of the knowledge, time, integration complexity, LLM model availability, and all other factors for assessing viability and time to implementation	Can we do this ourselves or do we need an implementation consulting firm? How fast? Do we need an open or closed LLM? If closed, how will it be built, or does it already exist? What systems, people, or other factors have to come together?
Data readiness	The quality, availability, and interoperability of the data upon which the GenAI will rely—the 4Vs of big data: volume, velocity, variety, and value	How clean are our data? How integrated are our data? Do we know source systems, order of operations and integrations, and data survivorship rules? How much structured versus unstructured data are there? How is our system interoperability?
Ethics, privacy, and security ease	The levels of ethical considerations, privacy and security compliance concerns, and regulatory oversight of the involved data	What are the ethical implications of this use case? What regulations apply to this data? What is our security posture? How much risk exposure does this application present?
User adoption likelihood	Willingness, speed, and continuity of user adoption	Why might users adopt/not adopt? How easy is resistance to overcome?
Scalability and flexibility	The degree to which this application can be scaled and/or easily tailored for other uses	How easy is this to scale? Can we use this for other purposes easily?
Measurability and monitoring	The ability to develop and measure key performance indicators (KPIs)	Can we measure performance? Do we have baselines upon which to compare performance? Who will track metrics?

provides readiness criteria that may aid in opportunity evaluation. Each criterion topic is framed with a description and key questions to consider.

In considering these readiness criteria, HEIs can evaluate individual use cases in their own institutions. The criteria could be scored from low (1) to high (5) for the fit and readiness of each individual use case given the current environment. For example, if a university’s top priority is increasing enrollment, a high “Strategic Alignment” score might be assigned to an admissions virtual agent. A high “Data Readiness” score for that use case would mean that the good quality data needed for this use case are available, including knowledge articles, university policy documents, and access to prospective student information. To compare scoring for use cases, a column could be added representing the total score across criteria assuming equal weighting. Weights could be added to each criterion based on organizational decision-making criteria. Figure 4 represents an example of a sample scoring table.

The scoring could also be used to compare readiness criteria across multiple use cases. For example, some prospective student use cases could score better on Data Readiness because they require less data effort (there is simply less data and fewer integrations), and these data are not as heavily bound by regulatory and other privacy and security constraints such as FERPA. In some institutions, an undergraduate program virtual agent might score higher on Feasibility and Implementation Speed than a graduate program agent, because graduate admissions criteria and processes differ widely across programs and are harder for an agent to handle. In the realm of student success, mentor matching may score lower than imagined on Data Readiness, because it requires significant data on both students and alumni, two different populations that are often ‘owned’ by entirely different institutional units with disconnected processes and systems.

4.2. Implementation Strategy Considerations

“Don’t overthink it. I think I would even say to the slow to adopt: I think it hinders the progress at of a lot of institutions. It’s the whole overanalyzing. You are paralyzed by all of the different people involved in the conversation and the opinions and the concerns…Don’t try to boil the ocean in higher ed. We’re at a place where AI needs to be adopted in in one application or use case or another, and you’ve got to dip your toe in or you’ll get left behind.”

(P7)

Study participants overwhelmingly believed that HEIs should consider where GenAI might best benefit their CRM applications and pilot possibilities quickly, learning and adapting as they find solutions. While HEIs should consider GenAI implementation in alignment with their strategic missions, adoption readiness, data environment, and financial capacity, participants warn against analysis paralysis, asserting that fast experimentation is vital for ramping up on capability and results in GenAI. Based on our findings, we offer the following implementation considerations.

4.2.1. Do Data First

“I’m afraid of the people who think that in the mediocre data environment that I find, I’m gonna have wonderful outcomes with AI…Here’s an opportunity to create bad information because the inputs were wrong, and we can do it at a blazing pace in creative new ways.”

(P4)

Considering the 3V’s of big data: volume, velocity, and variety, HEIs are rich in all three [9]. However, the most frequently identified challenge by study participants was the poor state of their institutional data, robbing them of a potential fourth V for value [11]. Simply put by one implementation consulting firm, “Data’s a mess” (P6). This issue is critical because vendors, consultants, and researchers consistently identify it as a key factor for successful GenAI optimization.

Data quality, integration, and ownership are notoriously challenging in higher education. HEIs still have many on-premises solutions, where the data and application are hosted on individual systems within the institution rather than cloud-based systems where data are more easily shared and applications more easily updated. Data are also fragmented across different units of the institution, who create micro-solutions to suit their individual needs and become artificial gatekeepers [48]. This leads to replication of data in multiple places with different formats and uses. Local information is often outdated and incomplete; central systems are disconnected and inaccessible.

Before undertaking an implementation, HEIs should inventory the data needed to feed the GenAI and evaluate the quality of the required data environment. Then they might choose use cases where the data are most ready while shoring up other data for more complex applications.

HEIs must find ways to collect, streamline, integrate, and share data to unlock its potential. Some study participants suggested that data orchestration could be undertaken in many ways, whether in multiple data repositories or few, provided the GenAI tools have adequate access and there are no conflicting data or there are rules for resolving conflicts. New data unification platforms are making the job of translation among systems easier, with less need for integration from system to system [49]. For several HEI study participants, their pilot programs focus first on data cleaning and aggregation to ensure readiness for GenAI applications, to “get the data to a place where it could use GenAI” (P1).

Implementation consulting firms and technology vendors included in our study stress the enormity of the data work required to make GenAI useful and safe for higher education CRM—to create guardrails for answers and recommendations, elicit quality responses, and provide accurate summaries. For externally facing GenAI uses like virtual agents, these guardrails will be especially critical. They note that LLMs have nuances and uniquenesses and that the higher education content on websites and FAQs was never written for LLMs and chatbots. It takes time and extensive work to curate content and test the AI in order to provide the best possible experience for constituents and to protect the institution.

4.2.2. Prioritize Credibility, Security, and Privacy

“We’ll need to do more than keep the human in the loop. We’ll need the human at the helm.”

(P13)

HEIs are stewards of sensitive information about many constituent types, from demographics to grades to health data to credit cards, and the responses GenAI provides will have significant potential impact. This data environment requires a sophisticated and highly protective security posture [50]. Therefore, HEIs should be judicious about their entry into GenAI, and carefully manage the tension between caution and speed.

HEIs require adherence to significant data privacy regulations, and participants raised the risk of leaked or hacked sensitive data. They referenced HEIs’ responsibility to protect sensitive information and evaluate how much risk they are willing to assume. HEIs would not want a virtual assistant to mistake one student for another and reveal personally identifiable information (PII), grades, or financial status (P5). However, they warned against too much caution causing delay, for they believe that if HEIs do not build suitable GenAI fast enough, students will simply use open GPTs: “if students use the tools on their own, they’re happily signing away their privacy” (P7).

The demands of AI-specific security require protection of AI models from tampering and unauthorized access. It is also necessary to ensure that inputs to AI models are sanitized to prevent adversarial attacks. Further, it should not be possible to reverse engineer training data to identify individuals. Models should be trained on decentralized data without transferring sensitive information. There is also a need to implement mechanisms to detect and mitigate biases in AI models that could lead to unfair outcomes.

Finally, HEIs must ensure that the GenAI applications in CRM offer reliable output and avoid “algorithmic opacity” [51]. Early GPTs trained on public data suffered from hallucinations and other challenges to their credibility. As they improve, HEIs will need to vigilantly train and check output to ensure trustworthiness and value. Trust in the responses GenAI provides, whether through an agent, summary, sentiment analysis, or code, will be paramount to adoption.

4.2.3. Measure Cross-Organizational ROI

“You don’t want to do AI for AI’s sake. There’s really no point to that.”

(P2)

Each GenAI use case, and experimentation with it, needs to have clearly stated and measured value to the organization. Two potential GenAI benefits that participants and research highlight are staff time savings and improved service leading to positive constituent outcomes [6,10,52]. While all see the potential for these outcomes, most HEIs are still in the exploration and pilot phases of GenAI in CRM applications, and they struggle with measuring return on investment (ROI).

Realizing ROI could pose challenges for higher education on the grounds of both cost and measurement. Participants expressed concerns about the high potential cost of GenAI development, noting that “this is actually pretty expensive at scale” (P5). GenAI is likely to be a costly technology considering the preparation of all of the data, the creation of the LLMs, the training of the models, the staffing, the organizational adoption strategies, and the need to keep up with the pace of technological change (P6). In purchasing external GenAI products, HEIs will find a landscape of expensive technologies that do not understand the higher education budget landscape. One participant stated that, “a top executive on campus shared the prohibitively expensive licensing cost to acquiring ChatGPT (https://openai.com/index/chatgpt/, accessed on 4 February 2025) for educational institution and that it was a non-starter” (P11). Not only are they expensive, but pricing is a moving target. What was originally seat licensing might suddenly change to a consumption model as GenAI technology vendors wrestle with how to best monetize this new technology [53,54,55]. With most HEIs still recovering from COVID-19 era budget challenges and facing additional financial strain, investing in new GenAI technologies will be a challenge. HEIs will need to utilize evaluation and readiness criteria to determine the likelihood of success balanced with potential ROI on each use case investment.

Measuring ROI is critical yet challenging based on higher education organizational accounting structures. Much research cites cost reduction from staff time savings as a major potential benefit of GenAI in CRM and other use cases [3,6,35,38,56]. Costs may be saved in the units providing service to their constituents, though costs in technical units are likely to increase. For example, one technology vendor observed, “The organizational structure is different at every institution. So who owns the budget? It’s a mix, right?” (P6). If HEIs cannot aggregate the savings and expenditures on account of GenAI technology across units, true ROI cannot be measured.

One way to justify anticipated GenAI-based staff savings comes from an Accenture 2024 report analyzing all job categories in the economy. By analyzing data from the Occupational Information Network (O*NET) of the US Bureau of Labor Statistics, Accenture estimates that GenAI will impact about 40% of the work we do today through automation or augmentation [41]. If staff time savings are to be realized, HEIs will need to be able to measure those savings through productivity metrics or outcomes.

4.2.4. Focus on Adoption

“AI in Education can only grow at the speed of trust.” [57]

GenAI has entered the workplace quickly, with little preparation and less training. University staff may not only be reluctant and distrustful but also fatigued while resetting from major COVID-19 era disruption. Focusing on overcoming adoption challenges may aid the uptake of new technology. While our research suggests that the key factors contributing to the success of CRM-related GenAI applications are mostly like factors in other technology projects, some adoption challenges are unique and require additional consideration [58,59].

An initial adoption challenge is trust in the GenAI itself—its models and explainability [60]. In open commercial GPTs, the output process is hidden behind the models and unknown training data. Even in closed, internally built and tuned models, explainability is challenging. One participant noted, “Even with the same prompt, because of the stochastic nature, you can ask the same question and get a different answer every single time” (P7). While both open and closed models are getting better, one study suggests that people trust their intuition more than AI and are more likely to hold on to their “sticky routines”, thus resisting adoption of AI [51]. A 2024 New York Times article reported on a study in which ChatGPT alone outperformed both doctors who used and who did not use GenAI, in part due to these routines [61]. Lack of training and overconfidence in their own experience led them to trust their own diagnoses even when the GPT offered conflicting evidence.

Not only do we often trust our own knowledge and experience over AI, but we also believe that human emotional capabilities exceed those of GenAI. However, new research may challenge this assumption. A recent study has shown that AI can be better at dealing with emotions than humans [43]. It has been found that, when compared with humans, AI demonstrates superior discipline in offering emotional support, an element crucial for student support services, while avoiding excessive practical suggestions, which may be less effective in achieving this goal. Even a recent Wall Street Journal article cited an Allstate Insurance discovery that their customers found their GenAI communications more empathetic than live agents [62]. If this capability can be harnessed appropriately, then constituent-facing GenAI applications could be highly suitable and valuable candidate CRM use cases, despite the initial reluctance to accept that an AI is capable of offering empathy.

The growing availability of GenAI in other enterprise technologies may help address some of these adoption challenges. University staff who would be engaged in CRM GenAI applications are likely already experimenting with GenAI in other applications. Most major technology platforms are adding GenAI to their products in ways that encourage use and increase familiarity. Microsoft Office and Google offer GenAI for writing emails and designing presentations. Microsoft Teams and Zoom offer AI-generated transcripts and meeting notes with action items that are widely used in higher education. These applications are doing the heavy lifting of overcoming GenAI resistance and fostering experimentation and adoption in secure, low-risk environments. Additionally, we are all customers experiencing experimentation in GenAI by commercial industries such as Allstate’s insurance agents, wireless carriers, and other service providers. As these technologies improve, staff members’ lived experiences may aid their acceptance of GenAI in their own work. In early GenAI projects, staff could be selected for their technological openness in addition to their functional capabilities. Once GenAI experimentation has started, the projects themselves can spark innovation and optimism, which Popa et al. suggest are operative competencies needed for GenAI adoption [52]. From there, ongoing training updated at the speed of AI’s rapid evolution will be of paramount importance.

Another major adoption challenge for GenAI in higher education CRM is the view of GenAI as a threat to staff jobs. With popular media predicting that significant portions of human work can be replaced with GenAI, employees may seek to protect their jobs by finding rationalizations for not implementing the new technology [41]. This is particularly true for staff in front-facing applicant or student service roles. Staff may be concerned that their performance will now be measured against GenAI, which they could soon expect to outperform them [63]. Addressing this challenge requires strong leadership support, which most current technology and CRM implementation literature identifies as a key factor for implementation success [64,65,66,67]. Good leadership implies clarity of intentions and measurement, clear communications of outcomes, and actions consistent with stated intentions. This includes leaders specifying ROI metrics for their projects and communicating how they will address the meeting or missing of those metrics. For example, if staff time savings is a project objective, communicating what is intended for those saved hours—repurposing toward different work, reducing roles by retraining or moving, or even acknowledging that current staff workload is already over expected hours, with the initiative intended to bring it to sustainable levels to reduce staff overload and burnout.

4.2.5. Embrace GenAI as a Co-Intelligence

“Always invite AI to the table.” [68]

Participants and research suggested that the most beneficial way to overcome GenAI adoption resistance is by approaching GenAI as a co-worker and partner in performance—what GenAI researcher Ethan Mollick describes as a “co-intelligence”, or a knowledgeable colleague, co-worker, and coach balancing and benefitting everyone’s performance [68]. Rather than answering questions about admissions or add/drop deadlines, staff can focus on conversations that utilize their higher-level skill sets. A Deloitte white paper suggests that, while AI benefits tasks requiring high accuracy and high creative difficulty, humans outperform AI in tasks with high context variability in the social relationship realm requiring emotional intelligence and complex decision-making within ambiguity [69]. These are the advising conversations in which admissions, academic, and career advisors can navigate complex organizational systems and foster positive student experiences—even help change lives.

We might also utilize GenAI as a co-intelligence to reduce human error and strengthen new hire competence. Some experienced student advisors and counselors may miss opportunities or concerns based on their own biases or mistakes. As shown in the “doctors versus AI” study, humans often make decisions based on intuition and experience, making their thought processes singularly unique; they might miss some detail or make a mistake [61]. Other staff are new to their jobs and do not have significant experience to bring to bear on the role, especially in departments with high turnover like admissions and academic advising. This would be an appropriate opportunity for a GenAI that is trained by experienced staff to double-check care plans or to help support new staff members with training, role play, quick answers, and ongoing coaching. With appropriate staff training and openness, GenAI has the potential to improve overall performance. It could even expand the thinking of experienced staff, and it could offer experience and ideas to newer staff without taking time away from other employees.

4.3. Limitations and Future Research

Our study reflects only the current state of GenAI, and the rapid advancement of the technology means that this research is limited to its early development. Future research would be able to evaluate industry experience as technology advances. The sample size for this research was also limited by the early stage of industry experience with GenAI in CRM. It may not reflect the diversity of experiences and perspectives of CRM and GenAI technology vendors, implementation consulting firms, or universities. Once the industry has more experience in this space, future research could incorporate a larger sample of perspectives, including the students, staff, and other relevant constituents.

Another limitation of our study is selection bias due to the participants’ inherently positive disposition towards GenAI in CRM. The sample for the study intentionally selected technology vendors, implementation consulting firms, and universities who were already investing in and experimenting with GenAI in CRM. This sample was chosen to ensure that the respondents would have experience with the technology and would be able to answer the study questions. In general, these participants had an optimistic view of GenAI and its potential benefits and this could introduce a bias in our study. A future study could be conducted with a random population sample to explore overall experience with and sentiment toward GenAI use in CRM applications.

This study utilized an exploratory methodology appropriate for the state of HEIs’ GenAI and CRM experience. No institution has at this time presented outcomes of implementations or even pilot projects. We were unable to evaluate GenAI outcomes on specific prototypes, as few HEIs have measurable results at this stage. Future studies could share measurable outcomes and learnings from implementations. To increase validity, additional research could be conducted related to the growing experimentation of GenAI use cases to identify results and best practices for future consideration.

Lastly, additional research opportunities would include a quantitative validation study of use case readiness evaluation criteria as the foundation for a critical success factor assessment as well as further work to understand the ethical implications of GenAI in higher education CRM.

5. Conclusions

This research has explored opportunities for GenAI in higher education CRM applications based on a qualitative study of technology suppliers, implementation consulting partners, and HEIs who are pursuing GenAI in CRM use cases. Our study identified the five types of GenAI with the most potential for CRM applications and showed their application through possible use cases within the higher education constituent lifecycle, including the highest prevalence of virtual agents. To assess readiness for GenAI in CRM projects, we have presented eight criteria of technical, organizational, and ethical factors for individual and comparative evaluation. Based on the industry’s unique strengths and challenges, we find that successful implementation strategies for GenAI projects must focus on technological elements like data readiness, data and model integrity, and fast experimentation methods, as well as on organizational elements like success measurement and staff adoption.

Our research suggests that technology vendors, implementation consulting partners, and HEIs who are exploring this space believe that GenAI applications within CRM use cases have considerable potential to alleviate higher education staff burnout and shortage concerns, reduce the friction of doing business with universities, increase university enrollment and funding, and allow learners to find their best possible fit for an educational experience and a lifelong learner relationship. At the beginning of the journey, HEIs must remember to put the human at the helm, taking calibrated experimental steps into GenAI for managing their many constituent relationships and then embracing those which provide the best benefit for the university and their constituents.

“As leaders, we are lucky if we have one opportunity in our careers to identify a genuine catalyst for monumental change. Gen AI is that opportunity.” [70]