The integration of Generative AI (GenAI) in higher education promises many tempting benefits – efficiency gains, cost saving, quality of service, prevention of human error, pride(!) – but it also presents a complex series of challenges and strategic dilemmas. In a recent talk I gave to university and business school leaders at the recent AACSB AI Conference in Pairs, I explored the critical decisions facing institutions as they look to adopt GenAI, the reasons behind both the successes and failures of past AI projects, and how institutions can position themselves for long-term success. This article expands on that talk:

Key Strategic Dilemmas: Build, Buy, or Borrow?

When it comes to integrating GenAI into the university ecosystem, the first question institutions must answer is whether to build their own AI tools, buy existing solutions, or adopt a hybrid approach. There are pros and cons to each option, but the choice is not black and white.

1. Build vs. Buy

Building a custom GenAI system offers complete control and customisation, allowing the technology to align perfectly with the institution’s unique needs. However, it demands significant investment in both talent and infrastructure. In contrast, buying a ready-made solution might be more cost-effective in the short term, but the trade-off is less flexibility and the risk that the purchased system may not align perfectly with the institution’s long-term needs.

2. Agency or In-house

Should the institution rely on external agencies to implement GenAI, or should it build internal teams? Agencies may bring specialised expertise and speed up deployment, but they often lack a deep understanding of the institutional culture. In-house teams, while slower to ramp up, can ensure long-term ownership of AI solutions and are better positioned to adapt them as the institution evolves. This is assuming that you are able to hire the necesssary technical experts in the first place, and retain them in the face of job market competition. The loss of a key technical person in a small project team can easily derail a development project.

3. Bootstrap or Big Budget

Another dilemma is how much funding to allocate. Is it better to start small, with a bootstrap mentality, testing GenAI on limited-use cases and scaling gradually? Or should institutions invest heavily from the outset, betting on a big-budget approach that assumes GenAI will be transformative? The answer depends largely on risk tolerance and available funding. Institutions should avoid falling into the trap of underfunding critical early stages, which can lead to implementation failure, but they should also be wary of overspending on technology without ensuring the validity of the solution. It wouldn’t be the first time that a shiny new tech development goes unused due to a lack of take up.

4. How to Future-Proof

A key concern is future-proofing AI systems. As technology evolves rapidly, institutions must plan for adaptability. Investing in modular, scalable systems and maintaining a flexible roadmap is essential to ensure that AI solutions remain relevant in the face of technological shifts.

5. Commercialisation or Competitive Edge

Finally, should universities seek to commercialise their AI innovations, rolling the new tool out as a start up supported by their entrepreneurship infrastructure. This allows for the tool to continue to grow and develop making use of funding from other institutional customers, or should the focus be on gaining a competitive edge by leveraging AI to improve the student experience, retention rates, and educational outcomes? This decision often hinges on the institution’s strategic priorities and ability to invest in and maintain the new technology.

Understanding Strategy through Failure

Many institutions (and private companies) have attempted to integrate AI, and their failures offer valuable lessons. These pitfalls often stem not from the technology itself, but from strategic missteps, misunderstandings, and misalignment of goals. The examples below are also not exclusive to AI projects but can often be applied to any technical development.

1. Failure of Leadership

Leadership plays a crucial role in guiding AI implementation. A common failure occurs when leaders optimise for the wrong problem, pushing AI to solve issues that don’t actually benefit from advanced technology. For example, using GenAI to automate routine administrative tasks might be an overkill solution, where simpler automation tools would suffice. Similarly, overconfidence in AI can lead to unrealistic expectations, and underestimating the time investment needed for successful implementation can cause projects to stall before they generate value.

2. Project Failure

On the ground, AI projects can fail due to a lack of suitable or balanced data, a critical component for machine learning models to perform accurately. Even when good data exists, AI initiatives can falter due to a lack of institutional or industry knowledge within the project team. Grassroots buy-in is another critical factor, and projects are more likely to fail when key faculty and staff aren’t invested in the technology’s success. Additionally, AI teams can sometimes become disconnected from the real-world problems they are trying to solve, which leads to solutions that look great on paper but don’t address practical needs.

Components of Success

In contrast, successful AI implementations focus on clear, strategic thinking, with a strong emphasis on understanding the problems at hand and creating sustainable support systems.

1. Focusing on the Problem, Not the Technology

The most successful AI initiatives start not with the technology, but with a clear focus on the problem to be solved. This backward design approach ensures that AI solutions are purpose-driven and directly aligned with institutional goals. By asking “What problem are we trying to solve?” rather than “How can we use AI?” institutions can avoid the trap of implementing technology for technology’s sake.

2. Choosing Enduring Problems

AI solutions should be aimed at solving enduring problems — issues that will persist and evolve over time, rather than trendy challenges that might be irrelevant in a few years. For instance, improving student retention, enhancing personalised learning, and optimising faculty workloads are long-term challenges that can benefit from AI-driven solutions.

3. Investment in Dedicated AI Support Teams

Institutions that have successfully implemented GenAI often do so with dedicated internal AI teams, ensuring that the project is fully integrated into the organisation. These teams are responsible for managing data, developing solutions, and providing ongoing support. By investing in internal talent, universities can create AI systems that evolve with the institution, rather than relying on external consultants who may lack long-term commitment.

4. Context Awareness in Technical Teams

Technical teams must have a deep understanding of the educational context in which they operate. This allows them to create AI tools that are not only technologically sound but also relevant to the specific challenges facing educators and students. Cross-functional collaboration between technical experts and educators is critical to ensuring that AI solutions are both feasible and practical.

5. Planning for Evolving Technology

Given the rapid pace of AI development, institutions must plan for continual evolution. This means setting up systems that can adapt as new technologies emerge, ensuring that today’s solutions don’t become tomorrow’s obsolete infrastructure. A phased approach, with regular reassessment and updates, is critical to staying ahead of the curve.

6. Early Pilot Testing

Pilot programmes are invaluable for testing GenAI implementations in real-world scenarios. By rolling out AI solutions in small, controlled environments, institutions can quickly gather data, refine their approach, and ensure that the technology delivers tangible benefits before scaling it across the entire institution. If the project is going to fail, it should ‘fail quick, fail small’.

7. Ongoing Training and Support

Even the most advanced AI systems won’t deliver results if faculty and staff aren’t trained to use and understand GenAI. It is as important to understand what can and as what can’t be handled by this new technology. Institutions should prioritise ongoing training and support to ensure that AI solutions are fully adopted and integrated into day-to-day operations.

8. AI Awareness and Training for Leadership

Finally, institutional leadership must be AI-literate. Leaders should have a clear understanding of AI’s capabilities and limitations, allowing them to make informed decisions about where and how to invest. Regular AI awareness training for university leadership can help avoid strategic missteps and ensure that AI initiatives are aligned with the broader goals of the institution.

---

Implementing GenAI in higher education is an exciting opportunity, but success requires careful planning, strategic alignment, and a focus on the long-term goals of the institution. By learning from past failures and focusing on the components of success, universities and business schools can harness the power of AI to transform the educational experience for both students and staff.