Online: Trending Now

Artificial Intelligence (AI) is rapidly becoming a cornerstone of contemporary practice in higher education. This summer is an ideal time to become AI literate for the fall.

Perhaps you just haven’t had time to keep up with the advent of agentic AI. Or, you simply didn’t realize that AI is not just a fad in higher education, rather it is our best hope to survive the pervasive budget cuts and dropping enrollments. Or, perhaps based on your look at AI tools in 2023, you found them too unreliable and subject to “hallucinations” to take them seriously. Here’s hoping that this summer provides you a bit of time spread across the season to “catch up” with the technology and begin the fall term with the knowledge and experience to make AI the best professional assistant you have ever had.

Instead of facing a stack of projects and problems alone, you will have a PhD-knowledge virtual assistant working at speeds far exceeding human thought:

According to OpenAI, o3 earned a record-breaking score on the ARC-AGI visual reasoning benchmark, reaching 87.5 percent in high-compute testing—comparable to human performance at an 85 percent threshold. The model also scored 96.7 percent on the 2024 American Invitational Mathematics Exam, missing just one question, and reached 87.7 percent on GPQA Diamond, which contains graduate-level biology, physics, and chemistry questions.

These assistants work 24/7, without vacation or holiday breaks. At the end of long day of work, you can pose complex problems to an advanced deep research model and it will conduct research, compose a detailed report, and prepare follow-up questions while you are eating dinner (I have experienced this myself; reading the results over dessert so much more satisfying). I have also awakened in the early hours of a Sunday morning with a great idea to pursue for work. I tapped it in directly to one of the advanced models and awakened to a 20+ page report complete with comprehensive citations and suggestions for further research later that morning.

AI developers have made great strides in avoiding hallucinations and off-target results. Among the improvements are the utilization of Retrieval-Augmented Generation (RAG), a natural-language processing technique that taps an expansive database to “enhance the context and accuracy of generated text.” Those who may have stopped using AI because of errors of the past will be surprised to see the far more consistent and accurate results of today. Despite these improvements, I continue to recommend that users submit nearly identical prompts to two or three of the leading models of AI. Although my motivation for that recommendation primarily is that you are likely to gain additional, useful information from the added results.

Today, most of the leading models provide “multimodel” features that can input, process and output various types of data beyond text, such as images, audio, and video. This enables engaging the prompter via voice communication. It supports the generation of stunning images and rather brief video segments. Google’s Notebook LM tool accepts input documents and related media about which it can create a podcast, allow listeners to ask questions and get audio responses, create crossword puzzles based on your lecture notes, and even create virtual debates.

The tools I most commonly use are those from OpenAI, particularly o3; those from Google, particularly Gemini 2.0 Flash for general work and Gemini 2.5 Pro for detailed research; and Anthropic’s Claude 3.7 Sonnet. However, there are many more models available today that may better meet your needs or preferences. One of the projects I have been pursuing lately is tracking research, new treatments and other emerging information about a particular disease. I am using Gemini 2.5 Pro and ChatGPT o3, running updates every week. The results have been comprehensive and well cited. Notably, ChatGPT o3 noticed that I had been asking for reports every Tuesday and asked if I would like it to run the same prompt every Tuesday and email the results to me. Such an action could be considered elementary agentic AI in which the tool can analyze needs, create a plan of action, and with permissions take autonomous actions:

Agentic AI is an AI system that acts autonomously, adapts in real time, and solves multi-step problems based on context and objectives. They are built of multiple AI agents that leverage large language models (LLMs) and complex reasoning. This enables them to have enhanced decision-making abilities and natural language understanding, facilitating more effective and intuitive user interactions.

The power of AI agents is only beginning to be realized; 2025 has been dubbed the year of the agent. It is anticipated that millions of agents will be created by the end of the year. Their potential is enormous, reaching beyond the individual to take actions on behalf of a human.

So, how can you get on top of this AI trend this summer? Prompt engineering, asking questions in proper context, detail and format, is a good place to begin. You might consider enrolling in one of the many low-cost or free prompt engineering online classes. The Google Prompting Essentials course is $49 through Coursera. It takes just a few hours, and successful completion results in a certificate. There are also a number of YouTube videos that condense the contents of the micro-course. You may want to browse the Coursera catalog section on prompt engineering that lists scores of classes from industry leaders, commercial vendors, colleges and universities that last from a few weeks to a few months in length. Many provide professional certificates.

I suggest you begin a course, many are self-paced, or one of the informal YouTube videos, then begin using the tools at every possible opportunity. “Iterative prompting” is the name of the game. Try reframing the prompt, providing additional information, and including examples of what you are seeking, for the tool to better understand your expectations.

Use your one or two chosen tools as often as possible. Ignore non-AI search tools for a while. You will notice that AI searching gets right to the best solutions rather than first listing the responses in order of those who paid for their place in the search response, such as “googling” a question does.

Searching topics as they arise four or five times a day, refining each of those search prompts to better understand the capabilities of each tool, and searching across a wide variety of topics and disciplines will advance your expertise and comfort with AI. By the time classes begin in the fall, you will be prepared to save much time and effort by using AI. You will also be able to integrate AI into your daily routine, become more productive, and share your expertise with your colleagues and students.

This article was originally published in Inside Higher Ed.