Daylong “Sprints” Reimagine AI Education

• In the One-Day Case Challenge at East China Normal University, students rely on AI tools to analyze a company’s digital transformation strategy and provide alternative solutions.
• Instead of accepting the first results from one AI platform, students seek a variety of perspectives, employ multiple tools, require end-to-end traceability, and continuously refine prompts.
• The school also hosts a boot camp in which participants use AI development frameworks to compress the entire innovation cycle into a day, ending up with a functioning prototype.

 
In 2026, China drew a clear line in the sand. The national “AI + Education” Action Plan issued an unambiguous directive: AI can no longer be an elective, a specialization, or a technological footnote. It must be a foundational public course offered by every higher education institution.

This directive makes sense for a country where the digital economy surpasses 8.45 trillion USD and accounts for more than 45 percent of the nation’s GDP. The Chinese State Council has set the ambitious goal of achieving 70 percent AI penetration in key sectors by 2027 and reaching a fully AI-powered economy by 2035.

The education action plan is designed to meet the needs of this increasingly digital economy. It calls for the creation of discipline-specific AI curricula and teaching materials as well as “short, practical, and innovative” frontier courses that keep pace with a fast-moving field. It also targets a wholesale upgrade of talent development programs across traditional disciplines.

For Chinese business schools, the policy carries a pointed message: The era of treating AI as a technical elective—relevant to computer science students but tangential to future managers—is over. Business schools are expected to turn out leaders who can translate complex AI breakthroughs into language that boardrooms understand and who can reverse-engineer business challenges into actionable technology solutions. As a result, many Chinese business schools are integrating science, technology, engineering, and math skills into their programs.

At the School of Economics and Management (SEM) at East China Normal University (ECNU) in Shanghai, we have responded by developing two signature programs. Each is a one-day event that compresses what used to take an entire semester into a single, intensive sprint. Together, the One-Day Case Challenge and the One-Day Boot Camp form the core of the hands-on, AI-first pedagogy that informs the school’s AI-Biz program.

A Case Competition Based on AI

SEM’s first new sprint, the One-Day Case Challenge, operates from a straightforward premise: What if AI is not merely a tool, but a collaborator?

Rather than treating AI as a background resource, the challenge places it front and center across every stage of the analytical workflow. Students must show how they craft prompts, build analytical scaffolding, stress-test their logic, iterate their arguments, and transform raw AI output into something genuinely their own.

In the first iteration of the challenge, held in May 2026, participants analyzed the way kitchen appliance company Fotile managed a lean digital transformation. Participants leveraged AI to examine complex business challenges, evaluate strategic alternatives, and develop actionable solutions. Their analyses spanned multiple dimensions of Fotile’s transformation, including intelligent manufacturing, cross-functional organizational collaboration, and employee-driven innovation.

As teams made their presentations, four advanced techniques stood out as valuable: seeking alternate perspectives, cross-validating across multiple tools, requiring end-to-end traceability, and iteratively refining prompts. Following is a closer look at each one.

1. Seeking Multiple Perspectives

Rather than conducting a single analysis, students instructed AI to examine the case from the distinct viewpoints of five professionals: a strategic consultant, a lean manufacturing expert, a supply chain analyst, a competitor analyst, and a critical reviewer. The students compared these viewpoints to identify common motivations and conflicting interpretations of the same information.

Then, students instructed the AI to conduct a self-refutation exercise in which it challenged its own arguments, identified the weakest links in its reasoning, explained why those weaknesses existed, and revised its conclusions accordingly.

Rather than treating AI as a background resource, the One-Day Case Challenge places it front and center across every stage of the analytical workflow.

Each time they used a prompt, students followed a standardized four-part structure in which they identified the role AI should play, the task in question, the constraints affecting the task, and ways the AI should format its recommendations.

This ensured consistency and analytical rigor. To discourage the AI from returning superficial recommendations, students explicitly prohibited generic suggestions such as “strengthen management” or “raise awareness.” Instead, they required every recommendation to be specific, evidence-based, and actionable.

2. Cross-Validating Across Platforms

Instead of depending on a single form of AI, students adopted a five-stage workflow that leveraged the complementary strengths of multiple tools. Students completed the following actions at the various stages:

Stage 1: Uploaded the original case to DeepSeek, Tabbit, and Monica and fed each one an identical prompt. After each tool independently analyzed the case, students compared the three outputs to identify consistencies, discrepancies, and complementary insights.

Stage 2: Used the synthesized findings as input for iterative discussions with DeepSeek, refining both the underlying arguments and the overall logical structure.

Stage 3: Divided responsibilities according to each tool’s strengths. For instance, DeepSeek expanded and refined the analytical content, while Tabbit generated recommendations for the presentation structure and slide organization.

Stage 4: Imported the presentation framework into YouMind, an AI-based creation studio, and used the Guizang PPT Designer Skill to produce a complete visual preview of the presentation.

Stage 5: Uploaded each preview slide to Kimi, which recreated the PowerPoint presentation by reproducing the preview’s layout, color scheme, visual hierarchy, and content organization.

Teams followed three principles as they completed the entire workflow: Never rely on a single AI-generated output, match each AI tool to the task for which it is best suited, and ensure humans are the final decision-makers.

3. Ensuring End-to-End Data Traceability

Before the one-day challenge began, students established IMA knowledge bases—AI-based workbenches where they could build and test their analytical models. They also imported classic business and management theories in Markdown format.

Once the case materials were released, students’ first instructions to AI were to extract only the data explicitly reported in the original case; cite the precise location of each data point; and avoid any estimation, extrapolation, or unsupported inference. They used the resulting output as a verified inventory of key evidence for subsequent analysis.

Teams accompanied every conclusion with an explicit source attribution, which enabled them to determine where the evidence originated.

During the analytical stage, teams accompanied every conclusion with an explicit source attribution. This enabled them to distinguish whether the supporting evidence originated from the case materials, company annual reports, industry reports, or team estimates.

After the presentation, students conducted a second round of verification to ensure that every figure, chart, and numerical claim on each slide was fully consistent with its documented source, reducing the likelihood of hallucinated data.

4. Iteratively Refining Prompts

To progressively deepen AI-assisted analysis, participants used a four-level prompt architecture:

• In the first level, which focused on case comprehension, students instructed AI to extract key facts and identify the most important information.
• In the second level, which emphasized analytical depth, students required AI to move beyond factual description and explain underlying causes, mechanisms, and implications.
• In the third level, students asked AI to transform analytical findings into presentation-ready visual structures.
• In the fourth level, students prompted AI to articulate the team’s analytical process, key lessons learned, and the broader implications of the case.

Each iteration was guided by explicit objectives for improvement. For example, the first prompts requested outputs that merely described the case; they evolved into prompts that required AI to explain why proposed solutions would be effective. The first analyses simply emphasized Fotile’s unique circumstances; later ones integrated established management theories and identified transferable lessons.

Together, these four strategies ensured that AI acted as a valuable partner in navigating the case challenge—but humans made the final decisions.

A Boot Camp That Brings AI to Innovation

In SEM’s second intensive AI-based event, the One-Day Boot Camp, professionals from different industries compress the entire innovation cycle into a single 10-hour sprint in which they move from a rough idea to a working prototype with genuine commercial potential. Participants create their models with the aid of AI development frameworks such as OpenClaw and Claude Code.

During a sprint, participants might be asked to combine AI with one of five focus areas that represent specific industry pain points: financial compliance, family healthcare, youth education, content creation, and cultural tourism and the consumer experience. Each track culminates in a functional prototype—not a slide deck, not a pitch, but a real, usable build.

During the One-Day Boot Camp, participants compress the entire innovation cycle into a single 10-hour sprint, moving from a rough idea to a working prototype.

In a recent One-Day Sprint event, one team addressed a very specific consumer pain point: the inconvenience of ordering at a fast-food restaurant during the lunchtime rush. Between dine-in orders, self-service kiosks, mobile pickup, and delivery apps, customers often face the frustration of slower fulfillment, long lines, and late deliveries.

This team tackled the chaos by building a digital twin of restaurant operations. The live, data-driven mirror of the store is designed to pull together information about incoming orders across every channel. It factors in workstation capacity, staffing levels, and the complexity of the food that needs to be prepared. It also tracks each store’s location and the capacity of nearby outlets.

This real-time model spots kitchen bottlenecks and order-pressure spikes before they become customer-facing problems. It then feeds store managers AI-driven recommendations for reallocating employees to the areas of greatest need, prioritizing orders, and managing front-of-house traffic to keep queues moving. It even adjusts menu recommendations to nudge customers toward items the kitchen can deliver quickly. As managers shift from reactive firefighting to proactive forecasting, the lunch rush turns from a crisis into a routine.

A second team in the One-Day Boot Camp focused on the “blind box” trend, in which customers buy boxes of goods from their favorite suppliers without knowing what’s inside. As part of this experience, customers film themselves opening the packages and reacting to the contents, then post their videos online.

This team created an app to serve Shanghai’s vibrant scene of pop-up blind box stores. These outlets thrive on spontaneity and buzz—but that’s also their weakness. They depend heavily on foot traffic and unpredictable social media momentum, and they don’t offer much in-store guidance. Meanwhile, they make little effort to ensure that customers film and post the emotional moments of unboxing.

To address these drawbacks, the team designed the AI Blind Box Companion Assistant, a mini-program that provides gamified, companion-style, emotionally intelligent interactions that turn a passive retail visit into a fun, guided experience. The program encourages customers to engage naturally with the products, share their reactions proactively, and return for future releases.

In doing so, the assistant helps pop-up stores build complete marketing funnels, guiding customers seamlessly from initial acquisition through conversion, social sharing, and repeat engagement. In this way, the AI assistant turns one-off foot traffic into a self-sustaining engagement loop.

What Travels, What Doesn’t

After piloting our One-Day Case Challenge and One-Day Boot Camp, we have distilled two key insights:

Not every aspect of the sprint model will translate to every business school context. The format requires reliable access to AI tools, a faculty team comfortable with rapid prototyping pedagogies, and institutional willingness to experiment with scheduling formats that break the semester mold.

Even so, the underlying principles remain broadly applicable. Schools can create intensive experiences if they compress the feedback loop between learning and doing, make AI a partner rather than a shortcut, assess capability rather than content coverage, and design learning experiences that mirror the actual conditions under which graduates will work.

A key question facing business education at this moment is: How do we make business education as fast, integrative, and relentlessly practical as the world our students are entering? Our school has responded by designing AI-powered One-Day Sprints. But it’s a question that every business school soon will need to answer in its own way.