Scaling Business Analytics Projects with AI-Assisted Development

Teaching Students to Build Shiny Apps in R Using Claude AI

---

🎯 Goal: Showcase how Claude AI can scale traditional business analytics projects — and how faculty can bring this approach into their own courses.

---

1. Session Overview

This article accompanies the presentation “Scaling Business Analytics Projects with AI-Assisted Development” delivered at the CSU AI Faculty Workshop (2026), Session 2.1: Teaching with AI in the Business Curriculum.

Three key ideas covered:

1. Why Shiny + AI is a powerful combination for business analytics education
2. A walkthrough of an AI-assisted development workflow in R
3. How this approach prepares students for skills that did not previously exist

---

2. The Problem We Are Solving

Most students can analyze data. Fewer can communicate it interactively.

“AI offers brand new solutions to cross-disciplinary areas that may lead to new ways of thinking and discoveries.”

The opportunity in front of us:

• Move students beyond disciplinary boundaries
• Recognize AI as a new engine for growth
• Prepare students for skills that did not exist before

Traditional analytics courses produce students who can run a regression or build a chart. But stakeholders — managers, clients, policymakers — need to interact with data, not receive a static PDF. Shiny bridges that gap.

---

3. Five Learning Objectives

Students who complete this project will be able to:

#	Objective
1	Inform or improve business operations and systems thinking
2	Cross disciplinary boundaries with data-driven tools
3	Recognize AI as an engine for opportunity and growth
4	Build new skills that reflect today’s workforce needs
5	Solve old problems in new AI-assisted ways

These objectives connect directly to the broader CSU AI initiative: preparing graduates who are not just AI consumers, but AI-empowered professionals.

---

4. Why Shiny?

Shiny turns R code into interactive web applications — no JavaScript required.

Traditional Analysis	Shiny App
Static chart in a report	Interactive, filterable chart
Emailed to one person	Shareable with anyone via a URL
Re-run manually each time	Updates instantly with user input
Code hidden from stakeholders	Results front and center

Shiny is a bridge between analytics and communication. It makes students’ work usable by non-technical audiences — a skill gap that employers consistently identify.

---

5. The Shiny App Model

A Shiny app has two parts that talk to each other:

┌─────────────────────┐        ┌─────────────────────┐
│         UI          │ ────>  │       Server         │
│  What the user sees │        │  Logic that reacts   │
│  Inputs + Layout    │ <────  │  to user input       │
└─────────────────────┘        └─────────────────────┘

Four core benefits for students:

1. Autonomy — Apps run independently in the cloud
2. Interactivity — Data becomes something users explore, not just read
3. Reactivity — When an input changes, outputs update automatically
4. Accessibility — Everyone can navigate and act on the dashboard

The key conceptual leap is reactivity: think of inputs as live variables, and outputs as expressions that “listen” to them. Once students grasp this, everything else follows.

---

6. UI Example — Sales Dashboard

The UI defines layout and controls, completely separate from business logic.

library(shiny)

ui <- fluidPage(
  titlePanel("Sales Dashboard"),
  sidebarLayout(
    sidebarPanel(
      selectInput("region", "Select Region:",
                  choices = c("North", "South", "East", "West")),
      sliderInput("year", "Year Range:",
                  min = 2018, max = 2024,
                  value = c(2020, 2024))
    ),
    mainPanel(
      plotOutput("salesPlot"),
      tableOutput("summaryTable")
    )
  )
)

Notice that students define the interface entirely separately from the logic. This separation encourages clear thinking about what the user needs versus how the data is processed.

---

7. Server Example — Reactive Logic

server <- function(input, output, session) {

  filtered_data <- reactive({
    sales_data |>
      filter(region == input$region,
             year >= input$year[1],
             year <= input$year[2])
  })

  output$salesPlot <- renderPlot({
    ggplot(filtered_data(), aes(x = year, y = revenue)) +
      geom_line(color = "steelblue", linewidth = 1.2) +
      labs(title = paste("Revenue —", input$region),
           x = "Year", y = "Revenue ($)")
  })

  output$summaryTable <- renderTable({
    filtered_data() |>
      summarise(Total = sum(revenue), Avg = mean(revenue))
  })
}

The reactive() expression is the most important concept here. filtered_data() is computed once and shared by both the plot and the table — efficient and clean. This is where AI shines: Claude can write the first draft of server logic in seconds, letting students focus on whether the logic makes business sense.

---

8. AI-Assisted Student Workflow

How students actually use Claude AI in this project:

[Student describes a business question in plain English]
         ↓
[Claude drafts the initial UI and server code]
         ↓
[Student runs the app, tests it, finds issues]
         ↓
[Student pastes errors into Claude for explanation]
         ↓
[Claude suggests a fix + explains why]
         ↓
[Student refines based on user feedback]
         ↓
[Repeat until the app answers the business question]

This workflow builds real professional skills: problem decomposition, debugging, iteration, and communication. AI accelerates the cycle — but students stay in the driver’s seat. They must validate outputs, check logic, and ensure the app remains meaningful.

---

9. What Students Ask Claude

Examples of effective AI prompts from this project:

✏️ “Write a Shiny app that lets users filter a sales dataset by region and displays a bar chart.”

🔍 “My renderPlot is not responding to input$region — what is wrong?”

➕ “How do I add a downloadButton so users can export the filtered table as a CSV?”

♻️ “Rewrite this server function to be more efficient using reactive expressions.”

The skill is knowing what to ask — not just accepting what AI returns.

Prompt engineering is itself a teachable and transferable skill. Students who write clearer, more specific prompts get better code — and this directly mirrors the professional communication skills we already teach in business courses.

---

10. Live Demo — Build an App in Minutes

Starting prompt given to Claude AI:

“I have a dataset called mtcars. Build a Shiny app with a dropdown to select a numeric variable and a slider for the number of bins. Show an interactive histogram and a summary statistics table below it.”

# Claude writes this — student runs, understands, and improves it
shinyApp(ui = ui, server = server)   # functional app in ~30 seconds

The result is a fully working analytics app, ready for students to customize. Claude handles the scaffolding; students handle the meaning and the business logic.

---

11. Bonus — Interactive R Markdown

Students can also embed Shiny widgets directly inside an R Markdown document, combining narrative, code, and a live app in one shareable file:

---
title: "Interactive Business Report"
output: html_document
runtime: shiny
---

selectInput("variable", "Choose a variable:", choices = names(mtcars))

renderPlot({
  hist(mtcars[[input$variable]],
       col = "steelblue", border = "white",
       main = paste("Distribution of", input$variable))
})

This format is especially powerful for capstone projects and client-facing reports — students can document their methodology alongside a live, explorable result.

---

12. Classroom Project Assignment

Deliverable: A Shiny app that answers a real business question.

Requirement	Purpose
One business dataset	Grounds the app in real context
At least two user inputs	Demonstrates reactivity
One chart + one table	Covers visual and tabular communication
Short AI reflection (~1 page)	Builds metacognitive awareness
3-minute class presentation	Practices explaining to non-technical audiences

Graded on: clarity of the business question, usability of the interface, and quality of the reflection.

The reflection is the most important differentiator — it forces students to think about how they used AI, not just that they used it. This is where the deepest learning happens.

---

13. Student Outcomes Observed

• Students who had never built a web app shipped a working dashboard in one week
• AI reduced time spent on syntax errors, freeing time for business interpretation
• Students reported higher confidence in presenting technical work to non-technical stakeholders
• Several students extended their apps into independent research projects

These outcomes map directly to the five learning objectives above — AI does not replace the learning; it removes the friction that previously blocked it.

---

14. Key Takeaways

1. Shiny + AI makes interactive application development accessible to all business students
2. Claude AI handles scaffolding so students can focus on business logic and design
3. Prompt engineering is itself a transferable skill worth teaching explicitly
4. Small projects build real confidence — and sometimes turn into real tools
5. The goal is not code — it is students who can move from insight to impact

---

Resources & Contact

Resource	Link
Claude AI	claude.ai
Shiny documentation	shiny.posit.co
R Markdown guide	rmarkdown.rstudio.com
RPubs	rpubs.com

---

“Jimmy” Zhenning Xu
Department of Management & Marketing, CSU Bakersfield
🔗 github.com/utjimmyx
📧 zxu3 at csub.edu

Presented at the CSU AI Faculty Workshop, Session 2.1: Teaching with AI in the Business Curriculum. Chair: Neil Shahrasbi, Department of Information Systems, SFSU.