The impact of AI retail training on 100,000 associates
As a boutique custom learning company, Cinecraft Productions has always been committed to designing high-quality e-learning solutions that are authentic, timely, accessible, relevant, engaging, fun and efficient, and in line with seven better learning principles. As global retailers with around 6,000 stores and 100,000 store associates approached us to help modernize retail training, there was an exciting opportunity to leverage artificial intelligence (AI) to meet their needs.
Learning Strategy
Retailers have adopted the same sales process for 15 years. The sales process was effective, but unfortunately it was underutilized due to too many confusion procedures. So the retailer condensed the sales process into three simple steps. Ultimately, this change will increase the value of the average shopping basket.
To achieve our goal, we recommended a blending approach involving AI-generated coaches that provide instant, authentic feedback, including behavioral modeling videos, video-based simulations, and review scenarios.
Since the new sales process aims to guide associates rather than providing scripts, it is recommended to use a dynamic AI-driven approach for review simulations. Instead of selecting the multiple selection option, associates write their own responses to customers. Custom Language Learning Models (LLMs) trained in the sales process enhance feedback in these simulations. Just like a real training coach, training LLM (also known as a model) allows associates to provide specific feedback based on what they enter in their answers. This approach builds the confidence of your associates and gives you personalized feedback.
Retail Training with AI: How did you do that?
As we developed an AI coach for Refresher simulations, there were many factors to consider. In addition to our standard processes, we provide the steps below to create an effective and safe solution.
Step 1: Determine your client’s needs
This required a thorough analysis of existing IT infrastructure, along with legal and security requirements.
The client did not have an existing AI platform, but wanted to host a new AI solution within the existing infrastructure. This requirement requires a robust and adaptable platform that can seamlessly integrate into the current ecosystem, while maintaining complete client control over the environment.
To meet these requirements, the AI platform and all associated data had to be securely sandboxed, ensuring that clients maintained ownership and governance of their data and workflows. Additionally, we proposed using intermediate servers to ensure the safety of data processing and minimizing risk. This ensures that learner responses and AI feedback remain safe and private.
Step 2: Decide on the best technology
The next step was to select the right technology to integrate AI into retailer sales process training. As with all effective learning solutions, accuracy and responsiveness are important. The AI model had to provide relevant and immediate feedback to its affiliated parties to support an attractive and dynamic training environment.
To ensure quality standards, we tested multiple AI models to determine which ones provided the most accurate results and the fastest response times. This rigorous evaluation process allowed us to select the model that best suits our client’s needs for efficiency and accuracy.
Effective AI integration is expensive, but is influenced by two key factors: the amount of data entry into the model and the number of queries or users accessing the service. To navigate through these variables to find the most efficient solution, we created a detailed cost matrix. This matrix evaluated various configurations and usage scenarios to determine the optimal balance between performance and cost-effectiveness for a client’s specific use case.
The solutions we chose ensured affordable prices without compromising quality and provided scalable solutions tailored to clients’ budget requirements and operational goals.
Step 3: Determine your technical workflow
Hardware retailers wanted to build courses using Articulate Storyline 360. This required learners to understand secure ways to interact with AI through the storyline interface. After extensive research and discussion, the following workflow was implemented:
Enter answers in the storyline – Learners will enter the responses into the storyline course by watching a video where customers enter the store or ask questions. Intermediate Server Processing – Learner responses are sent securely to servers owned and controlled by clients for pre-processing. AI Platform Processing – Intermediate servers send insensitive data to the AI platform and generate contextually relevant feedback to learner responses. Sensitive information is stored on intermediate servers and is not passed on to the AI platform. Intermediate Server Processing – AI feedback is sent to intermediate servers where it is refined and formatted to be delivered to the storyline. Feedback Delivery to Storyline – Learners receive immediate and practical feedback from AI coaches directly within the Storyline Training module.
This behind-the-scenes process occurs every time a learner answers a question and only takes a few seconds to complete!
Step 4: Train the model
An AI model was needed to serve as the ideal in-store performance coach for hardware retailer associates. This means that they had to teach them everything about the new sales process for their clients, as well as other expected actions for their associates, systems and resources they might use in their jobs. This was a meticulous process. Instead of developing a custom model, we used a base model from the AI platform. A detailed educational context was provided to meet the specific goals of the retailer. This includes training models to recognize industry-specific terms, common customer scenarios, and retailer policies and procedures. This content is outlined in a scenario grid and full storyboard similar to the process of regular training courses.
Step 5: Test the model
After providing all this information to the model, we had to make sure it was trained effectively. If the AI coach provided an answer that matches the client’s goals, we were successful! Otherwise, we had to reorganize the model with different information. The testing process was first carried out by users who were familiar with training content but were not associates at the store. After improving the model, the simulation was launched as a pilot of a selected number of store associates to try. They provided ideas about the ease of use, relevance and feedback accuracy of the training.
Step 6: Improve your model based on feedback
During the test phase, areas of improvement were revealed. For example, the model response had to be improved to ensure consistent tone and accuracy in the retailer’s communication style. After multiple iterations and adjustments, the desired performance and learner satisfaction levels were achieved.
Conclusion
The integration of AI into retail associate training has proven this global hardware retailer’s transformation. By leveraging cutting-edge technology along with sound instructional design principles, we have created scalable solutions that increase relevant reliability, improve customer service and provide measurable business outcomes. For learning and development experts exploring AI, this case study highlights the importance of thoughtful implementation and the commitment to quality e-learning principles.
Cinecraft Productions
Cinecraft is a boutique content development agency that collaborates with some of the world’s most renowned brands to improve employee performance. Better Learning – Better results.
