AI: Use Case Studies

Cloud First Company Mission:

Adopting Machine learning as a paradigm-shifting technology.

Context:

Artificial intelligence is designed to optimize the company’s activities, open previously unreachable horizons, reduce costs, create a competitive advantage and allow people to engage in innovative activities, rather than in a daily routine.

Principles

Machine Learning represents the shift from machines representing the world through predetermined rules, to the use of constantly evolving probabilistic models.

Machine Learning can be thought of as a field of computation that uses algorithms to derive probabilistic models of complex events by training these models with a set of observations (data).

AI Use Case.

Please see some case studies below we have gathered below.

Each differ in terms of goals however overarching AI context & principles apply.

Key Clients Business Aims:

  • Reduce costs using AI tooling.
  • Decentralise Reporting
  • Increased BI maturity/ Business Insights/Analytics
  • Better visibility of Issues
  • Process Optimisation
  • Reduce low value/repetitive tasks for employees
  • Trust & Governance is AI adoption

 

Introducing our Use Cases for ML / AI for our AI Client.

Negotiation Engine

Our AI Client deals with a global network of suppliers across both its traditional retail business and it’s own private-label operations. Despite hundreds of suppliers, only about twenty of these suppliers account for ~70% of all purchases. But contract negotiations with other suppliers take up a disproportionate (>70%) of the procurement organization’s time. This is despite the fact that procurement officers try to reduce friction through auction systems and go through relatively routine contracting processes. Recognizing that the same few terms appear over and over again, the Chief Procurement Officer recently proposed using AI to analyze contract red lines and present a proposed negotiation strategy.

 

Recruiting Screener

The HR team’s recruiting function has been under significant pressure as a result of our AI Client’s continued growth and expansion. With each new store, hiring needs to be considerably expanded, and given the high employee turnover of retail, the team struggles to support each location’s hiring needs in addition to central staffing. As a result, the team is preparing a recruiting screening tool that uses the profiles of successful past employees in order to analyze and screen employee applications prior to interviews.

 

New Offering Tool

Our AI Client is constantly looking for ways to “private label” or otherwise produce in-house product offerings. Historically the company has focused only on sales of 3rd party offerings, taking “best-sellers” and looking for opportunities to offer similar products at a reduced cost. As the company’s density of in-house offering has expanded, they’ve begun to consider launching products not yet available in market. Thus far, these efforts have been championed by our AI Client’s in-house creative team. But inspired by an article on generative learning for new product design, the team has proposed an AI-driven approach for creating new offerings by analyzing the characteristics of historic best-sellers and synthesizing them into entirely new offerings.

 

Employee Pay Assessor

Our client has over 10,000 full and part-time staff at any given time. They are spread across one hundred stores, several operations warehouses, as well as central corporate staff. These employees have historically been categorized into roughly 70 different types of roles, ranging from cashier to floor manager to accounts receivable analyst to marketing specialist. Each year, Our AI Client’s undergoes an annual compensation evaluation period to assess pay raises in light of inflation, performance, and market demand. This is separate from any promotions which are determined on a case-by-case basis. The HR team has historically allocated each team manager a flat raise “budget” and allowed discretion by managers in how their pay allowance is allocated among employees. Hoping to add sophistication to this system, the HR team has recently analyzed their employee retention data and hopes to optimize the company’s workforce by using predicted likelihood of retention to help allocate this budget. For example, if an employee is being paid at below market rate and is a high performer, the system may suggest a significant pay increase to try and retain the employee.

 

Capital Project Tool

Our AI Client owns more than one-hundred locations across the United States. One of its competitive differentiations is that it does not use a franchising model. As a result, each new store opened is a capital intensive project that requires purchase of land, building or refitting of store location, new inventory risk, and up-front marketing costs. The cost of new locations is relatively fixed and Our AI Client’s has historically opened new stores only in densely populated areas. As they continue to expand, the company is looking to use a data-driven approach, drawing on inspiration of their e-commerce business’ digital marketing strategy, to look at deeper characteristics of the neighborhoods they’ve operated in historically and assess where they might be successful in the future.

 

Segmentation Tool

Our AI Client’s leadership team has always been clear that the company is a customer-centric business and that knowing who the customer is and what motivates them to purchase is essential. To that end, the marketing and merchandising teams have historically maintained buying “personas” that they used to form a common language around groups of buyers. Curating these personas has always been a manual process, but the company’s new CMO recently proposed the use of spectral clustering, an unsupervised learning technique, to analyze anonymous customer data, match it to publicly available demographic data, and form rich, data-driven customer personas that can update in real time.

 

Digital Channel Chatbot

More and more of Our AI Client’s business has been shifting online as they do their best through same-day in-store pickup and other tools to compete with large online players. Our AI Client’s uses a shared call-center to support customer service for all orders at a significant price. But the marketing team has recently suggested using online messenger channels to deflect some of this call volume. These channels also present the opportunity for the use of chatbots instead of humans for certain routine customer calls.

 

Supply Chain Optimizer

Our client has over 100 stores that are serviced by a system of ten warehouses located in the EU. Each of these facilities receive shipments from vendors, overseas manufacturers, and other warehouses. In addition, they share e-commerce responsibilities and help physical store locations maintain their inventory needs. Because shipping costs are significant, Our AI Client’s tries to optimize the use of their trucks and prevent items from being moved more times than necessary. To date, this is handled via spreadsheet by supply chain coordinators at each warehouse, each of whom have a fleet of trucks at their disposal. The VP of supply chain has suggested centralizing this function in order to reduce waste and minimize cost, using a linear optimization model, while ensuring inventory needs are met.

 

Purchase Propensity Predictor

Predicting demand for new products is among Our AI Client’s biggest challenges. Historically the company has used a simple average of 5-10 comparable product launches over the last five years to gauge overall magnitude and then spread inventory across stores based on proportion of gross sales for the relevant category. Frustrated with stockouts, the VP of EU sales has suggested using some of the more sophisticated data being collected across the company to provide a more robust projection.

 

Visual Classifier for Competitive Analysis

The retail electronics market is highly competitive and Our AI Client’s competes with several local and online brands. One way the company has historically differentiated itself is by offering unique and differentiated products, even in mundane categories, and closely monitoring competitive offerings. But as e-commerce has become the biggest competitor, Our AI Client’s has struggled to analyze competitive offerings because of the sheer breadth and scale of their online competitors. The merchandising and buying team recently proposed using computer vision in order to monitor new online offerings and ensure Our AI Client’s product line remains differentiated.

 

Natural Language Product Classifier

An absolutely critical function for the business is to ensure that with every new product released, it is tagged within the company’s existing taxonomy. This tagging process allows all functions from marketing, supply chain and accounting to communicate in a common language across the vast portfolio. Today, this process is handled by buyers. Whenever a buyer selects a new item, they are required to publish a written description of the item that is approximately 150-300 words long. The buyer then applies appropriate tags based on this description documenting its size, color, material, category of use, etc. The COO of the company recently proposed using natural language understanding and classification in order to pre-tag these written summaries, saving buyers valuable time

 

Machine Learning Insights- Industry Quadrant