From Data to Decisions: 4 Tiers of Practical AI 

A question we often hear from our clients and partners in the automotive industry is, “How can my business recognize value from AI?”

One useful framework for understanding the value that AI can deliver consists of four tiers of outcomes: 

• Insights – what happened?  

• Predictions – what’s going to happen?  

• Recommendations – what should I do?   

• Automation – do it for me!   

The common thread across all these tiers and outcomes? Quality data is crucial to AI success. AI is only as good as the data that it’s trained on, which is why we, at Cox Automotive, spend so much time and effort making sure our data is high-quality, secure, and usable. 

Let’s apply this 4-tiered framework to some automotive examples.  

Consumer Marketing: Personalization  

• Data Source: Online consumer shopping activity from Autotrader and Kelley Blue Book. 

• Insight: John has been actively browsing SUV listings, indicating a strong interest in this vehicle category. 

• Prediction: Based on behavioral data, there is a 75% probability that John will purchase a new SUV within the next 90 days. 

• Recommendation: Target John with SUV-specific marketing campaigns, using channels and timing optimized for engagement. 

• Automation: Dealer software automatically delivers personalized SUV ads to John via social media, aligned with his preferences and optimal engagement windows. 

Logistics: Adaptability 

• Data Source: Historical vehicle shipment records, carrier performance metrics, and market trends from Ready Logistics and Central Dispatch. 

• Insight: Seasonal demand and market dynamics are driving fluctuations in shipping costs. 

• Prediction: The algorithm forecasts an upcoming increase in transportation costs across specific origin-destination pairs. 

• Recommendation: Lane- and load-specific transportation price range recommendations — so carriers can offer quotes that balance competitiveness with profitability, and shippers can optimize between cost and delivery time. 

• Automation: Future state possibilities exist to match carriers with the right loads based on lane preference, capacity availability, and equipment, to enhance speed, efficiency and profitability for both shippers and carriers.

Dealer Sales: Efficiency 

• Data Source: Incoming leads to VinSolutions CRM, market insights from vAuto, consumer shopping insights from Autotrader and Kelley Blue Book. 

• Insight: Mary submitted a new car sales lead along with a trade-in that matches the dealership’s inventory needs. 

• Prediction: This lead is high-priority since it is likely to convert to a sale, and the trade-in has a high probability of selling within two weeks at a high price-to-market. 

• Recommendation: AI composes a draft email or text response expressing interest in the trade and offering a test drive appointment for the new vehicle, but the dealership edits the communication before sending. 

• Automation: Future state possibilities exist for AI to auto-respond to the lead with a similar personalized message – even after hours – and auto-schedule the test-drive appointment based on Mary’s availability. 

As you move from insights to automation, the value AI provides becomes greater. At the same time, each tier requires a higher level of testing and trust in the algorithm than the tiers before it. The more complete, timely, accurate, connected, and trustworthy the input data, the more effective AI solutions can be. 

Ben Flusberg, Chief Data Officer
Dr. Ben Flusberg is the Chief Data Officer at Cox Automotive, where he oversees companywide data strategy and governance, AI and machine learning innovation, and data products. He has worked at Cox Automotive for nearly a decade, where he held roles across product management and digital operations before becoming the Chief Data Officer in 2020.