Designing an AI-Assisted Car Buying & Negotiation Experience.

Buying a car is a high-cost decision surrounded by fragmented information, unfamiliar terminology, dealership interactions, and uncertainty about whether a buyer is making the right choice.

Our capstone began with a broad question: If car shopping were designed today around the buyer, from scratch, with AI, what could it look like?

Rather than automating the entire purchase, we explored how AI could help buyers understand information, make decisions, and negotiate while keeping the user involved throughout the process.

DESIGN QUESTION

How might AI reduce uncertainty without taking control away from the buyer?

Our process combined product-space research, research into emerging AI interactions, interviews with car buyers and dealers, prototyping, usability testing, and continuous iteration.

We looked at the experience from both sides of the transaction to understand not only what buyers struggled with, but also what information dealers felt was being lost in existing online car-shopping experiences.

1. Product space research
2. AI research
3. Buyer + dealer interviews
4. Ideation
5. Prototyping
6. User testing
7. Iteration

1. 01

   BUYERS ARE ALREADY DOING A LOT OF RESEARCH

   Car buyers often move between multiple websites, pricing tools, vehicle-history sources, reviews, and personal tracking systems before they feel comfortable making a decision.

   “I basically built my own database to keep track of everything.”

2. 02

   TRUST IS FRAGILE

   Buyers questioned price fairness, dealership intentions, and whether important information was missing or difficult to interpret.

   “I just wanted to feel confident I wasn’t making a mistake.”

3. 03

   FIRST-TIME BUYERS LOOK FOR REASSURANCE

   Less experienced buyers frequently relied on parents, friends, experienced buyers, and online communities to help them judge whether they were making the right choice.

   “[If I were to go through the buying process again], I would try on my own and then bring in my parents.”

4. 04

   NEGOTIATION CREATES ANXIETY

   Negotiation was repeatedly described as stressful, confusing, uncomfortable, and difficult to evaluate.

   “You’re working toward a number, but you don’t actually know if you did well.”

1. TRIGGER / RESEARCH

   User realizes they need a car, defines a budget, and begins browsing.

2. COMPARING OPTIONS · FOCUS

   User compares vehicles, prices, trims, reliability, history, and ownership factors.

3. DEALERSHIP / NEGOTIATION · FOCUS

   User test drives, communicates with dealers, evaluates offers, and negotiates.

4. PURCHASE / PAPERWORK

   User completes pricing, financing, insurance, and final documentation.

WHERE WE FOCUSED

We ultimately focused on two moments with particularly high uncertainty and decision pressure: Choosing the car. Negotiating the deal.

Research initially generated opportunities across the entire car-buying experience.

We explored ideas involving:

• search and initial exploration
• deal finding and comparison
• unique vehicle value
• communication and connection
• negotiation
• paperwork
• broader interface support

The goal at this stage was breadth. We generated many possible ways AI and interface design could reduce effort, increase clarity, and support buyer decision-making.

We then evaluated which ideas most directly supported the project goals and where users were likely to perceive meaningful value.

Some concepts were prioritized immediately, some were saved for possible future development, and others were removed because they did not fit the core opportunity.

THE CONCEPT

This narrowing process led to one product with two connected experiences: Smart Shortlist + Negotiation Co-Pilot.

A decision hub for choosing the right car.

By the time buyers reach the shortlist stage, they are often comparing several plausible cars across different dealerships and trying to keep track of price, reliability, mileage, vehicle history, features, and advice from other people.

Smart Shortlist consolidates those decisions into one workspace while allowing the buyer to determine what matters most.

One of the central research themes was that buyers did not want AI making important decisions for them.

Instead of declaring one vehicle objectively ‘best,’ Smart Shortlist lets the buyer rank factors such as price, reliability, mileage, fuel economy, technology, ownership cost, and vehicle history. The interface then reorganizes the shortlist around those priorities.

AI supports the decision.
The buyer defines the decision criteria.

Research showed that buyers were frequently piecing together information from multiple websites and keeping track of details themselves.

The comparison view brings important differences into a single side-by-side format so users can evaluate price, mileage, reliability, service history, ownership factors, and other vehicle information without rebuilding the comparison themselves.

Vehicle-history information can be valuable without necessarily being easy to interpret.

The prototype separates factual service-history information from AI-generated interpretation. Service History provides the underlying records. Mechanic Insights explains potential risks, upcoming maintenance, possible ownership costs, and questions a buyer may want to investigate.

AI-generated interpretations shown in the prototype were simulated product behavior.

Once a buyer chooses a vehicle, the decision problem changes. Research consistently surfaced three concerns:

• Pricing is unclear
• Buyers feel unequipped to negotiate
• Fear of overpaying

AN IMPORTANT DISTINCTION

Negotiation Co-Pilot was designed as a support layer around the dealer conversation rather than an autonomous negotiating agent.

The interface combines:

• the dealer conversation
• dealer context
• market negotiation information
• suggested messages
• real-time guidance
• multiple active negotiations

The goal was not to remove the buyer from negotiation. Instead, the interface gives users information and suggested actions while allowing them to choose what to say, when to respond, and whether to continue negotiating.

Some interviewees expressed concern that age, gender, or perceived experience could affect how they were treated during negotiation.

The concept therefore keeps identifying buyer information hidden during the initial dealer conversation while communication remains within the platform — a design concept intended to reduce unnecessary identity information during early negotiation.

“Feeling confident that you’re going to be listened to is definitely a big deal.”

Negotiation is difficult when buyers lack context about both the seller and the realistic pricing range.

The concept surfaces dealership information alongside market-based negotiation guidance so the buyer can better understand the conversation they are entering.

• Distance
• Response time
• Rating
• Years in business

• Market reference
• Listing price
• Potential negotiation range
• Target guidance

These numbers are prototype examples, not live market predictions.

Less experienced buyers told us they often did not know what language to use or what information they should push back on.

The prototype provides optional suggested messages that users can select, edit, or ignore. After dealer responses, the interface can also surface a short explanation of what may matter and suggest possible next steps.

The capstone also explored how the concept could create value within the existing CarGurus ecosystem rather than functioning only as a standalone interface exercise.

JumboDrive ultimately became a two-stage concept centered on two moments where buyers had told us they needed the most support.

Smart Shortlist helps users decide which car best fits their priorities. Negotiation Co-Pilot helps them understand and navigate the dealer conversation.

Across both experiences, the product philosophy remained the same: use AI to reduce cognitive burden and explain complexity without removing the buyer from the decision.

This project strengthened my interest in the space between user research and product development.

The most valuable design decisions did not come from asking where AI could be added. They came from understanding where users felt overwhelmed, uncertain, or unsupported and then deciding whether AI could meaningfully reduce that burden.

It also reinforced the importance of preserving user control. In a high-cost decision like buying a car, automation can be useful, but users still want to understand what is happening and remain involved in important choices.

If I continued the project, I would expand dealer-side testing, further validate the negotiation experience, and explore how the prototype could integrate with real listing, pricing, and dealership data.

Interactive product concept. AI behaviors and marketplace data shown in the prototype were simulated for design and usability testing.