Artificial intelligence is changing how businesses make pricing decisions. Companies can now analyze sales patterns, customer behavior, competitor prices, and market demand to determine what people are willing to pay. But having access to more information raises an uncomfortable question: Just because a business can predict how much a customer will spend, should it use that information to charge them more?
For small business owners and startup founders, the challenge is finding ways to use AI to improve profitability without making customers feel they’re being treated unfairly. A pricing strategy that increases revenue today may not be worth much if it drives loyal customers away tomorrow.
We recently explored this issue in our PowerHomeBiz article, How Small Businesses Can Use AI to Set Better Prices Without Losing Customer Trust. We also looked at how major companies are approaching the issue in Walmart vs. McDonald’s: What Their Different Approaches to AI Pricing Teach Small Businesses, published on LearningFromBigBoys.com.
To get a better understanding of what responsible AI pricing looks like in practice, we spoke with James Sun, CEO and Co-Founder of Kapnova, an AI company focused on helping businesses improve revenue and profitability through data-driven decisions.
Sun, an MIT Sloan MBA and three-time venture-backed founder, has worked with consumer brands across the beauty, retail, and other industries. His experience has shown him how easily businesses can mistake rising sales for improved profitability, or assume that a marketing campaign worked simply because revenue increased afterward.
In this written interview, Sun explains the differences between dynamic, localized, and personalized pricing; which types of customer data businesses should avoid using; and how entrepreneurs can test pricing changes without risking customer relationships. He also shares why AI-generated recommendations still need human judgment—and why sometimes the smartest pricing decision is not to change prices at all.
Table of Contents
Question and Answer with James Sun on AI Pricing
1. How do you distinguish between dynamic, personalized, localized and AI-assisted pricing, and why does it matter?
Dynamic pricing changes based on demand or timing, like airline tickets. Localized pricing adjusts for different markets. Personalized pricing means two people might pay different prices based on what a company knows about them.
AI-assisted pricing uses algorithms to analyze data and recommend prices. It doesn’t necessarily mean charging customers different amounts.
The distinction matters because there’s a difference between using AI to understand market demand and using it to extract the maximum amount from an individual.
At Kapnova, we focus on helping companies make economic decisions that increase revenue and profit, not exploiting individual willingness to pay.
2. What data should companies use when setting prices, and what should be off-limits?
Businesses should use historical sales, inventory, product costs, competitor pricing, seasonality and overall demand.
I’d draw the line at using sensitive personal information, such as someone’s health, financial situation or vulnerabilities, to charge them more.
There’s a difference between knowing sunscreen demand increases in summer and knowing a particular customer desperately needs a product and might pay twice as much.
Just because AI can analyze certain information doesn’t mean businesses should use it.
3. Where is the line between understanding market demand and estimating what an individual will pay?
It’s the difference between studying market economics and exploiting individual behavior.
Suppose AI determines that raising a skincare product from $40 to $43 could improve profit without significantly affecting demand. That’s legitimate pricing optimization.
But if AI determines one specific customer would pay $55 based on their browsing history or personal circumstances, that’s different.
I’m not against personalization. Loyalty discounts can benefit customers. But rewarding loyalty isn’t the same as secretly charging someone more.
My rule: If you can’t comfortably explain your pricing logic to customers, reconsider it.
4. Can you share an example where AI suggested a price change that produced measurable results?
Working with hundreds of consumer brands in beauty, I’ve seen companies focus on sales growth without fully understanding what drives profit.
Here’s an illustrative example of the decisions we’re building Kapnova to address.
A beauty product sells for $30 with a $12 product cost. At 10,000 units, that’s $300,000 in revenue and $180,000 in gross profit.
Raise the price to $33, and if volume drops only 5%, gross profit increases to $199,500. That’s an additional $19,500 without increasing marketing spending.
But if volume drops 20%, gross profit falls to $168,000.
Same price increase, completely different outcome.
This is why Kapnova combines AI with causal inference and mathematical optimization. We want to estimate what a decision will actually do to profit, then measure whether it worked.
5. Have you seen situations where maximizing short-term revenue could hurt a brand long-term?
Absolutely, especially in beauty, where loyalty and repeat purchases matter.
A company might raise prices and make more money this quarter, but if customers switch brands or purchase less frequently, it could lose much more over time.
Traditional analytics often looks at metrics in isolation. Revenue increases, but what happened to retention, repeat purchases, acquisition costs and gross profit?
Sometimes the right decision is to sacrifice short-term revenue for long-term customer value.
AI should inform those tradeoffs, but entrepreneurs still need to apply judgment.
6. What information is useful for small ecommerce businesses without large data-science teams?
Start with what you already have: selling prices, product costs, units sold, promotions, inventory and repeat purchases.
A $2 million ecommerce business might discover its best-selling product generates less profit than one selling half as many units. Or that a 20% discount increases orders but reduces total profit.
Those insights don’t require millions of transactions.
AI is making sophisticated analysis accessible to smaller businesses that couldn’t previously afford teams of analysts.
But start with a specific business question, not an AI tool looking for a problem.
7. How should businesses test new prices without risking customer backlash?
Start small. Don’t change prices across your entire catalog because an algorithm recommends it.
Test a few products or markets. Establish a baseline and compare results against similar products or markets where prices stayed unchanged.
Measure gross profit, conversion rates, customer complaints and repeat purchases, not just revenue.
The biggest mistake is assuming what happened after a price change happened because of it.
That’s why causal inference matters. You need to understand the actual impact of your decision, not just compare before-and-after numbers.
8. What should companies disclose about algorithmic pricing?
Companies don’t need to reveal proprietary algorithms, but they should explain why prices differ.
Consumers generally understand seasonal sales, loyalty discounts and regional pricing.
What they wouldn’t appreciate is discovering they paid more because an algorithm analyzed their personal behavior and decided they were willing to pay more.
My test is simple: Can you explain the price difference in plain English without embarrassing your company?
Customer trust is expensive to rebuild.
9. What should business owners understand about the FTC’s attention to personalized pricing?
Business owners should ask three questions: Can AI do it? Is it legal? And is it good for the business?
AI can estimate individual willingness to pay using enormous amounts of consumer data. But maximizing every transaction doesn’t necessarily maximize long-term business value.
There’s a difference between offering relevant discounts and charging customers more based on private information.
I’d focus AI on better pricing, promotions, inventory and marketing decisions. There’s tremendous profit opportunity without invasive personalization.
Businesses should also review their practices against applicable privacy and consumer protection laws.
10. What’s the biggest pricing mistake companies make by confusing correlation with causation?
Assuming that because sales increased after a discount, the discount caused the increase.
A retailer offers 20% off and sales rise 30%. Everyone celebrates. But what if sales would have increased 25% anyway because of seasonality or increased advertising?
The company gave away margin for very little incremental demand.
I experienced a similar problem spending heavily on influencer marketing. We generated activity and engagement, but those metrics didn’t necessarily translate into incremental revenue.
That experience led me to study causal inference at MIT and eventually build Kapnova.
Predictive AI finds patterns. Causal inference helps determine whether an action actually produces a result. That’s a critical distinction when real money is involved.
11. Can AI become too sophisticated and undermine the pricing predictability customers value?
Absolutely. An algorithm optimizes whatever objective you give it.
Tell it to maximize revenue per transaction, and it might recommend constantly changing prices or charging different customers different amounts.
But customers value consistency, simplicity and fairness. Those have economic value too.
That’s why we believe enterprise AI needs more than language models and autonomous agents. It needs economic models, business constraints and human judgment.
Sometimes the best recommendation is to leave prices alone.
12. What three guardrails should small businesses establish before using AI-assisted pricing?
First, optimize for profit, not just revenue. Understand how price changes affect margins, demand and retention.
Second, establish clear data boundaries. Use legitimate market and business data, avoid sensitive personal information and make pricing decisions explainable.
Third, test before scaling. Start small, measure incremental results and keep humans involved in important decisions.
My advice to entrepreneurs is simple: AI shouldn’t just help you make more decisions faster. It should help you make better decisions.
That’s our thesis at Kapnova. AI finds the opportunities. Math determines the answer. And businesses measure whether those decisions actually create value.
