Is Artificial Intelligence Good for Pricing?
The Jury is Still Out
Just about every other pricing article you see these days talks about the impact of AI.
AI blew up SaaS subscription pricing in 2025 and no doubt this will continue well into this year and beyond. Blogs such as Good Better Best and Growth Unhinged track the industry’s latest price page changes weekly. It’s enough to make one dizzy.
This leads us to consider two related questions:
What about other (non SaaS) industries? When and how will AI change pricing strategies for companies in the non-tech space such as industrials, financial services and healthcare, to name a few?
Will the outcomes of all this be GOOD or BAD for business? Who will ultimately benefit more: buyers or sellers?
Already pricing professionals across many industries are using AI to enhance their advanced analytics, enable real-time responses to competitor and market moves, and do better predictive revenue modeling. No surprise here, similar shifts are occurring in other professions.
AI tools are shifting from general-purpose Large Language Models (LLMs) to industry-specific “vertical AI” and task-oriented work agents “agentic workflows.” With respect to B2B pricing, there are two emerging innovations that are catching my attention: AI-driven value modeling and purchasing agents.
AI-Driven Value Models
Most familiar to me are AI-driven value modeling tools. In the past year I test drove four different ones, not to mention building my own prototype using Lovable. In this respect I am a uniquely qualified reviewer. For this reason: I have probably built more value models, either in spreadsheet format or by using customer value management platforms (namely LeveragePoint and Ibbaka) than anyone else. I’ve probably done close to a thousand of them for products across all industries, including non-profit organizations.
Let me state clearly that as an independent value coach, I am totally neutral about these new AI value modeling tools. So, I’ll make no endorsements, nor will I mention any vendors by name at this time. My evaluation process is still ongoing. Later, when I have spent more flight hours on them I will write a detailed review.
However, my quick collective assessment of them are:
They all are a tremendous improvement compared to earlier generation tools. By earlier generation, I mean tools where you have to design, build and find customer data to quantify value. All of these AI tools are capable of cranking out a decent value model for a specific customer and product very quickly. What I like best is that they will pull in and document data sources (including proxy estimates) for the value quantification. Overall, a huge time saver.
It’s still early days. In my objective opinion, none of these AI tools are fully ready to be used by novice users because the quality of the output is spotty. 70-80% of it tends to be solid, 10-20% is impressive (meaning it suggests unique value drivers I overlooked) and 10-20% is unusable (either flawed or impractical measures of value). Although this is adequate for an advanced value modeler, the bigger deal breaker for me is the limited editing and presentation functionality.
They will revolutionize value selling. As we learned when I worked at LeveragePoint, the relevant job-to-be-done here is NOT building value models. The primary and most frequent use case for value models is closing deals. Once AI-driven value modelling tools evolve to include coaching sales users through the entire selling process, including listening in on all client interactions and providing guidance around telling an effective value story and even price negotiation; then at that point it becomes a real-time coach (thereby making my job obsolete).
AI Buying Agents
Much of what I know about this comes from my colleague Steven Forth, founder of ValueIQ.ai. Steven is regarded as one of the most forward thinking experts about pricing innovation. He is “all-in” on AI’s impact on sales and purchasing workflows, arguing that companies must fundamentally change how they present pricing information to accommodate autonomous buyer agents.
Steven predicts that companies will need to create pricing and value content primarily designed for AI consumption versus humans. AI will fundamentally transform the RFP (Request for Proposal) and vendor selection process. Further, traditional packaging such as Good-Better-Best will dissolve and be replaced by real-time pricing proposals based on automated value assessments.
Steven foresees a future where significant portions of the purchase and sales processes are automated via agentic AI with minimal human intervention.
Happily Ever After or Paper Clip Napoleon?
So where will AI-led pricing eventually lead us? Let me offer up two exaggerated and radically different future scenarios.
Happily Ever After. A golden age for value suddenly becomes possible because much of what was considered unproven or unmeasurable about customer value creation becomes fully transparent and objective. Hardball negotiation and gamesmanship become relics of the past. Both buyers and sellers are assured of getting a fair deal and the overall purchasing process becomes more efficient. Revenue and costs become more predictable. Customer value continues to grow because sellers have relevant data they need to improve their product/service performance.
Paper Clip Napoleon. This darker scenario is inspired by the 2024 book Nexus by Yuval Noah Harari. Harari has a pessimistic perspective about AI, framing it as a unique threat to human civilization. He prefers calling AI “alien intelligence” because it possesses intellectual powers that surpass humans in ways we may not fully comprehend.
To illustrate the risk of autonomous business agents, Harari cites a thought experiment, i.e., “Paper Clip Napoleon” put forth by philosopher Nick Bostrom. I quote it here:
…imagine that a paper-clip factory buys a superintelligent computer and that the factory’s human manager gives the computer a seemingly simple task: produce as many paper clips as possible. In pursuit of this goal, the paper-clip computer conquers the whole of planet Earth, kills all the humans, sends expeditions to take over additional planets, and uses the enormous resources it acquires to fill the entire galaxy with paper-clip factories. (Nexus, pages 271-273)
Spy vs. Spy is Perhaps More Likely
While the chances of a Clippy-esque galactic dictator are quite remote, a more likely scenario is akin to the classic “Spy vs. Spy” comic strip from the late, great Mad magazine. In this comic strip, two implacable secret agent foes constantly battle each other via outlandish schemes and booby traps.
Isn’t it ironic that AI applications are called “agents”?
Jokes aside, it is the zero-sum tone of this potential interaction that worries me. A true value philosophy is win-win. This may undermine the basic trust assumption implicit and essential in all free market transactions.
What is to prevent a seller agent from deceiving a buyer? Or a buyer agent from detonating a WMD (weapon of mass discounts)?
It could get messy. And as Harari cautions in his book, we can never be certain how these agents really think or what crazy schemes they will think of next.
The more useful distinction is whether the value AI creates remains governable as systems scale, drift, and fail. Boards get into trouble when optimization outpaces accountability and advantage quietly converts into exposure.
Thank you for the shoutout. I learn a lot about value modelling from your work. I look forward to your reviews of the emerging value modelling agents.
It would be interesting to construct a classic scenario planning view of the future impact of AI on pricing. What are the two key critical uncertainties that we would use to structure this?
But the most important, how do we make the adoption of AI in pricing a positive sum game for buyers, sellers and society as a whole.
One thing that generated value models do is make what was once scare plentiful. Value models used to be scarce and expensive. No one created all of the value models that one really needed (For example, how many companies have value models relative to each of the competitive alternatives?). Value models will be abundant, easily configured and dynamic (updating in near real time). What will that change?