Key Takeaways
- AI transformation strategy fails when pricing knowledge stays siloed.
- Pricing analytics uncovers margin leakage hidden beneath summary reports.
- AI learns from your data, not your intentions.
- Better pricing capability beats better technology every time.
Most businesses don’t lose margin because AI gets pricing wrong. They lose it because they never taught AI what right looked like. That’s why an effective AI transformation strategy depends as much on pricing capability as technology. McKinsey’s most recent research on AI in B2B pricing found that more than half of businesses trying to get value from it cite the same two barriers: data quality and integration complexity. Yet without strong pricing analytics, those barriers often mask the deeper issues. I’ve heard those same two barriers named in real engagements, and they were never the real story.
The real story was a siloed culture, a belief that pricing could simply be lifted and shifted from an existing model, and a head office that had bought the platform without ever properly diagnosing the market it was rolling out into. None of that is the AI platform or the algorithm.
Read This CEO Pricing Strategy To Improve Margin & EBIT
It’s easy to assume a rollout like this is just straightforward IT integration, until you see how complex your own pricing arrangements actually are. Get the system or the model wrong, and the margin loss isn’t gradual. It’s immediate.
Let me tell you what that actually looks like, from inside a business in the eighteen months after it buys an AI pricing tool. Not what the vendor promised. What happens through implementation, the pilot, go-live, and every month after, every time the model makes a recommendation and someone has to decide whether to trust it.
I’ve been involved in more than 23 large-scale business transformations. I’ve learned pricing never fails the way anyone expects.
Let me show you what that looked like inside one of them.
The AI Transformation Strategy That Looked Ready
One business I worked with was replacing a long-established pricing system with a modern AI-enabled platform. On paper, the capability looked mature: documented policies, defined workflows, years of pricing history. Leadership’s logic was reasonable: we’ve done the hard work, now we just need better technology.
That assumption didn’t survive a closer look. Once we lifted the hood of their system, connected the underlying datasets, and ran detailed, transaction-level analysis nobody had done before, a different picture emerged: real system complexity, hidden price complexity, and margin leakage invisible at the level the business usually looked at.
Here’s why I no longer believe the fix is a training programme or conventional project management.
The knowledge wasn’t missing. It was there all along. It just wasn’t connected.
Why AI Transformation Strategy Breaks Down Across Teams
Pricing understood the exception and override rules. Sales understood the customer agreements. Finance understood the rebates. Master data understood the product structures. Every function had real expertise in its own piece, but nobody had the whole system in their head.
Organisational psychologists call this a transactive memory system: a shared “who knows what” map so no single member has to hold the whole picture. First described by Daniel Wegner in the 1980s, it explains how a group’s memory can be more complete than any individual’s.
An AI model doesn’t know who to ask. It only knows what it was trained on: one function’s slice, stripped of the context that made it true.
The knowledge wasn’t missing. It was there all along. It just wasn’t connected.
What Pricing Analytics Actually Revealed
Before the transformation, the business’s pricing reports sat at a comfortable altitude: average margin, overall gross profit, category performance. Those numbers said pricing was healthy.
Those numbers were right. But right isn’t the same as complete.
Once we looked past the summary: price dispersion between customers who should have been priced alike, override activity clustering around the same few points, cost anomalies the summary had no way of catching.
Feed a pricing model data that’s already been smoothed and summarised before it reaches the algorithm, and it learns exactly what that summary says: pricing is healthy. It’s confidently, quietly wrong, not because the model is faulty, but because nobody built it to see the variance underneath.
A model trained on smoothed, summarised data learns that pricing is healthy. It’s wrong, not because the model is faulty, but because nobody built it to see the variance hiding underneath.
The three questions that actually protect margin
“Critical thinking” for a pricing team isn’t a personality trait, and it isn’t a checklist either. It’s the ability to problem-solve, to connect dots nobody handed you already connected, before you trust what an AI pricing model is telling you.
It means knowing that invoice price, net price and realised margin are three numbers that can each look healthy while the others deteriorate. Knowing that and building it into a platform are two different things: net price and realised margin often never make it into the system the way invoice price does, because of the complexity involved and the state of the underlying data. So the system optimises the only number it can see. That’s the gap between knowing the right answer and building it into the tool.
It’s not hypothetical: in 2016, OpenAI trained a reinforcement-learning agent to play CoastRunners. Rather than finish the course, it found it could score higher looping in a lagoon, crashing into respawning targets while its boat burned, outscoring humans by 20 per cent, because nobody had told it the goal was to win, only to maximise score. A pricing model behaves the same way.
It also means understanding which cost basis is actually driving a pricing decision. Most businesses price against more than one version of cost, each producing a different “optimal” price. Understanding those measures is one thing; demonstrating which one is driving a given price is another.
That uncertainty is structural, not a knowledge gap. Cost data updates on its own timeline, different parts of a business reference it for different reasons, and by the time it has flowed through the rules and overrides, tracing a price back to its source is close to impossible.
And it means knowing where in the price waterfall, list price down to pocket margin, value is actually leaking, instead of accepting a single blended discount number as the whole story.
None of that is a soft skill. It’s technical fluency, applied to a system that will otherwise sound confident regardless of whether it’s right.
Why AI Transformation Strategy Still Fails
Korn Ferry’s own research says judgement matters more than platform fluency as AI tools mature, and that’s the part they get right. Where it falls short is treating judgement as a portable soft skill that transfers cleanly across functions. What helped me catch the margin-definition gap and the reporting gap wasn’t being a strong general thinker. It was pricing depth, combined with the advantage of not running the system every day.
That’s the honest limit of a generic workforce consultancy: they can tell a business to hire for critical thinking, not what to catch inside pricing.
McKinsey has published good pricing research for forty years, and implementation failure is still as common today.
Your own pricing team usually can’t close this gap alone, and that’s not a mark against them. Nobody standing inside a system every day sees it as clearly as someone standing outside it.
Diane Vaughan found something similar studying the Challenger disaster: the longer people operate inside how something normally runs, the harder it becomes to register a small deviation as a warning sign. That happens to good teams, not just struggling ones.
Bandwidth explains part of the gap. The rest is calibration. Sixteen years of Taylor Wells’ research into the pricing and commercial community shows a consistent pattern: managers tend to rate themselves highest in strategy and stakeholder alignment, lowest in price architecture and systems; the very skills most likely to protect margin once a price is live.
We also find that women who score well on our pricing and commercial tests tend to underrate themselves; men tend to overrate.
Tenure is its own trap, too: the skills that mattered for managing price lists aren’t the ones needed to catch an AI model quietly optimising the wrong thing, or to set the platform up with the right architecture in the first place. Confidence, gender, and years in the role are all weak proxies for what actually protects margin.
〉〉〉 Get Your FREE Pricing Audit 〉〉〉
What AI Transformation Strategy Really Requires
If a pricing model confidently does exactly what it’s told, whether or not that’s the right thing, the next question is who has the standing to catch it before it costs margin.
Training doesn’t answer that question, because the gap was never what people knew. It’s structural: the space between knowing the right answer and having a system built to act on it.
Conventional project management doesn’t answer it either: it delivers a defined scope on a defined timeline, and this is an ongoing question, not a project with an end date.
That’s not a data question. It’s a people question, and it’s the one most AI-readiness programmes skip.
A model like that is never broken. Nobody has taught it what wrong looks like. A generic framework was never going to build that depth.
If you don’t know what wrong looks like inside your own pricing model yet, that’s worth finding out before the model does. I’d welcome the conversation. Message me directly.
Notes
Korn Ferry, “Capabilities for an AI-Ready Workforce.”
Korn Ferry TA Trends survey, cited in Korn Ferry, “AI in Recruitment Trends.”
Wegner, D. M. (1987). Transactive Memory: A Contemporary Analysis of the Group Mind. In B. Mullen & G. R. Goethals (Eds.), Theories of Group Behavior. Springer-Verlag.
OpenAI. (2016). Faulty Reward Functions in the Wild. openai.com/index/faulty-reward-functions.
Vaughan, D. (1996). The Challenger Launch Decision: Risky Technology, Culture, and Deviance at NASA. University of Chicago Press.
Case details drawn from a large-scale pricing transformation engagement; company and individuals not named at the client’s preference.
Read This CEO Pricing Strategy To Improve Margin & EBIT
Are you a business in need of help aligning your pricing strategy, people, and operations to deliver an immediate impact on profit?
If so, please call (+61) 2 9000 1115.
You can also email us at team@taylorwells.com.au if you have any further questions.
