LISTEN
Listen to this perspective
Written by Oleg Cohen. Companion narration uses the Brian voice.
Every company rations attention, and most have stopped noticing.
The best estimator reviews the largest bids, not all of them. Pricing gets a hard look for the top fifty accounts, and the other few thousand run on schedules set years ago. Audits sample. Service calls start cold, because preparing every one would cost too much. Customers get sorted into tiers, because personal attention for all of them doesn’t scale. The reasoning behind a decision lives in the head of whoever made it.
None of this is a failure. It’s design. Over decades, companies have shaped their processes, service levels, org charts and even what they sell around one constraint. Experienced attention is scarce and expensive. You can’t put your best person in every situation, so you decide which situations get them and accept a rougher answer everywhere else.
That constraint is moving. For an operating company, that is the real reason AI matters, and it’s the part most AI programs miss.
What actually changed
Software has long handled calculation and anything that can be written down as explicit rules. What it couldn’t do was the interpreting. Reading a situation, assembling context, connecting facts and preparing a decision all needed a person, so they were rationed. AI can now do portions of that work, inside systems and alongside people, at a cost that makes it worth doing in places a person never could be.
I don’t mean intelligence is free, uniformly reliable, or a substitute for judgment. I mean the constraint has moved far enough that arrangements a company once rejected as uneconomic are worth reconsidering. Economists would say a frontier shifted. I’d put it in business terms: the boundary around what the company can practically do has moved, and most companies haven’t redrawn their map.
The word “impossible” doesn’t belong here. Most of what sits on the other side of that boundary was always possible in principle, with unlimited staff. It was impractical at the scale the business needed. That distinction matters, because it tells you where to look.
In Your AI Is Getting Smarter. Your Company Isn’t. I suggested that some of a company’s limits are inherent in the business, while others only reflect what knowledge, judgment and coordination used to cost. That essay raised the question. This one is about how to tell the two apart.
Where to look
Asking where to use AI sends a company looking for tasks to speed up. That keeps the business as it is, with better assistance. The more useful question is what the company has been rationing, and what would change if it stopped.
The signals are usually in plain sight:
- something you do only for the biggest customers, deals or sites;
- something you do quarterly that would be worth doing continuously;
- something you sample instead of checking in full;
- a decision routinely made with less preparation than it deserves;
- a handoff where context is lost and rebuilt over the phone;
- a service you once considered and dropped, because delivering it consistently would take too many experienced people.
Each is a place where the company accepted a worse answer because the better one cost too much attention.
Consider a distributor with four thousand accounts. Its pricing team gives real review to the top fifty. The rest run on discount schedules set years ago and adjusted when a customer complains. Nobody thinks this is ideal. It’s what the team’s attention allows. Now suppose every account could get the review the top fifty get, covering cost to serve, freight, payment behavior, what similar customers pay and what this one has accepted before. A pricing manager still decides. But the long tail stops being priced by neglect, and for many distributors the long tail is where margin quietly leaks. This is an illustration, not a case, but the shape is common.
That isn’t a faster version of the existing process. It’s a capability the company didn’t have before: pricing judgment applied to every account. Earlier essays in this series argued that the unit of value is a capability, not a tool. Looking for what you ration is how you find one.
Three tests
Not everything rationed deserves to be un-rationed. I’d apply three tests.
Is it valuable? Doing it everywhere should change revenue, margin, retention, risk or capacity, not just make the work tidier.
Is it yours? The better answer should depend on the company’s own history, customers and judgment. If any competitor buying the same tool gets the same result, the gain will pass through to customers, as the first essay in this series, EBITDA Up, Value Down, argued. Look for rationed work where your own experience is the input.
Can you see whether it worked? If the company can compare the outcome with what it expected, it can learn. The price held or it didn’t, the margin came in or it didn’t, the customer stayed or left. Without that comparison, you have automated a guess.
The third test is where most of the lasting value sits. A distributor that reviews every account and then sees which prices held learns something every month. One that only prices faster learns nothing.
What it takes
Finding the opportunity is business work. Making it run touches everything else. Someone needs authority to change a price. The company’s history has to be usable, not buried in a system. Several systems have to feed the decision, and the results have to come back to it. That is why this can’t simply be handed to IT, as I argued in The Missing Owner, and why a quick prototype won’t produce it, as I argued in Busy With AI. Owning the outcome, designing the capability and delivering the technology are three different jobs, and all three are needed.
Why I see it this way
I should say why I look at it this way, and why I think it can be done. I didn’t arrive at business through AI. I arrived at AI through more than thirty years of building businesses and complex systems.
My first training was in theoretical physics, including quantum chromodynamics. That left me with a habit of asking which constraint shapes a system and what happens when it moves. Three decades of designing and delivering business systems taught me the other half, which is how much work lies between a good idea and something that runs every day.
Two experiences matter most here. For seventeen years, AssureBridge, essentially a two-person company, has served Fortune 500 customers around the world. A company that small can’t meet enterprise demands by adding people, so the leverage has to be built into how the work is done. That is the argument of this essay, practiced long before AI made it easier. At ShakeIQ, the ambition for the platform outran what the funding would support on the conventional path. The architecture and the delivery model had to change until the platform could be built and delivered. A constraint that seemed to rule the product out turned out to call for a different design.
Kainora is where I’m applying the same ideas now, including to how Kainora itself operates. AI is new. My approach isn’t. What’s new is how much further it can go.
Where I sit: I advise through Kainora and am building technology aimed at this problem, turning work a company used to ration into capability it can repeat and learn from. It’s in production with one company and in pilots with several more, and I’d still rather be tested by readers than taken on trust.
Two questions for you. What does your company ration today? And which of those, if you stopped, would a competitor find hardest to copy?