Innovation Is Not Reinvention. The Difference Matters.

Part 2 of a three-part series on innovation, reinvention, and the migration of customer value.

Part 1: https://www.valueprop.com/blog/VP-Insider-03

Here's something that should bother you a little: a company can be genuinely, legitimately innovative — great engineers, billions in R&D, new products landing every year, patents stacked to the ceiling — and still lose the thing that matters most.

It can still watch someone else walk off with the next important customer relationship. The next profit pool. The next position of control.

That's why I keep two words apart that most people use interchangeably: innovation and reinvention.

Innovation creates or improves a product, a process, a service, or a capability. Reinvention changes what the company is built to be valuable for.

Sometimes one leads to the other. Sometimes innovation just makes the existing business a little better — polish on the same car. And sometimes a company keeps innovating energetically while the market quietly walks value out the back door.

I want to show you what I mean using four companies you already know: IBM, Microsoft, Google, and Amazon.

Let me say up front — none of these are operating models for your $8 million shop. Their resources and options aren't yours, and they never will be. But the patterns travel just fine down to your scale. Watch for them:

• IBM shows how a successful reinvention can quietly become the next constraint.
• Microsoft shows how to keep the old cash engine running without letting it dictate the future.
• Google shows that inventing the technology doesn't guarantee you lead the market it creates.
• Amazon shows that being physically essential doesn't protect you from losing the decision.

Put together, they teach something uncomfortable:

The company that invents the future isn't necessarily the company that captures its value.

IBM: the reinvention that saved the company

Let's give credit where it's due first. Lou Gerstner didn't wreck IBM by pivoting it toward services. He rescued it.

When Gerstner walked in the door in 1993, IBM was bleeding — $8.1 billion lost in 1992 alone, as the PC and distributed computing gutted the economics of the old mainframe-centered model. This was existential, not a rough quarter.

Gerstner's insight was simple to say and hard to execute: IBM's future couldn't depend on selling customers a self-contained IBM technology stack anymore. Instead, IBM would get paid to make everyone else's increasingly fragmented technology work together — hardware, software, networks, data, business process, all of it. Not the machine. The capability to make the whole mess function.

It worked. Revenue grew nearly 40% between 1994 and 2000, net income nearly tripled, and by 1999 IBM was running e-business services for over 10,000 clients. Gerstner would later say the company had found its confidence again and its ability to set the industry's agenda.

That's not a rebrand. That's a real reinvention, and it worked because Gerstner read where value was actually moving: corporate technology was getting more heterogeneous, and the customer's problem stopped being "get computing power" and became "make it all work together reliably." IBM had the breadth, the relationships, and the execution muscle to own that problem.

Here's the harder question, though — not whether Gerstner was right. He was. The question is: how long should that have remained IBM's governing assumption?

When the reinvention becomes the doctrine

IBM's integration model rested on a few load-bearing beliefs: enterprise tech would stay fragmented, big companies would stay nervous about giving up control of critical systems, and somebody big and trustworthy would always be needed to coordinate the chaos.

True for a long time. Then, cloud computing started dissolving the problem the model was built to solve. Infrastructure became a subscription. Software became a subscription. Security, identity, governance — increasingly just part of the platform rather than something you hired someone to bolt together.

Customers still had integration headaches, old systems, regulatory tangles, and weird internal politics. But more of the complexity was getting absorbed by the provider itself, before the customer ever had to think about it.

And trust flipped. The old logic was we're safer if we own and control the infrastructure ourselves. The new logic is that we may be safer on a platform that's professionally run and updated continuously at a scale we could never match.

That didn't erase IBM's role. It moved IBM later in the process, from the first call that designs the whole environment, to the specialist you bring in once the standard platform hits its limits. 

I'll give IBM this: it never stopped inventing. IBM Research is still formidable — AI, quantum, hybrid cloud, all of it. But its own 2025 annual filing admits, in plain language, that the company risks losing ground if it can't commercialize and scale its innovations fast enough, and that new delivery models could shrink demand for products it already sells.

That's the real story. IBM's problem was never a shortage of new technology. It was repeatedly turning new technology into a new center of gravity for the company. The Gerstner playbook got institutionalized. The Gerstner act — the willingness to fundamentally reread the market and move — didn't get repeated.

The IBM lesson

A successful reinvention is not a permanent constitution. It's a response to a market condition that existed at a moment in time. The very strategy that saves a company can, years later, become the wall it can't see over — if nobody in the room keeps re-testing the assumptions underneath it.

For an industrial company, the lesson isn't "go become a software business." It's simpler than that, and it should sting a little:

When your value depends on managing complexity for the customer, you'd better be watching closely for the market's attempts to remove that complexity.

The expertise can still be real. The hard cases can still need you. And you can still slide from the customer's first call to their last resort. That's value migration, and it doesn't announce itself.

Microsoft: turning the old identity into an asset

Microsoft's problem looked nothing like IBM's, and that's what makes it instructive. Windows and Office weren't failing. They were two of the most valuable franchises in business history. Which, weirdly, made reinvention harder, not easier — nobody feels urgency around a golden goose.

When Satya Nadella took over in 2014, Microsoft was still enormously profitable — and still deeply identified with a computing world built around Windows and the PC. Meanwhile, the world was sprinting toward mobile, the cloud, subscriptions, and software that had to work everywhere, not just on a Microsoft machine.

Microsoft could have kept treating Windows as the center of the universe. It's what everyone inside the building had been rewarded for their whole careers. Instead, Nadella started describing the company as "mobile-first, cloud-first." One of the earliest tells: putting Office on the iPad, and rolling Office out on competitors' platforms. Microsoft quietly stopped requiring you to buy Windows before it would sell you Microsoft.

That's not a distribution decision. That's a change in strategic hierarchy. Windows still mattered. Office still mattered. Neither one got to decide where Microsoft was allowed to compete anymore.

Look at where that landed. In fiscal 2025, Microsoft Cloud revenue hit $168.9 billion — up 23% — with Azure and related cloud services up 34%. Windows and Devices, still a real business, is now a much smaller slice of what actually drives the company.

Notice what Microsoft didn't do. It didn't kill the cash cow. It changed the cash cow's job. Windows and Office became the thing that funded, distributed, and lent credibility to the new business, instead of the fence around what the new business was allowed to become.

That distinction matters more than it sounds like it should, because I see the opposite of it constantly in industrial companies. A mature product line is genuinely valuable — it throws off cash, it gets you in the door, it builds credibility, it buys you time. The trouble starts when that mature business also gets a veto over the future.

Sound familiar? A sales team that quietly resists a new offer because it messes with comp plans. A distributor that slow-walks digital ordering because it shrinks the salesperson's role. A manufacturer that avoids a service model because it might cannibalize parts revenue. A company that won't simplify the product because complexity is what the engineering team gets paid to manage.

In every one of those cases, nobody's rejecting innovation out loud. They're protecting the economics and the org chart of the business they already understand.

The Microsoft lesson

The old business doesn't have to die for the company's identity to change. That might be one of the more sophisticated moves in strategy, and it's worth memorizing:

Preserve the economics of what already works — without letting it define the ceiling of what comes next.

Let the mature line keep serving customers and throwing off cash. Just make sure it's funding the next reason customers will prefer you, not blocking it.

Google: inventing the technology isn't the same as leading the market

Google wasn't caught flat-footed on AI. Its own researchers were central to inventing the thing. In 2017, a team out of Google Research and Google Brain published the paper that introduced the transformer architecture — the backbone underneath basically every modern large language model. Google had sophisticated language models buried inside Search years before "generative AI" was a phrase anyone outside a research lab used.

Then OpenAI released ChatGPT at the end of November 2022, and everything changed — not technically, but experientially. Suddenly, the public had a new way to interact with information: ask, get a synthesized answer, refine, keep going. Google had its own public answer, Bard, out within a few months.

A few months doesn't sound like much of a lag. Strategically, it was enormous.

Google wasn't caught without AI. Google was caught without the product experience that suddenly redefined what everyone assumed AI could do.

Here's why that gap opened at all, and it's the interesting part: Google had the research, the talent, the data, the infrastructure. It also had one of the most profitable business models ever built. Classic search shows you links; advertisers pay to sit near those links. A synthesized answer changes the number of searches, the number of clicks, how ads get shown, and the whole relationship with the websites Google's answers are quietly built on top of.

OpenAI had none of that to protect. It could ship a conversational answer engine without first figuring out how it would gut an existing empire. Google had more capability and more to lose by using it well.

That's a very common incumbent trap, and I'd bet good money you've felt a smaller version of it in your own business: you can see the future clearly — you may have even helped invent it — and still be institutionally unable to move on it, because moving on it means damaging something that currently pays everyone's salary.

To Google's credit, it's now responding hard — AI Overviews rolled out broadly in the U.S. in 2024, Gemini is getting stitched into Search, Workspace, Cloud, everything. Alphabet's 2025 annual report is candid that these AI-driven experiences may simply monetize differently than the businesses they grew up on. And the numbers are real: Search-related revenue hit $224.5 billion in 2025, Cloud revenue grew 36% to $58.7 billion, and Gemini is reportedly running across all fifteen of Google's major products with half a billion users.

Google may still land this. Responsive reinvention can absolutely win — Microsoft didn't invent the public cloud either, Amazon did, and Microsoft still built Azure into a juggernaut by reacting decisively and leaning on its enterprise relationships. Google has comparable assets. The open question isn't whether Google can build excellent AI. It obviously can. It's whether it's willing to disrupt the search business as aggressively as a competitor with nothing to protect.

The Google lesson

Technical leadership doesn't automatically become market leadership. Sometimes the obstacle isn't a shortage of ideas, talent, or technology at all. It's the success of the model you already have.

Here's a question worth asking in your own conference room, uncomfortable as it is:

What would we offer customers if we weren't worried about cannibalizing how we currently make money?

You don't have to launch the answer tomorrow. But if leadership can't say it out loud honestly, you probably don't understand where the actual threat is coming from yet.

Amazon: AI can move the decision, but it can't move the box

Amazon is the odd one out in this list, and I think it's the most instructive.

It started as a bookseller, expanded into general retail, opened itself to third-party sellers, built one of the largest fulfillment networks on earth, spun up AWS almost as an internal side project, and grew a major advertising business along the way. Every expansion built on the last one — but each one also changed what Amazon fundamentally was. Books were never the destination. Books were the door.

By 2025, that engine included Amazon's own retail sales, third-party seller fees, advertising, subscriptions, a global fulfillment network, and AWS, which alone did $128.7 billion in sales that year.

AI is going to change Amazon again, and the most obvious pressure point is product selection. Picture a genuinely capable shopping agent — one that asks what you actually need, compares options honestly, weighs tradeoffs, checks prices, and just tells you what to buy. That could beat scrolling through filtered listings, sponsored placements, and eight hundred conflicting reviews without much of a fight.

But here's the catch, and it's the part that should make Amazon's leadership a little uneasy: for that agent to be trusted, it has to optimize for the shopper's decision, not for Amazon's next transaction. Sometimes that means recommending something Amazon doesn't sell. A direct-from-manufacturer purchase. A local shop. A repair instead of a replacement. No purchase at all.

Otherwise, it's not really an agent. It's just a fancier merchandising engine wearing a friendlier face.

So why would Amazon build something that occasionally sends the customer somewhere else? Because owning the decision might end up being worth more than winning every individual transaction. Amazon's already testing the edges of this with features that surface and help you buy products from brand websites even when Amazon doesn't carry them — explicitly framed around helping you find and buy the thing you want, wherever it lives.

That's not generosity. That's intelligence-gathering with a checkout button. An open shopping agent tells Amazon in real time which products customers want that Amazon doesn't have, which brands keep beating Amazon's own alternatives, where price points are landing, which manufacturers Amazon should be courting, and which categories deserve better assortment or an acquisition. A lost sale today can become next quarter's sourcing strategy.

This is the same logic that built Amazon Marketplace in the first place. Letting third-party sellers onto the platform looked, on the surface, like giving away sales Amazon could have kept for itself. Instead, it expanded selection, pulled in more traffic, generated fees, and made Amazon more useful — which made Amazon stronger, not weaker.

Own more of the decision. Then use what you learn to strengthen everything downstream of it.

Amazon's physical advantage doesn't disappear

Here's the part I want you to sit with, because it's the most directly relevant one to your business.

AI can transform discovery and comparison, and selection. It cannot transform physics. A product still has to be made, warehoused, picked, packed, routed, delivered, and sometimes returned. Amazon runs its own enormous fulfillment network and keeps pouring money into robotics — by 2025, it had deployed its millionth robot across more than 300 facilities.

That's very hard to replicate. AI can make that network smarter — better forecasts, better routing, better placement — but it doesn't replace the network. Until somebody invents a Star Trek transporter, electrons still can't move a box from a warehouse to your customer's dock.

The real risk is a different sentence entirely:

Someone else's electrons may end up deciding whose box gets moved.

A third-party AI agent could own the whole customer relationship — do the research, pick the product — and only route the order to Amazon when Amazon happens to offer the best price and delivery. Amazon stays essential. It also quietly becomes a fulfillment utility sitting behind somebody else's decision screen.

That's why Amazon has to defend two completely different layers at once: the physical muscle to execute the transaction, and the informational position to influence the decision that leads to it.

The Amazon lesson

Owning production or fulfillment does not mean owning customer preference. A company can stay physically indispensable and still become commercially interchangeable — those are not the same thing, and I don't think most industrial owners have fully absorbed that they aren't.

You may keep making the part. Your distributor may keep stocking it. Your integrator may keep installing it. But if somebody else owns the specification, the comparison, the recommendation — that somebody else captures the customer data, sets the buying criteria, and claims a bigger piece of the margin.

The product stays put. The value walks around it.

Four companies, four different problems

Here's the thing I want you to notice: innovation isn't actually the variable that separates these four stories. All four companies innovate constantly. What separates them is something else entirely.

IBM shows what happens when a successful reinvention hardens into a permanent doctrine.

Microsoft shows how to turn an old identity into fuel instead of a fence.

Google shows that you can help invent the future and still get outmaneuvered commercially by someone with less to protect.

Amazon shows that physical execution can stay rock-solid while the decision layer underneath it quietly migrates to somebody else.

So the useful question was never which company innovated the most. It's:

Which company most clearly saw where value was moving — and rebuilt itself to go get it?

That's the question that actually applies to you.

Innovation leaders aren't always reinvention leaders

I hear a version of this from owners all the time: "We've invested in new equipment. We launched a new product. We're piloting AI in customer service." Good — genuinely. Those things can matter.

But none of them, by themselves, answer the strategic question.

You can invest heavily and still be reinforcing the wrong position. You can get very good at a capability that customers are starting to take for granted. You can automate a middleman role that the market is slowly erasing. You can produce a technical breakthrough that someone else quietly packages into the experience customers actually prefer. You can keep doing the physical work beautifully while losing all say in why and how customers choose you.

The real job — your job — is connecting the innovation to the migration. Where is customer trust actually moving? Where is expertise getting embedded into someone else's product? Where is complexity being stripped out from under you? Who's getting the customer's first call now? Who's collecting the data that tells them what customers really want? Which of your current strengths is quietly becoming table stakes? And which of your current profit pools should be funding whatever comes next?

Those questions matter exactly as much at $8 million in revenue as they do at $280 billion. The answers will look completely different. The discipline of asking doesn't.

The point was never to imitate the giants

I'll say it plainly: don't try to build Microsoft's R&D budget, Amazon's fulfillment network, Google's research bench, or IBM's global consulting arm. That's not the takeaway, and if that's what you walked away thinking, I've failed to make my point.

But an $8 million manufacturer can absolutely make the same category of mistake these giants made.

You can treat a value proposition that's worked for fifteen years as if it's permanent. You can let a profitable product line quietly block investment in a more valuable capability. You can have the best technical knowledge in your market while somebody else builds the easier customer experience around it. You can keep making and shipping an essential physical product while losing your grip on the specification and the buying decision entirely.

None of that is a technology company problem. It's a business-model problem, and business-model problems don't check your revenue before they show up.

No market is completely static. The product may stay. The customer's need may stay. Your people, your machines, your materials — all of it may stay exactly where it is.

But the reason customers prefer you can still change out from under you. That's why innovation is necessary but not sufficient — and it's why the real test isn't whether you built something new. It's whether you became more valuable while the market changed what it values.

Next in the series: the reinvention questions every industrial owner should be asking — and a practical way to decide where innovation belongs in your business, and where it doesn't.

• Which of these four — IBM, Microsoft, Google, or Amazon — looks most like a mistake you've seen (or made) at a smaller scale?
• Where in your business does the mature product line have quiet veto power over something newer and better?
  
  
  

Research Source Notes:
-IBM's official corporate history, "A Prescient Vision" / The Birth of E-Business (ibm.com/history); IBM FY2025 Form 10-K, Item 1A "Risk Factors" (SEC filing, Feb. 2026).
-Satya Nadella's day-one CEO memo, reported in "Satya Nadella's Vision For A New Microsoft," TechCrunch (July 2014); Microsoft FY2025 Annual Report (microsoft.com/investor).
-"A New Chat Bot Is a 'Code Red' for Google's Search Business," The New York Times (Dec. 2022); Alphabet FY2025 full-year financial results (abc.xyz/investor).
-"Amazon launches a new AI foundation model to power its robotic fleet and deploys its 1 millionth robot," Amazon official newsroom (aboutamazon.com, July 2025); Amazon 2025 Annual Report / Form 10-K (SEC filing).