The Myth:

AI is the technology that changes everything.

The reality: the biggest wins still come from boring, proven technology applied to the right problem.

Stephen Andriole observed this in 2017 in his original piece for MIT Sloan. He argued that most short-term impact comes from conventional operational technology: networking, databases, enterprise software. Not from the hot new thing. The companies that won were the ones who applied mainstream tools already in consumers' hands.

Think about Uber. The technology that made Uber possible wasn't exotic. It was the phone in your pocket, GPS, mobile payments, a well-designed app. All built on infrastructure that already existed. The innovation was in applying proven technology to a business model problem.

AI follows the same pattern. The tool is extraordinary. The way most companies apply it is ordinary. They bolt it onto broken processes and expect it to fix what's underneath.

Where the Money Actually Goes (and Where It Should)

MIT's "GenAI Divide" report from July 2025 found something that should redirect a lot of budgets. Companies were allocating more than half of their AI spending to sales and marketing tools. But the highest ROI came from somewhere else entirely: back-office automation. Document processing, compliance workflows, internal operations.

The boring stuff.

Deloitte's 2026 State of AI report reinforced this. 74% of organizations hope to grow revenue through AI. Only 20% are actually doing so. That's a 54-point gap between aspiration and execution. And the companies in the 20% share a common trait: they didn't start with the flashiest use case. They started with the most friction.

Sangeet Paul Choudary's February 2026 piece in Harvard Business Review adds a layer that's worth sitting with. He studied Figma vs. Adobe and Shein vs. traditional fashion houses, and his conclusion is that incumbents fail not because they lack AI, but because they bolt new technology onto old structures without rethinking how work gets organized.

His Shein case is particularly telling. Traditional fashion companies still organize around the seasonal collection. Design, forecasting, buying, merchandising, production, all coordinated around a long-cycle calendar. Even when they use generative AI to speed up design sketches, the underlying structure stays sequential. They made the season faster. They didn't rethink the season.

Shein reduced the unit of work from the seasonal collection to continuous small-batch experiments. Every design brief becomes a testable product concept. AI doesn't sit on top of this process as a fancy addition. It connects sensing, decision-making, and action into a single learning loop. The company learns through constant, low-cost experimentation during operations rather than through periodic reviews.

That's not AI changing everything. That's a company rethinking how it works, then using AI (and a lot of boring operational technology) to make the new structure possible.

The SAMR Trap

Dr. Ruben Puentedura's SAMR model keeps showing up in my work because it describes what I see in the field with uncomfortable precision.

Substitution: swap 1 tool for another, no functional change. Augmentation: add some improvement to an existing process. Modification: redesign the work because the technology makes it possible. Redefinition: do something that wasn't conceivable before.

Klarna did Substitution. They swapped humans for AI in the same workflow. The workflow stayed the same. The unit of work stayed the same. The coordination stayed the same. They just made it cheaper and faster. Until the quality collapsed, and it turned out to be neither cheaper nor faster.

Adobe did Substitution. They moved from boxed software to cloud subscriptions, but the file-based, sequential editing model stayed intact. Figma changed the unit of work from the file to the element inside the file. That's Modification. Adobe changed the delivery mechanism. Figma changed how design work was organized.

The pattern repeats across industries. Companies adopt the new tool (AI, cloud, mobile, whatever the wave is) and apply it to the existing way of working. They get incremental gains. They call it transformation. They wonder why the returns plateau.

Donella Meadows wrote that the most powerful place to intervene in a system is often at the level of information flows and delays, not at the level of new components. If information moves slowly through your organization, if decisions take weeks because data lives in 7 different spreadsheets that nobody reconciles, adding AI to that mess is like putting a turbocharger on a car with flat tires.

Fix the tires first. Then talk about turbochargers.

What Actually Produces Returns

I've worked with over 100 businesses on automation and AI implementation. The projects that produce the fastest, most measurable return almost never involve AI. They involve:

Connecting 2 systems that don't talk to each other.

A client was manually transferring data from their CRM to their invoicing system. 3 hours a day, every day. The fix wasn't AI. It was a $50/month integration tool and 2 hours of setup.

Removing manual steps from a workflow nobody has examined.

Another client's approval process required 4 different sign-offs for routine purchases under $500. Nobody remembered why. Removing 2 of those approvals cut processing time by 60% and freed up a senior manager's calendar by 5 hours a week.

Building a simple decision process for routine cases.

A service company was routing every customer request through the same triage process regardless of complexity. A simple rules-based system handled 70% of the requests without human intervention. No AI required. Basic conditional logic.

These aren't impressive. They won't make a conference keynote. But they produce measurable ROI in weeks, not months. And they create the clean, connected, well-documented foundation that AI can actually build on later.

MIT's research backs this up. The 5% of companies that succeeded with AI picked 1 pain point, executed well, and partnered with vendors who understood their workflows. The investment bias that's failing companies, dumping AI budgets into sales and marketing instead of operations, is the same mistake Klarna made. Chasing the headline instead of fixing the plumbing.

The Uncomfortable Question

If Uber's competitive advantage wasn't exotic technology but proven tools applied to a reimagined business model, what does that tell you about your AI strategy?

If Figma beat Adobe not with better features but with a different way of organizing work, what does that tell you about the tools you're buying?

If Klarna's $10 million in AI savings evaporated because they didn't redesign the work before they replaced the workers, what does that tell you about where to start?

The uncomfortable question for Myth 2 is not "are we using AI?" It's "have we fixed the boring stuff that would produce returns right now, with technology we already have?"

In my experience, the answer is almost always no. Companies skip the boring stuff because it's not exciting. Nobody gets promoted for connecting 2 systems. Nobody writes a press release about removing a manual step from an approval chain.

But that's where the money is. And it's where the foundation gets built that makes AI actually useful later.

Where to Start:

If you're running a company, here's the sequence that works:

1. List every manual handoff in your core business process.

Every time a human copies data from 1 system to another, sends an email to trigger the next step, or re-enters information that already exists somewhere, that's a manual handoff. Each 1 is a candidate for automation with proven, boring technology.

2. Calculate the cost of each handoff.

Time spent per instance, times frequency, times the loaded cost of the person doing it. Most companies are shocked when they see these numbers. A 15-minute task done 20 times a week by a $75K employee costs roughly $9,400 a year. If you have 5 of those, you're looking at $47K annually in manual work that doesn't need to be manual.

3. Fix the top 3 with tools you already pay for.

Most companies use less than 30% of the capabilities in their existing tech stack. Before you buy anything new, check whether your CRM, your project management tool, or your accounting software already has the automation feature you need. It probably does.

4. Measure the before and after.

Hours saved. Errors reduced. Cycle time shortened. These numbers become your business case for the next round, and they build the internal credibility you need when you're ready for AI.

5. Then, and only then, ask the AI question.

With clean data, documented processes, and connected systems, AI has something to work with. Without those things, you're Klarna. Bolting a sophisticated tool onto a broken foundation and calling it progress.