Steve Anderson on Successful Failure & AI: Why Playing It Safe Is the Riskiest Move for Small Businesses
Q1. The “Door One vs. Door Two” Illusion: In The Bezos Letters, you highlight Jeff Bezos’s framework of separating irreversible “Door One” decisions from easily reversible “Door Two” experiments. Why do small business owners treat a $20/month AI subscription like a Door One marriage—and is their hesitation actually about the software, or are they terrified of admitting they don’t know how their own workflows really operate?
First, a small correction because Bezos’s wording makes the point clearer. In his 2015 letter to shareholders, he split decisions into Type 1 and Type 2. Type 1 decisions are one-way doors. Once you walk through, you can’t come back. Type 2 decisions are two-way doors. Walk through, look around, and if you don’t like what you see, walk back out.
A $20-a-month AI subscription is about as two-way as a door. Cancel it, and you’re out twenty bucks.
So why do owners treat it like a marriage? I don’t think it’s the software. Your second guess is closer to the truth. To test an AI tool, you have to describe the work you want it to do step by step. Most owners have never written down their workflows. The steps live in their heads or in the head of the one employee who’s been there 15 years. The first honest test of an AI tool exposes that gap, and nobody enjoys realizing they can’t explain how their own business runs.
That discovery is worth far more than the subscription. This week, before you buy anything, write down one process from start to finish. Then hand that write-up to the AI and see what it does with it.
Q2. Risk Aversion Masquerading as “Due Diligence”: You’ve built a career studying risk and have noted that people aren’t afraid of failure, they’re afraid of the consequences of failure. When a founder spends six months “researching” AI tools without testing a single one, at what point does prudent risk management become pure cowardice disguised as strategy?
I’d use a softer word than cowardice. The owner, stuck in research mode, is protecting something. A friend of mine put it this way years ago, and I’ve repeated it in nearly every talk since: people aren’t afraid of failure; they’re afraid of its consequences. For an owner, those consequences are looking foolish in front of the team or wasting hard-earned money.
Here’s the test I’d use. Research is prudent when it’s meant to inform a decision. It becomes avoidance when it’s meant to feel ready. If you’ve read twenty reviews and haven’t logged into a single tool, you’re past the point where more information helps.
Bezos has a rule for this, also from his letters: make most decisions with about 70% of the information you wish you had. Wait for 90%, and you’re probably moving too slowly. Spending six months reading about AI tools without testing one is waiting for 100%. You’ll never get there because the tools change every few weeks.
Pick one task you do every week. Run it through two tools on Monday. By Friday, you’ll know more than six months of research would tell you.
Q3. Polishing Broken Processes (Innovation vs. Invention): You recently argued that business owners need to “stop innovating and start inventing” with AI. When owners agonize over picking the “perfect” AI tool to speed up a legacy task, aren’t they just using shiny technology to automate a process that shouldn’t exist in the first place?
Often, yes. That’s the heart of the talk I gave at TEDxBackBay at the Harvard Club, “Stop Innovating. Start Inventing.”Innovation improves what already exists. Invention creates something that didn’t exist before. Most owners point AI toward innovation: make the old report faster, write the same follow-up email in less time.
Michael Hammer foresaw this in 1990. His Harvard Business Review article, titled “Reengineering Work: Don’t Automate, Obliterate,” warned of it. Thirty-six years later, we’re making the same mistake with a smarter tool.
Speeding up routine work is still worthwhile. One of my 14 growth principles is Accelerate Time with Technology. But before automating anything, ask one question: if I were starting this business today, would this step even exist? Many legacy tasks exist only to work around a limit that no longer exists. Someone retyped data because two systems couldn’t communicate. Someone built a weekly report because no one could ask the numbers a direct question. AI removes those limits. Polishing the workaround misses the bigger opportunity.
The invention question is different. What could you do for your customer now that wasn’t possible two years ago? That’s where the growth lies.
Q4. The Sunk-Cost Trap of “Tech Stack” Loyalty: Most small businesses suffer from tech fatigue, yet switching costs for modern AI tools are practically zero compared to legacy enterprise software. Why are founders so quick to fire an underperforming vendor, yet they’ll cling to a mediocre AI workflow—or refuse to test a competitor—just to avoid feeling like they wasted three hours of setup time?
I’d push back on half the premise. In my experience, founders aren’t quick to fire underperforming vendors. Most keep mediocre vendors for years because switching feels like a project. They carry that habit into AI.
But you’re right about the sunk-cost piece, and AI raises the cost of giving in. Three hours of setup feel like an investment, but they aren’t. They’re tuition. You learned what you needed and how to explain your work to a machine. That knowledge travels with you to the next tool.
Legacy software locks you in with contracts, data migration, and retraining the entire staff. Most AI tools require none of that. Your prompts and process notes are plain text, so you can paste them into a competitor’s system within an afternoon.
Run a bake-off every quarter. Take your three most-used AI workflows and run each through a second tool. Keep whichever wins. Sometimes the incumbent will win, and that’s fine. You’ll know it earned its place.
Q5. The Real Threat Isn’t Picking the Wrong Tool: If an AI tool fails to deliver ROI after 30 days, the worst-case scenario is losing a few bucks and a weekend of tinkering. What is the actual existential risk to a small business owner who refuses to make messy, reversible AI bets right now—and how fast does that window close before a competitor eats their lunch?
The real risk is slow, and that’s what makes it dangerous. In his 2016 letter, Bezos described Day 2: “Day 2 is stasis. Followed by irrelevance. Followed by excruciating, painful decline. Followed by death.” He added that the decline happens in extreme slow motion. You don’t feel it until it’s too late.
Blockbuster is my favorite example. At its peak, it operated 9,000 stores and generated $6 billion in revenue. In 2000, Netflix offered to sell itself to Blockbuster for $50 million, but Blockbuster declined. By 2010, it was bankrupt. One store remains in Bend, Oregon, and it’s a tourist attraction.
The owner who refuses to make small AI bets faces the same problem. Each bet teaches you something about your customers and your processes. Those lessons compound. A competitor who’s been running low-cost experiments for two years understands what AI can and can’t do for their business. You can buy the same tools they bought, but you can’t buy their two years of learning.
How quickly does the window close? I can’t give you a date, and I’d be wary of anyone who claims to be able to. I’d rather you act as if it’s closing now. A failed experiment costs you a weekend. Waiting costs you time on the learning curve.
Qx. Anything else you wish to add?
One distinction I’d add. Most owners think of AI as operational: tools that sit within a process and perform a task. That’s useful. The bigger payoff for a small business owner is what I call Leverage AI, using AI as a thinking partner you’ve taught about your business. Test an idea before you spend money on it. Have it poke holes in your plan before your customers do.
If you want a place to start, I built a free AI version of my book, The Bezos Letters Experience. Bring a real question about your business, and it walks you through it using the book’s principles, starting with a Risk & Growth Assessment. It’s not Jeff Bezos, and it won’t make decisions for you. It will ask better questions than most people.
……………………………………………………………………..

Steve Anderson helps small and mid-sized business owners grow like Amazon. He’s the author, with Karen Anderson, of The Bezos Letters: 14 Principles to Grow Your Business Like Amazon, a Wall Street Journal and USA Today bestseller translated into 22 languages. A keynote and TEDx speaker and an AI Strategist, Steve has led The Anderson Network since 1999 and co-founded Catalyit, where he served as CEO from 2021 to 2024. He has 320,000 followers on LinkedIn. thebezosletters.com.