The arrival of Claude for Small Business earlier this week marked an interesting moment—and a savvy strategic move—for Anthropic. Rather than saddling web browsers with more AI slop or trying to slather AI onto perfectly good user interfaces that don't need improving, Anthropic is attempting something both less flashy and potentially more fruitful: finding a practical, agentic AI-powered application for everyday business owners looking to make ends meet.
The bag of tricks included in Claude for Small Business is somewhat predictable, running the gamut from "ready-to-run" agentic workflows to connectors for PayPal, QuickBooks, HubSpot, Canva, DocuSign, and more. With these tools, business owners can use Claude to help plan their payrolls, reconcile their books, analyze their cash flow, spin up promotional campaigns, and so forth. That's all well and good, but it also entails trusting Claude to perform those bookkeeping and promotional duties accurately and thoughtfully. I'm guessing many small business owners will (quite rightly) balk at the prospect of handing their tried-and-true workflows to an unpredictable AI model, even one as powerful as Claude.
Small business users were equally hesitant about computers in general at the dawn of the PC age. Sure, an Apple II or a Commodore 64 could balance checkbooks and track inventories, but not much better (or faster) than a human could. Why bother coughing up $1,500 (in unadjusted 1979 dollars) for an Apple II that wasn't much better at bookkeeping than a person with an old-school ledger?
Then as now, what was missing was a killer app—a game-changing application that takes a familiar task and transmutes it into something simple, elegant, and (in retrospect) seemingly inevitable. What could that killer app be? Back in 1979, the answer was VisiCalc, the very first computer spreadsheet. It was a brilliant tool that perfectly leveraged the power of the Apple II (and later many other PC platforms). All of a sudden, small business owners weren't just tracking their expenses and revenue—they were forecasting them, all by changing a single number in a cell. Thanks to VisiCalc (which was later eclipsed by Lotus 1-2-3, and then later Microsoft Excel), the $1,500 price tag for an Apple II looked like a bargain.
I'd argue that we're at a similar pre-VisiCalc moment with AI. Yes, I know, Claude Code is a killer app, but it's only killer for a narrow slice of users. For the rest of us, AI remains an imperfect fit, like (if I can recycle a metaphor) trying to use a socket wrench to slice a wedding cake. With Claude for Small Business and Claude Cowork, Anthropic is nibbling around the edges of what VisiCalc accomplished: finding a truly useful and unique application for AI that offers tangible value to small business owners—and, by extension, to everyday users everywhere.
But trying to shoehorn the agentic AI abilities of Claude Code into the world of small business is, I would argue, a dead end. What makes AI terrific at crafting code—its endless creativity—is what makes it so worrisome when it comes to business. Yes, AI can build meticulously crafted spreadsheets and beautifully crafted bar charts in seconds, but they're useless if you can't trust the data behind them. The key is harnessing AI's power in a different way, applying its strengths to the right applications while turning its flaws (especially its runaway creativity) into virtues.
In short, AI needs a VisiCalc—a killer app that transforms ChatGPT, Claude, and Gemini from slot machines (or slop dispensers) into truly useful tools, ones that make the lives of everyday users—not just coders—tangibly better. All we have to do is find it. Easy, right?
Historical Context: The VisiCalc Revolution
To understand what AI is missing, it helps to look back at the personal computer revolution and the role VisiCalc played. Before 1979, computers were primarily used by hobbyists, large corporations, and universities. The Apple II, introduced in 1977, was one of the first successful mass-market microcomputers, but its sales were modest. Then came VisiCalc, created by Dan Bricklin and Bob Frankston. VisiCalc turned the computer into a dynamic tool for financial modeling and analysis. Business owners who had never considered buying a computer suddenly saw the value: you could change a single number and instantly recalculate an entire budget. This was a profound shift from paper ledgers and calculators.
VisiCalc became the "killer app" that drove sales of the Apple II and later the IBM PC. It didn't just replace an existing tool; it enabled new ways of thinking and working. By 1980, more than 100,000 copies of VisiCalc had been sold. The spreadsheet became synonymous with personal computing. Lotus 1-2-3 improved on the concept in the 1980s, and Microsoft Excel eventually dominated. The point is that a single application made the hardware indispensable. AI is still waiting for that moment.
Current Attempts and Their Limitations
Anthropic's Claude for Small Business is one of several efforts to find a killer app for AI. Other companies are also trying: Google has its Workspace AI features, Microsoft has Copilot, and many startups are building vertical AI agents for specific industries. Yet none have achieved the ubiquity or transformative impact of VisiCalc. Why? Because most of these tools either replicate existing functionality or add friction to workflows.
For example, AI writing assistants can generate text, but they often produce generic or inaccurate content. AI coding assistants like Claude Code are powerful, but they serve only developers. AI chatbots handle simple customer queries but frustrate users when they fail. The promise of AI agents that can automate complex tasks is compelling, but trust remains a major barrier. Users fear that AI will make costly mistakes, especially when dealing with finances, legal matters, or medical advice.
Moreover, AI models are inherently probabilistic, not deterministic. A spreadsheet calculates exactly the same result every time; an AI might give different answers to the same prompt. This unpredictability is fine for creative tasks but unacceptable for many business applications. The very creativity that makes AI good at generating ideas makes it unreliable for precise tasks.
What Would a True AI Killer App Look Like?
A true AI killer app would be one that leverages AI's unique strengths—its ability to process vast amounts of information, recognize patterns, generate creative solutions, and adapt to context—in a way that is transparent, reliable, and immediately valuable. It would not just automate existing tasks but enable entirely new categories of work, much as VisiCalc enabled dynamic forecasting.
One candidate might be a personal AI assistant that manages your digital life: scheduling, email, finances, health, and learning. But such an assistant would need to be trustworthy and private, which current cloud-based models struggle to guarantee. Another candidate is AI-powered research tools that can synthesize knowledge from across the web and produce accurate, cited reports. Some tools like Perplexity AI are trying, but they still hallucinate too often.
In the business world, AI that can autonomously handle regulatory compliance, tax preparation, or supply chain optimization could be game-changing. But again, the stakes for error are high. Perhaps the killer app will be something we haven't imagined yet—a new interface that combines AI with augmented reality, or an AI that learns your personal preferences so well it can anticipate your needs.
The search for AI's VisiCalc continues. Every few months, a new tool emerges that claims to be the breakthrough, but so far none have crossed the chasm from interesting utility to indispensable necessity. The challenge is not just technical but psychological: humans need to trust the tool before they will rely on it. And that trust can only be built through consistent, error-free performance over time.
Until then, AI remains in the pre-VisiCalc era: a powerful but underutilized technology waiting for its killer app to unlock its true potential. The industry is filled with brilliant engineers and entrepreneurs working on this problem, and it's likely that the breakthrough will arrive when we least expect it—just as VisiCalc did in 1979. But for now, we watch and wait, trying to balance excitement with skepticism.
Source: PCWorld News