I know someone who's getting rid of QuickBooks.
A few years ago that would have sounded like a bad idea — a small business owner abandoning their accounting software for… what exactly? Now it sounds like a calculated move.
Here's what he's doing instead: keeping all his transactions on a local server and using AI prompts for any analysis or reporting he needs.
“What was our ad spend in November and how much revenue did it generate?”
“Which clients are trending down this quarter?”
“What's our margin by product line for the last six months?”
QuickBooks can't answer those questions in a conversational way. You'd need to know which reports to run, which filters to apply, and then interpret the output yourself. With AI and raw transaction data, he just asks.
So he stopped paying for QuickBooks.
Why This Is Happening
The software categories most exposed to AI disruption share a common characteristic: their primary value is providing structured access to your own data, or generating outputs from templates you could replicate with a good prompt.
Jasper was one of the early casualties. It launched as an AI writing tool that charged a significant monthly subscription for generating marketing copy. When OpenAI introduced custom GPTs, a lot of people realized they could build a “Jasper” for their specific brand voice and use case for basically nothing. The subscription evaporated.
The same logic is now spreading to other categories. When AI can replicate what the software does — and often answer questions the software was never designed to handle — the subscription becomes harder to justify.
Which Tools Are Most at Risk
Single-function reporting tools. Any software whose main job is to slice and display your existing data in a particular format is vulnerable. You don't need a dedicated tool to query your data if AI can query it directly.
Template-based content generators. Tools that charge to generate variations of a specific content format — social captions, email subject lines, ad copy — have been among the first to lose customers to custom GPTs and Claude. The underlying generation is now a commodity.
Simple workflow automation with no native integrations. Tools that charge primarily for their automation logic, without deep integrations or proprietary data, are increasingly replaceable by AI agent builders.
Specialized research or analysis tools. Tools built to process a specific type of document or answer a specific type of business question may be replaceable by a capable AI model with the right prompt and access to your data.
Which Tools Aren't Going Anywhere
Core infrastructure with compliance requirements. Accounting systems used for tax compliance, payroll platforms with legal obligations, systems that need to produce auditable records — these aren't going anywhere. Compliance isn't a prompt problem.
Deep integration hubs. Software that serves as the hub connecting your entire stack — a CRM with deep integrations across your sales, marketing, and service tools — is harder to replace because the value is in the network, not just the features.
Specialized tools with proprietary data or networks. Software where the value is the data inside it (industry databases, market intelligence, professional networks) rather than what it does with your data is in a different category.
Communication and collaboration infrastructure. Email, messaging, and video conferencing are table stakes — they're not being replaced by prompts.
How to Audit Your Own Stack
Go through your active software subscriptions and ask two questions about each one:
1. Does this tool primarily help me get structured access to my own data?
If the answer is yes, and your data could be exported and made accessible to an AI model, there's a case for evaluating whether a custom solution could replace it.
2. Does this tool generate outputs I could replicate with a well-crafted prompt?
If the main deliverable is a type of content, summary, or analysis that AI can produce, the standalone subscription is worth questioning.
For the tools that fail both tests — infrastructure, compliance, network-value software — keep them. For the ones that pass, the question is whether building or prompting is worth the effort compared to the subscription cost.
A commercial real estate firm I worked with uploaded their ledgers and balance sheets directly to Claude Code and started asking it to identify issues and build financial models. What used to take days of manual analysis now happens in minutes. That's not a special case — it's a preview of how a lot of specialized business tools will get displaced.
The Skill That Matters Now
Knowing which tools to question is becoming a real competitive advantage.
The businesses that are paying full price for mid-tier software that AI has made redundant are subsidizing a slow transition. The ones doing the audit and rebuilding the tools worth replacing will have lower overhead and more flexible infrastructure.
This isn't about eliminating software. It's about understanding what the software is actually doing, and whether AI can do that thing better, cheaper, or on terms you control.
Start with your smallest subscriptions. Those are the easiest to experiment with. Find one tool where most of the value is “it organizes my data so I can query it” — and see what happens when you move that logic into a prompt.
One thing to try this week: Open your subscriptions list and find one tool that primarily summarizes, reports on, or generates content from data you already own. Export a sample of that data and ask an AI model the question you'd normally use the tool to answer. See how close it gets.
