From Turing to today —
why AI feels different this time
Investors have seen tech hype cycles rise and fall before. To separate real structural change from marketing noise, you must understand the fundamental shift in how software is now being made.
Start here: The paradigm shift
For seven decades, artificial intelligence was deterministic. Software engineers wrote explicit, rigid rules: “If X happens, do Y.” This required humans to anticipate every possible scenario and write code for it. While this worked for basic databases, it made software highly fragile, expensive to maintain, and completely unable to handle messy, real-world data.
Today, AI is probabilistic. Instead of writing instructions, we feed algorithms massive datasets and let them figure out the patterns themselves. When you interact with a modern AI system, it is not executing a pre-written rule; it is predicting the most likely next step based on everything it has analyzed. This allows software to understand context, reason through complex problems, and adapt without being reprogrammed.
“Legacy software was a machine built from gears we had to design by hand. Modern AI is an engine that grows its own parts. It fundamentally changes the capital required to scale a business.”
The three waves of software economics
To understand why this moment is different, look at how the cost structure of software has evolved through three distinct eras:
Four structural shifts that change deal dynamics
This is not a normal software upgrade cycle. The move to probabilistic systems introduces four fundamental shifts in how you must analyze target operations.
From Code to Context
In legacy software, intellectual property lived in the proprietary code. In the AI era, code is rapidly commoditizing. The real value and competitive moat have shifted to proprietary data pipelines and operational context.
Linear vs. Exponential Scale
Traditional SaaS scaling was bottlenecked by human processes (support, data entry, basic analysis). Generative AI can execute these cognitive tasks at near-zero marginal cost, fundamentally decoupling revenue growth from headcount growth.
The Hallucination Liability
Traditional software is either right or broken. AI software can be completely wrong while appearing perfectly correct. This changes the risk profile from predictable software bugs to unpredictable operational and legal liabilities.
Unprecedented Velocity
Previous tech shifts (like mobile or cloud) took a decade to play out. The current AI infrastructure is scaling globally in months. Companies that fail to adapt their workflows are losing competitive advantages on a highly compressed timeline.
Why this shift changes your underwriting models
For decades, software margins were highly predictable: high gross margins (80%+), offset by significant R&D and Sales & Marketing costs. The new paradigm alters both sides of this equation. R&D productivity is skyrocketing as AI tools help engineers write code 30-40% faster.
However, if an asset relies heavily on external API calls to power its intelligence, gross margins can degrade quickly under heavy usage. When modeling a target’s future EBITDA, you must stress-test their “AI unit economics.” If their gross margin doesn’t improve as they scale, they are not leveraging the structural advantages of the technology.
The single signal that separates builders from bubble-riders
In our due diligence work at idbokx, the most reliable indicator of a company that actually understands this paradigm shift is Workflow Re-engineering.
Laggards simply bolt an AI chat interface onto their old legacy software and call it a day. Leaders completely redesign their underlying business processes. They look at where humans are slowing down decision-making and use AI to automate the entire pipeline from input to action. Look for companies that are shrinking cycle times, not just adding shiny features.
Assessing an asset’s AI maturity and defensibility?
Our AI Opportunity & Risk Assessment Framework helps deal teams cut through the generative AI hype, evaluate the defensibility of a target’s data moats, and identify immediate operational efficiency plays.
