Traditional SaaS models benefited from near-zero costs for new users. AI's high computational and token costs upend this, creating deeply unprofitable users and workflows unless firms carefully manage implementation and pricing.
Many companies claim their proprietary data is a key asset for AI. However, this data is worthless if the underlying software architecture is too brittle or poorly designed to extract and feed it into modern AI models, making the supposed moat an illusion.
Technically skilled CTOs at companies with mundane products can get bored and over-engineer solutions. They build overly sophisticated architectures that read "like an AWS catalog," introducing unnecessary complexity and cost that doesn't align with actual business needs.
When a founder or CTO builds an entire custom tech stack—including their own programming language and compiler—it's a massive red flag. This creates an unmanageable key-person risk, makes hiring impossible, and signals a mindset resistant to collaboration or external validation.
Most technical problems discovered during diligence can be fixed. The real deal-killer is a loss of trust. When a company actively hides major issues, like a failed penetration test, it signals a fundamental dishonesty that makes a future partnership untenable, leading to an immediate abort.
Contrary to the narrative that AI will eliminate junior developers, their value is increasing. Young engineers often have an AI-native "growth mindset," making them more willing to experiment with and adapt to new agentic development workflows than more experienced developers.
The ultimate goal for portfolio management is shifting from episodic due diligence to a continuous process. By constantly assessing and improving a company's tech posture, a future sale becomes a "non-event" where a comprehensive vendor fact book can be generated on demand.
In a sale process with multiple potential PE buyers, vendor due diligence acts as a critical shield. An external firm can handle initial tech workshops for all bidders, protecting the internal development team from being distracted and unproductive for months.
Companies may build features, like a mobile app, that see almost zero user adoption. However, these can have immense marketing value, with customers citing their mere existence as a key reason for purchasing the product, even if they never actually use them.
Tech due diligence is no longer a post-LOI, checkbox "IT audit." In the AI era, it has "shifted left" to become a critical, pre-LOI analysis of a company's strategic defensibility, AI maturity, and ability to innovate, often starting with outside-in signal gathering.
The new benchmark for engineering maturity is "agentic development." This isn't just auto-complete; it's a full workflow where AI agents write code, open pull requests, and perform reviews overnight, guided by senior engineers who act as mentors to the "smart but inexperienced" AI.
A company's career page is a crucial source of truth during due diligence. The technologies listed in job postings reveal the actual tech stack. This can expose a major disconnect between an investor's thesis (e.g., modern, AI-native) and the on-the-ground reality (e.g., hiring for legacy Delphi developers).