4 Minutes
The narrative was almost too clean to question: AI rises, old software falls. Investors bought into it fast. Then reality started to interrupt.
Brad Lightcap, OpenAI’s COO, isn’t buying the idea that legacy software companies are drifting into irrelevance. If anything, he sees the opposite happening behind the scenes. The companies many assumed would lag are moving with surprising urgency, reshaping their products and rethinking how they serve customers in an AI-first world.
Speaking on the Uncapped podcast, Lightcap pushed back on the recent market panic that wiped billions off software stocks. The assumption driving that sell-off was simple: if AI can build tools, why rely on traditional software vendors? But that logic, he suggests, misses how aggressively those vendors are adapting.
“They’re moving as fast as startups,” he said, pointing to deep customer relationships as an advantage newer players simply do not have.
What the market got wrong
The downturn, dubbed by some as a “software apocalypse,” picked up speed in February after new AI tools demonstrated the ability to automate legal and clerical work. Investors feared a domino effect. If AI can replace workflows, maybe it can replace the software behind them too.
That fear hit hard. Companies like Salesforce, Snowflake, and Microsoft saw shares drop between 24% and 30% this year. The underlying concern was that businesses might skip vendors altogether and build custom tools using AI.
But inside those same companies, a different story is unfolding. According to Lightcap, they are not just adding AI features. They are redesigning entire user journeys, exploring new markets, and integrating automation at a fundamental level.
The real shift is not replacement. It is reinvention.
AI doesn’t remove complexity. It multiplies it
Dan Rogers, CEO of Asana, sees the same pattern from a different angle. His company was among those hit hardest during the sell-off, yet he argues AI actually increases the need for structured software.
Why? Because coordination becomes harder, not easier.
When humans and thousands of AI agents start working together, someone or something still has to organize tasks, track progress, and manage workflows. That layer does not disappear. It expands.
“The coordination problem grows exponentially,” Rogers noted, suggesting that AI makes platforms like Asana more essential, not less.
This idea cuts against the popular narrative that AI simplifies everything. In practice, it often introduces new layers of complexity that demand better systems, not fewer.
The limits of building everything with AI
There is also a practical question: even if companies can build their own tools with AI, should they?
Anish Acharya of Andreessen Horowitz is skeptical. In his view, using advanced AI models to recreate standard business software like payroll or ERP systems delivers limited returns. The savings, he estimates, might hover around 10%.
That raises a bigger point. AI is a powerful resource, but it is not always efficient to deploy it on problems that are already solved.
“Why aim this level of innovation at rebuilding existing systems?” he asked, hinting that the real opportunities lie elsewhere.
A contrarian bet that’s gaining traction
Nvidia CEO Jensen Huang has also dismissed the idea that software companies are on the way out. The belief that AI will replace software tools entirely, he said, does not hold up under scrutiny.
Instead, AI will rely on those tools. It will use them, extend them, and plug into them. Not erase them.
Lightcap’s takeaway lands in a similar place, but with an investor’s twist. If you are optimistic about AI startups, it may also make sense to stay optimistic about legacy software companies. Not in spite of AI, but because of it.
The race is not AI versus software. It is everyone running at full speed, trying to figure out what software looks like when intelligence is built in from the start.
Source: businessinsider
Comments
Tomas
Seen this at my last job, we layered AI on top of existing tools and the coordination chaos got worse, not better. So yeah platforms that organize it matter. still messy tho
coreflux
wait, so legacy vendors are sprinting to catch up? sounds plausible but i keep thinking, how many will reallly pull it off vs just patching old stuff. market panic felt premature, maybe but curious.
Leave a Comment