The Great Software Shakeout , Who Wins When AI Writes the Code?

You can hear the same pitch every day in the SoMa neighborhood of San Francisco: “Our agent automates your entire workflow.” The demo functions flawlessly. Instant responses to Slack messages. reports that are produced in a matter of seconds. Impressed, investors nod. The product then silently fails when it encounters a chaotic real-world dataset for the first time.

AI software seems to be exploding at a Cambrian pace right now. Every month, hundreds of new businesses appear that promise to revolutionize the nature of labor. Founders speak with the urgency of those who think the future is coming sooner rather than later, venture financing is flowing quickly, and demos are polished. But underneath the optimism lies a sense of unease. It’s the start of the Great Software Shakeout.

Generative AI hasn’t just made software better. It has caused instability in the SaaS economy. Software firms prospered for 20 years thanks to layered integrations, gradual feature releases, and dependable subscription arrangements. AI agents now promise to create conversational interfaces by collapsing entire stacks. It’s probable that a large portion of the SaaS ecosystem of today was founded on presumptions that are now incorrect.

Software firms will not be protected by incumbency alone, according to a new survey conducted by RBC Capital Markets. Survival will be determined by innovation rather than market share. Uncomfortable antecedents can be found in history. Retail used to be controlled by Sears. Video rentals were owned by Blockbuster. Both undervalued turning points in technology.

You can see the craziness if you scroll through LinkedIn. Startups are referring to their platforms as “agentic AI.” Autonomous bots. Products that claim to take the place of marketing coordinators, customer service representatives, and junior analysts.

Many of these agents are still fragile, though. They struggle with complexity but do well in controlled demos. Behind the scenes, what appears to be autonomous frequently needs ongoing human adjustment. It seems as though we are living in the Autonomy Illusion age.

The majority of AI firms today use playbooks made for early SaaS and mobile apps to pursue high-margin growth. AI agents, however, are not the same. They are based on probability. They experience hallucinations. They need supervision. They are more difficult to scale than cloud storage. The shakeout will take place there.

Companies like Microsoft seem to be in a good position, according to RBC. Azure, Office, Teams, Dynamics, and LinkedIn are all impacted by its AI plan. Few can match its depth thanks to its in-house development and partnership with OpenAI. The extent to which AI is now integrated into Microsoft’s main products may still be underestimated by investors.

Another understated competitor is Intuit. It made significant investments in machine learning well before the current craze. The addition of new AI agents to TurboTax and QuickBooks suggests that a business is changing gradually rather than rushing to keep up.

HubSpot appears to be well-positioned to benefit from automation in mid-market companies because to its integrated CRM stack and AI-native products like ChatSpot and Breeze. Less ostentatious but fundamental, MongoDB offers advantages as AI-native applications demand adaptable unstructured data management.

Often disregarded, pegasystems stand to benefit from increased enterprise complexity. Its Blueprint workflow builder makes it possible for businesses to use natural language to modernize processes—a small but significant change. Not everyone appears to be as ready.

Although Salesforce’s Agentforce effort has garnered notice, analysts believe that a large portion of it is still in pilot stage. There is a significant difference between automation and actual agentic autonomy. When promises exceed the maturity of the product, confidence may be damaged.

The challenge for ZoomInfo is more existential. Large language models run the risk of becoming commoditized if they are able to produce contact intelligence on their own. RBC compared their change to the Yellow Pages trying to compete with Google. It seems like a harsh contrast. However, history hasn’t always been kind.

The atmosphere among AI founders was tense yet electrifying when I recently walked through a co-working facility in New York. The funding rounds are coming to an end. Burn rates are high, though. Infrastructure costs are still high, especially when inferring big language models. Strategies for monetization are currently in the experimental stage. Which company models will stabilize is still up in the air.

Those with the most viral demos might not be the most likely winners. They might be businesses incorporating AI into long-lasting ecosystems that already have distribution, data, and trust. The benefits of infrastructure suppliers might be more predictable than those of the agents who interact with customers.

A more significant cultural change is also taking place. AI coding tools are being used by developers more and more. Auto-generated snippets are used by junior engineers. In a matter of hours, entire prototypes can be scaffolded. The value moves when AI writes the code.

It turns toward orchestration, data quality, system design, and supervision. Businesses that adjust to this value redistribution might prosper. Those who continue to use outdated pricing schemes can find it difficult.

It will take time for the Great Software Shakeout to occur. However, the early tremors are evident in strategic pivots, startup closures, and market volatility.

As I watch this develop, it seems more like an evolutionary filter than a solitary disturbance. Numerous species will appear. Most will disappear.