The Incumbent’s Dilemma in the Age of AI

Why software incumbents can’t postpone transformation anymore.

‍

Anthropic’s release of new AI tools, specifically Claude Cowork, caused a significant market correction in software stocks. Investors worry that AI will undercut traditional software business models and compress their future growth and margins. This is driven by the following

• Fear of disruption: AI agents, AI‑native tools or in‑house AI systems could replace existing SaaS products
• Pricing pressure: if AI makes it easier to build and run software, there would be downward pressure on subscriptions and revenue per customer would come down for established vendors
• Reduced competitive advantage: AI lowers barriers for new entrants and for customers building internal tools, which would erode the moat for incumbent SaaS vendors

As a result, valuation reset is happening for many companies. The market is already pricing the incumbent software companies in anticipation of AI disruption. It is a forward looking judgment, not a post-mortem i.e. the market is not saying today’s software companies are broken, but it is signaling the future business models look far less secure.

The question is how do incumbent software companies actually respond? Valuation resets can happen overnight, but the business transformation cannot.

Why AI transformation is hard

For most incumbents, this transformation will be structurally hard for a multitude of reasons.

Legacy tech debt, product complexity, and data silos

SaaS products typically have many layers of services, databases, permissions models, and integrations that bind these. These layers are often built over the years, over different architectural cycles, through integration of 3rd party services and sometimes through acquisition. By contrast, the pre-SaaS legacy software stacks are monolithic and lack modular, modern architecture, which makes them even more brittle today.

AI systems (especially agents) require clean and coherent data models. But incumbent companies usually have fragmented schemas, duplicated records and data silos that were not designed for consumption by AI models.

Fixing this requires foundational re-architecture that includes data unification, workflow re-design, and may need rewriting some core systems. Incremental changes will be both hard and not yield sufficient value.

KTLO overwhelms innovation (customer and board expectations)

For most (software) companies, Keep The Lights On (KTLO) dominates reality. Boards expect revenue and margin. Customers expect stability. So, uptime, compliance, and customer commitments consume the majority of engineering and product capacity. In such an environment, AI transformation competes directly with current obligations.

Transformation (especially AI) requires some experimentation and iteration, some of which can fail. So, the transformation requires embracing risk, whereas revenue-generating organizations are generally structured to minimize risk. As a result, taking on initiatives that are truly transformative is often hard.

Organizational friction, incentives, and unclear ownership

For AI (especially agents) to provide transformative value, it will have to cut across functional boundaries i.e. product, engineering, data, security, legal, customer success. Then the question becomes

• Who owns an AI agent that spans CRM, support, billing, analytics?
• Who is accountable when it makes a wrong decision?
• Who approves changes?

In addition, the incentives for different functions continue to remain aligned with shipping products, protecting ARR, and hitting quarterly targets. The cross-functional nature of AI workflows can create organizational friction due to incentives being at odds. For example, product teams may push to launch AI features quickly for innovation and roadmap, while engineering teams are incentivized to slow things down to preserve system stability, reliability. Data teams care about governance, quality, and risk reduction, while AI developers want broader data access.

Without explicit executive ownership and incentive realignment, AI becomes everyone’s priority in theory and no one’s responsibility in practice.

AI talent gaps and the new culture

Companies face AI talent gaps because talent is still scarce and expensive. AI development also needs a different mindset. Traditional software engineering optimizes for determinism and repeatability. In contrast, AI systems are probabilistic; development requires experimentation since technology is still at early stages, and it can sometimes be hard to fully define compared to traditional systems. So, AI adoption and transformation requires both sustained investment in talent upskilling and a cultural shift at the organizational level.

Transformation Is No Longer Optional

None of this is easy, but it is also unavoidable. For incumbent software companies, the transformation is not optional anymore. While it will be a struggle, many of these companies are also well-positioned to win. Incumbents possess what AI-native upstarts often lack: deep customer intimacy, decades of domain knowledge, embedded workflows, trust earned over time, and proprietary datasets. These assets are the true moat in an AI world. The challenge is about whether and how they can use it. Those that align organizational incentives with clear executive ownership, modernize their stacks, and treat AI as a workflow transformation can emerge stronger.