At Davos, Anthropic’s CEO suggested that AI could do “most, maybe all” of what software engineers do end-to-end within six to twelve months.

It’s easy to see why this narrative spreads: it compresses a complex socio-technical reality into a single, simplified timeline.

But in practice, that statement collapses a critical distinction: AI can accelerate coding tasks, while software engineering is the discipline of delivering, operating and sustaining systems in production.

Why the 6-12 month timeline is misleading

Most real-world software work is not “writing code.” It includes everything required to make systems reliable at scale:

• defining constraints and requirements.

• designing architecture that can evolve.

• securing identity, data and access boundaries.

• building tests and deployment guardrails.

• operating systems with observability and incident response.

• controlling cost under cloud economics.

• proving compliance and auditability.

These are not “extras.” They are core engineering responsibilities that separate demos from production systems. This is also why DevOps research consistently frames delivery performance as a combination of technical practices and human systems, not as the output of a single tool that “automates everything.”

The real bottleneck: verification, governance and trust

As AI generates more code, the verification surface grows. Production-ready teams must be able to answer, consistently and quickly:

• Are changes correct under real traffic?

• Are they secure under your threat model?

• Do they comply with industry and regulatory constraints?

• Are they observable and debuggable?

• Can you roll back safely?

• Can you explain decisions to auditors and stakeholders?

Developer sentiment reflects a persistent trust gap. For example, the 2025 Stack Overflow Developer Survey reports more developers actively distrust the accuracy of AI tools than trust it.

And in some contexts, AI-assisted development can reduce net productivity when review, correction and integration overhead dominates. A METR randomized controlled trial found experienced open-source developers took longer with AI tools in that setting.

The implication is straightforward: the work does not disappear, but moves, with more value shifting upstream (intent, architecture, constraints) and downstream (verification, governance, operations).

What we expect to be true in 2026

Over the next 12 months, we expect four outcomes to become increasingly clear:

1. More code will be generated by AI.

2. Engineering roles will shift toward higher-leverage responsibilities such as architecture, reliability, security and governance.

3. Teams with strong delivery systems will compound gains because they can validate and ship changes safely.

4. Teams without operational foundations will accumulate new forms of technical and organizational debt, faster than before.

This is not a reason to slow down. It’s a reason to invest in production discipline.

What leaders should do instead of debating replacement

If the question is “will AI replace engineers?”, the leadership translation is more practical: are we building production systems where AI-generated changes can be trusted, governed and reversed safely?

If not, the most impactful investment is not a larger model but stronger guardrails:

• testing strategy and quality gates.

• observability and SLO-driven feedback loops.

• policy-as-code and access controls.

• deployment controls (progressive delivery, rollback strategies).

• clear ownership and approval paths.

If yes, AI becomes a force multiplier. Engineers spend less time on repetitive execution and more time on resilience, system quality and business outcomes.

A pragmatic path we see working repeatedly is: observability first → automation second → AI/agents third. Without the first two layers, agents amplify volatility. With them, agents amplify operational intelligence.

Closing perspective

The most useful way to read bold predictions is not as prophecy, but as pressure. AI won’t erase software engineering in months. But it will make production readiness, governance and operational maturity the differentiators between teams that scale and teams that scramble.

How is your organization adapting: treating AI as a coding shortcut, or as a production capability that requires discipline and oversight? If you don’t already know the answer, talk to one of our experts.