The Shape of Things to Come

The focus among leaders today seem to be "how do we implement AI in our existing structure?" The answer might be "we don't."

TL;DR

• Small-to-mid businesses with life-or-death stakes are higher quality datapoints than Enterprise headline posturing and think tank projections.
• Roles and workflows are reshaped quietly in teams that ship products.
• AI collapses marginal coding cost. Lower execution cost increases total software production. The constraint is no longer coding capacity.
• The limiting factor is human decision-making, system design, and orchestration, not the hours humans spend typing. The AI agent acts as a magnifier for human intent.
• Bespoke systems become economically viable in more domains. Maintenance can be wholly or partially automated, weakening historical anti-bespoke arguments.
• The developer career ladder compresses and mutates. Middle-tier engineering roles are most exposed.
• UX and QA consolidate and become strategic. PO becomes more central.
• Integration and open standards replace vendor-driven standardization.
• The core transformation is organizational, not technical.
• The firms that redesign roles around abundant execution will outperform those that merely add AI tools to legacy structures.

Human vs AI: A Time and Leverage Perspective

• In practice, a human software developer in Sweden (and similar Western markets) spends roughly 5% of their calendar time writing code that directly generates value.
• By contrast, a single AI agent can operate 24/7/365, often delivering at 10x the speed of a human and at a fraction of the cost. It executes tasks without the overhead of context switching or downtime.

This amounts to two magnitudes decrease of implementation time and three magnitudes decrease of cost for the parts of development done by AI, if the organization is positioned to leverage 24/7/365 agentic development.

This transforms the fundamental economics of execution:

• Human developers are leveraged far more efficiently when AI handles the bulk of boilerplate and repetitive tasks.
• A single human can now supervise multiple AI-driven “builders” instead of writing code directly.
• The ratio of POs to builders compresses because execution scales multiplicatively via AI rather than linearly via human effort.

This Is an Organizational Shift, Not a Technical Upgrade

• The primary constraint in software is moving from execution capacity to decision quality.
• Coding throughput is no longer the dominant bottleneck.
• Alignment, prioritization, architectural coherence and governance are becoming the limiting factors.
• Teams structured for scarce coding capacity are misaligned for abundant coding capacity.
• The transformation pressure is structural, not tooling-driven.

If the above is correct, then role composition must change.

The Cost of Programming Is Collapsing

• AI-assisted development reduces feature delivery time by multiples, not percentages.
• Boilerplate, scaffolding, tests and refactors are increasingly automated.
• Runtime debugging, profiling and incident analysis are already AI-assisted and improving quickly.
• Continuous security scanning and CVE patching can be agent-driven.
• Architecture conformance checks can be automated.

If the marginal cost of implementation drops sharply:

• Total software volume increases.
• Internal tooling becomes economically rational.
• Bespoke systems compete with commercial SaaS in more domains.

This follows the logic observed by William Stanley Jevons. Efficiency increases consumption.

Maintenance Is No longer a Counterargument

The common objection to bespoke or in-house development is the assumption that it will bring increased maintenance overhead. That assumption may be outdated.

• AI agents can continuously scan dependencies and patch vulnerabilities in real-time.
• AI can refactor legacy modules safely under supervision.
• AI can monitor architecture drift and enforce standards.
• AI can simulate regression testing at scale.

Maintenance may become:

• More in-house than outsourced to vendors.
• Faster than traditional subscription patch cycles.
• Potentially more secure if monitored continuously.

This undermines the argument that bespoke systems are inherently more expensive long term.

The Bespoke vs Off-the-Shelf Equation Is Shifting

Historically:

• Buy if not core, Build if differentiating.

Now:

• Build becomes cheaper in many internal and semi-core domains.
• SaaS subscription stacking becomes relatively more expensive.
• AI enables lean domain-specific systems with minimal excess features.

Where bespoke now wins:

• Internal workflow automation.
• Data pipelines.
• Niche B2B services.
• Rapidly evolving product ideas.

Where packaged solutions still win:

• Highly regulated functions.
• Commodity systems such as payroll and accounting.
• Domains governed by strict liability and compliance.

Small and mid-sized companies benefit disproportionately.

• AI tooling democratizes capabilities previously limited to large enterprises.
• Smaller firms face less legacy friction.
• Adoption speed becomes strategic advantage.

This dynamic mirrors elements of The Innovator's Dilemma: Incumbents are disrupted because they are structurally constrained by their own operating models.

The Developer Career Ladder Compresses and Shifts to AI Fluency

• AI impacts skill bands unevenly.
• Pure boilerplate work is automated first.
• Mid-level engineers who historically added value through speed and familiarity are the most exposed.
• Senior engineers who design systems, manage risk, and orchestrate AI gain leverage.

Traditional ladder:

• Junior writes repetitive code, learns patterns by doing.
• Mid-level increases speed, autonomy, and breadth of execution.
• Senior designs architecture after years of implementation experience.

AI-native ladder:

• Repetition is no longer the primary learning mechanism.
• Pattern exposure comes through AI-assisted exploration.
• Debugging, profiling, and runtime analysis are AI-augmented from day one.
• Career progression is determined by ability to orchestrate AI, reason about architecture, and integrate systems, not raw coding throughput.
• Middle-tier roles compress; fewer positions justified solely by execution capacity.
• Transition from assisted contributor to system-level thinker is faster.
• Emphasis on architectural reasoning, integration, and AI governance occurs earlier in a career.

Net effect:

• The role of developer does not disappear.
• The definition shifts from writing code to designing and governing systems that produce code.
• AI fluency, not years of coding, becomes the primary measure of skill and career advancement.

Role Reshuffling Is the Core Impact

If execution cost collapses, then:

• Fewer execution-heavy engineers per product.
• More emphasis on product clarity.
• More emphasis on architectural coherence.
• More emphasis on integration standards.

Likely structural outcome per product:

• 1 product strategist with strong technical fluency.
• 1 to 3 AI-augmented engineers.
• UX centralized or fractional across multiple teams.
• QA strategic and automation-focused rather than embedded per team.
• Strong shared platform and integration layer.

The historic 8 developers per PO model becomes economically questionable.

If execution is no longer scarce, strategy must carry more weight.

UX and QA Role Definitions Refine

UX

• Execution tasks such as mockups and layout generation are AI-assisted.
• Focus shifts to research, differentiation, and service design.
• Likely 1 senior UX now serves multiple teams instead of 1 per team.
• The role is refined, emphasizing impact over output volume.

QA

• Test generation, regression, and security scanning are AI-native.
• Human QA focuses on edge-case modeling, risk evaluation, and AI validation.
• Embedded QA per team is optional in low-risk domains.
• The role is refined, moving from execution to strategic oversight.

AI reshapes rather than remove these functions. Responsibilities are elevated, execution is automated, and skill focus is refined.

Integration and Open Standards Become the New Standardization

If bespoke systems multiply:

• Vendor-driven standardization weakens.
• Open APIs and interoperability become dominant coordination mechanisms.
• Internal integration layers gain strategic importance.

Standardization shifts from product vendors to protocol and integration layers.

Companies that master integration outperform those that standardize on monolithic SaaS stacks.

The Innovator’s Dilemma Cuts Both Ways

Enterprises are slow to adopt due to:

• Governance.
• Risk models.
• Business model inertia.

However:

• Once adoption begins, scale advantages can amplify AI leverage.
• Proprietary data becomes multiplier fuel.
• Capital enables large-scale internal transformation.

Small-to-mid firms adopt, test, and refine AI workflows first. Large companies follow later, scaling the changes but rarely innovating them.

Not a Question of Headcount

The interesting tension is:

• Does productivity grow faster than product proliferation?

That determines absolute developer demand.

But the more important shift is qualitative:

• Execution-heavy roles shrink.
• Orchestration-heavy roles expand.
• Middle-tier compression increases.
• Strategy and system thinking become scarce assets.

If this model holds, the winning software organization will be smaller per product, more strategic per decision, and heavily automated in execution and maintenance.