Strategic Framework
December 2025 · Vincent Verdet

Re-Use Before Build:
How AI Is Rewriting the EA Playbook

The Future of Build vs Buy in Enterprise Architecture

7 min read

Introduction

For decades, Enterprise Architects have relied on a simple hierarchy: Re-Use before Buy, Buy before Build. This principle has guided countless IT investment decisions and saved organisations from the risks and costs of unnecessary custom development. But what happens when building becomes as fast and cheap as buying?

AI-assisted development is fundamentally changing the economics of software creation. What once required months of development and teams of specialists can now be prototyped in hours and refined in days. This isn't a marginal improvement—it's a paradigm shift that forces us to reconsider one of Enterprise Architecture's most sacred principles.

As someone who has spent the last six months actively working with AI coding tools, I've witnessed this transformation firsthand. Projects that would have been automatic "Buy" decisions a year ago are now viable "Build" candidates. The question is no longer whether AI will change the Build vs Buy calculus, but how quickly organisations will adapt their decision frameworks.

Key Takeaway

We're moving toward a "Re-Use Before Build" EA principle where organisations leverage existing assets and AI to create on-demand solutions, potentially bypassing traditional software procurement for an expanding range of use cases.

The Traditional Principle

Before examining what's changing, let's understand why "Re-Use, Buy, Build" became the dominant framework in Enterprise Architecture.

1
Re-Use First

Leverage existing enterprise assets—APIs, services, components, or entire applications. This maximises prior investments, reduces integration complexity, and ensures consistency across the technology landscape. Re-use is nearly always the lowest-risk, lowest-cost option.

2
Buy Second

When no existing asset fits the need, evaluate commercial off-the-shelf (COTS) solutions. Vendors have already invested in development, testing, and support. The shared cost model means enterprise-grade functionality at a fraction of custom development costs.

3
Build Last

Custom development is the option of last resort. Reserved for truly unique requirements where no existing solution provides competitive differentiation. Building carries the highest cost, longest timeline, greatest risk, and ongoing maintenance burden.

Why This Hierarchy Made Sense

Cost Economics

Custom development required expensive specialists working for months. Buying amortised those costs across thousands of customers.

Risk Profile

Build projects frequently overran budgets and timelines. Commercial products offered predictable pricing and proven functionality.

Quality Assurance

Vendors invested in testing, security, and compliance that few internal teams could match.

Maintenance Burden

Custom software required ongoing support, updates, and technical debt management indefinitely.

This framework served organisations well for decades. But every assumption it rests upon is now being challenged.

The AI Disruption

AI-assisted development doesn't just make coding faster—it fundamentally restructures the economics of software creation.

What's Actually Changing

Development Speed

Before AI

Weeks to months for functional prototypes

With AI

Hours to days for working applications

Skill Barrier

Before AI

Required specialised developers for each technology

With AI

General technical competence enables broader creation

Iteration Cost

Before AI

Change requests meant significant rework

With AI

Rapid iteration through conversational refinement

The New Economics of Build

Consider a typical departmental tool—perhaps a workflow tracker, a reporting dashboard, or a data validation utility. Under the traditional model, the analysis would look something like this:

$ Traditional Build vs Buy Analysis

Buy Option

$10,000-50,000 annual licensing. 3-6 month implementation. 70% feature fit. Vendor dependency for customisation.

Build Option (Traditional)

$100,000+ development cost. 6-12 month timeline. 100% feature fit. Ongoing maintenance burden. High failure risk.

Obvious Decision

Buy. The economics aren't even close. Accept the 70% fit and work around the gaps.

AI AI-Era Build vs Buy Analysis

Buy Option

Same as above: $10,000-50,000 annual. 3-6 month implementation. 70% feature fit. Vendor lock-in.

Build Option (AI-Assisted)

$5,000-20,000 development cost. 2-4 week timeline. 100% feature fit. Iterative refinement. Internal ownership.

New Decision

Build becomes viable. Perfect fit, faster delivery, lower long-term cost, no vendor dependency.

This isn't hypothetical. It's happening now, across industries, as organisations discover that the "Build" option has fundamentally changed.

Toward Re-Use Before Build

If AI makes building cheaper and faster, what happens to the traditional hierarchy? I believe we're moving toward a new model: Re-Use Before Build.

1
Re-Use: Foundation Layer

Leverage existing enterprise assets—APIs, data sources, authentication systems, design systems. These become the building blocks that AI-generated solutions consume and extend. Re-use remains essential, but its role shifts from "complete solution" to "enabling infrastructure."

2
Build: AI-Assisted Creation

Use AI to rapidly construct solutions on top of the re-use foundation. Generate applications, integrations, and workflows that perfectly fit requirements. The "Build" step becomes a form of sophisticated assembly rather than ground-up construction.

3
Buy: Strategic Platforms Only

Reserve purchasing for truly complex enterprise systems—ERP, CRM, core banking. Systems where the vendor's domain expertise, compliance investment, and ecosystem integration genuinely exceed what AI-assisted development can achieve.

Software on Demand

The logical extension of this paradigm is "Software on Demand"—the ability to generate fit-for-purpose applications when and where they're needed, rather than procuring and configuring generic solutions.

The Vision

Imagine a business user describing a need: "I need a tool that pulls data from our inventory system, cross-references it with supplier lead times, and alerts me when we risk stockouts." In the traditional model, this triggers a procurement process, vendor evaluation, contract negotiation, and implementation project. In the Re-Use Before Build model, an AI assistant generates a working prototype within hours, iterating until the solution fits perfectly.

This isn't science fiction. The foundational capabilities exist today. What's needed is the organisational readiness to embrace this approach.

When Buy Still Wins

Let's be clear: this isn't a "Build everything" manifesto. Commercial software remains the right choice for many scenarios. The question is where the boundary lies—and that boundary is shifting.

Buy Remains Optimal For:

1 Complex Enterprise Systems

Examples

ERP systems, core banking platforms, supply chain management, healthcare information systems.

Why AI Can't Replace

Decades of domain expertise, regulatory compliance, complex integrations, and proven scale that no AI can replicate in reasonable timeframes.

AI's Role

Integration layers, custom extensions, reporting, and workflow automation around purchased platforms.

2 Regulated Industries

Examples

Financial services compliance, pharmaceutical validation, medical device software, aviation systems.

Why AI Can't Replace

Certification requirements, audit trails, and liability frameworks that require vendor accountability and proven compliance track records.

AI's Role

Documentation generation, compliance checking, test automation—augmenting rather than replacing certified systems.

3 Ecosystem Dependencies

Examples

CRM systems with marketplace integrations, collaboration platforms, industry-specific networks.

Why AI Can't Replace

Value derives from network effects, partner ecosystems, and marketplace integrations that custom solutions cannot replicate.

AI's Role

Custom integrations, data synchronisation, bespoke workflows connecting platform capabilities.

The Expanding "Build" Zone

What's changing is the scope of viable Build candidates. Applications that were previously automatic "Buy" decisions now deserve fresh evaluation:

Departmental Tools

Workflow trackers, approval systems, scheduling tools, internal portals—often better served by custom solutions that fit exactly.

Integration Layers

Data synchronisation, API orchestration, event-driven workflows—AI excels at connecting existing systems.

Reporting & Analytics

Custom dashboards, data visualisations, automated reports—rapidly generated from existing data sources.

Process Automation

Validation workflows, approval chains, notification systems—bespoke logic without RPA platform overhead.

Impact on IT Organisations

The shift toward Re-Use Before Build has profound implications for how IT organisations are structured, staffed, and governed.

Role Evolution

Traditional IT

Developers

Code writers, implementation specialists

Architects

Solution designers, technology selectors

Transitional

Developers

AI collaborators, code reviewers, integration specialists

Architects

Platform enablers, governance designers

Future State

Orchestrators

Solution specifiers, AI directors, quality governors

Enablers

Platform builders, re-use curators, capability providers

New Capabilities Required

  1. Prompt Engineering: The ability to translate business requirements into effective AI instructions becomes a core skill. Clear specification leads to better generated code.
  2. AI Output Validation: Teams must develop expertise in reviewing, testing, and securing AI-generated code. This requires understanding code without necessarily writing it.
  3. Platform Thinking: Re-use becomes more valuable when existing assets are designed for AI consumption. APIs, data services, and components need to be AI-ready.
  4. Rapid Iteration: The value of AI development comes from speed. Organisations need processes that support quick cycles, not waterfall governance.

Governance Implications

Critical Consideration

More building means more governance. Organisations must establish frameworks for AI-generated code review, security scanning, and quality assurance before scaling AI development. The speed of creation demands automated validation pipelines.

Code Review Policies

Who reviews AI-generated code? What level of scrutiny is required? How do we ensure security and compliance?

Ownership & Maintenance

Who owns AI-generated applications? How do we prevent proliferation of unmaintained solutions?

Quality Standards

What testing is required before deployment? How do we validate AI code meets enterprise standards?

Intellectual Property

How do we handle IP considerations for AI-generated code? What are the licensing implications?

Implementation Guidance

Phase 1: Foundation 0–6 months

  1. Audit Re-Use Assets: Catalogue existing APIs, services, and components. Identify gaps in discoverability and documentation that would hinder AI-assisted development.
  2. Pilot AI Development: Select low-risk use cases for AI-assisted building. Departmental tools, internal utilities, and proof-of-concepts are ideal starting points.
  3. Establish Review Processes: Create lightweight governance for AI-generated code. Security scanning, code review checklists, and approval workflows should be in place before scaling.
  4. Train Key Personnel: Develop prompt engineering and AI collaboration skills in your most capable developers. These early adopters become champions and mentors.

Phase 2: Expansion 6–18 months

  1. Revise Decision Frameworks: Update your Build vs Buy evaluation criteria to account for AI-assisted development economics. Create decision trees that incorporate new cost/time factors.
  2. Enhance Re-Use Layer: Invest in making existing assets more AI-consumable. API documentation, example code, and integration patterns accelerate AI development.
  3. Expand Practitioner Base: Move beyond early adopters to broader AI development capability. Consider citizen development programmes with appropriate guardrails.
  4. Measure and Iterate: Track outcomes from AI-built vs purchased solutions. Use data to refine decision frameworks and identify optimal use cases.

Phase 3: Transformation 18+ months

  1. Platform Evolution: Transform your re-use layer into a true "AI-ready" platform. Self-service capabilities, automated testing, and deployment pipelines enable scale.
  2. Vendor Strategy Revision: Renegotiate vendor relationships with new leverage. Some solutions may be candidates for replacement; others become more strategic as platforms.
  3. Organisation Restructure: Evolve IT organisation structure to reflect new operating model. Platform teams, AI enablement functions, and governance centres of excellence.
  4. Continuous Optimisation: Establish feedback loops between AI development outcomes and platform capabilities. What gets built should inform what gets re-used.

The Path Forward

The Re-Use Before Build paradigm isn't about abandoning procurement—it's about having more options. Organisations that master AI-assisted development gain agility, reduce vendor dependency, and create perfectly-fit solutions faster than traditional approaches allow. The question isn't whether to adopt this model, but how quickly you can build the capabilities to execute it effectively.

Comments

Share your thoughts on how AI is changing your Build vs Buy decisions.