Buy
The capability is standard, stable, broadly shared, well served by the market, and does not encode how the company wins.

Build vs. Buy Enterprise AI · A CIO–CFO Decision Framework
AI has changed the cost of creating software, but it has not eliminated the responsibilities of owning it. Buy commodity capabilities. Build the operating advantage. Compose the two around the workflow.
FIELDRUNTIME / BUILD VS. BUY
EXECUTIVE POSITION
The decision is no longer simply whether to rent software or recreate an application. It is which capabilities to buy, which workflow intelligence to own, and how to combine them into the smallest system that improves a measurable business outcome.
WHAT CHANGED
The old renewal question had two answers.
Traditional software economics favored renting because building required a large team and a long implementation. AI-assisted engineering changes that boundary. The right unit of comparison is now one end-to-end workflow and the complete system required to operate it safely.
Renew the current product or replace it with another product?
Which capabilities should we buy, configure, compose, build, or retire?
Annual license price versus the cost of producing a convincing prototype.
Three-year cost, measurable value, control, adaptability, reversibility, and ownership of operating intelligence.
FIVE DECISION PATHS
Choose at the workflow level.
An enterprise may use all five paths across its portfolio. The goal is not maximum ownership. It is intentional ownership.
The capability is standard, stable, broadly shared, well served by the market, and does not encode how the company wins.
A commercial system fits the core job and limited configuration can accommodate the organization without creating a fragile exception layer.
Existing systems should remain systems of record, but the workflow, context, decisions, and coordination across them are distinctive.
The workflow contains proprietary judgment, fast-changing rules, unique economics, or learning that should remain an enterprise-owned asset.
The software mainly preserves obsolete steps, duplicated records, manual reconciliation, or work that should disappear in the redesigned operation.
DECISION MATRIX
Start with operating reality, not a platform preference.
| Workflow condition | Path | Why | Evidence required |
|---|---|---|---|
| 01Standard + stable | Buy | The market can spread development, compliance, and maintenance across many customers. | The workflow can adopt the product’s standard process without material workarounds or loss of advantage. |
| 02Standard core + bounded variation | Configure | The product solves the durable core and the variation is limited, visible, and maintainable. | Configuration stays inside supported extension points and survives upgrades without specialist rescue. |
| 03Commodity records + distinctive coordination | Compose | CRM, ERP, ticketing, or data platforms can remain while an enterprise-owned workflow layer carries the work across them. | Interfaces, state, authority, evidence, failure ownership, and replacement boundaries are explicit. |
| 04Proprietary judgment + continuous change | Build | The workflow is a source of operating advantage and improves through company-specific exceptions, corrections, and outcomes. | The organization can own the product, evaluation, security, support, and controlled learning responsibilities. |
| 05Duplicated or obsolete work | Retire | Recreating an unnecessary process—through SaaS or custom software—preserves cost instead of removing it. | The redesigned workflow can eliminate the step without losing a necessary control, record, decision, or customer outcome. |
BUYER EVIDENCE
Architecture and economics must clear the decision together.
THREE-YEAR ECONOMICS
Neither the license nor the initial build is the full cost.
Use the same time horizon, operating volume, risk assumptions, and business baseline for every option. Keep modeled value separate from observed value.
Implementation + platform + integration + operation + review + change + expected failure cost.
Capacity + cycle time + throughput + revenue or margin + cash + risk reduction, adjusted by adoption and confidence.
Value minus TCO, considered alongside control, reversibility, bargaining power, learning, and strategic option value.
WHAT TO OWN
Models and tools are replaceable. Operating intelligence compounds.
Even when the enterprise buys most of the technical stack, these intelligence artifacts should remain portable, inspectable, and controlled by the organization.
How work moves, branches, stops, recovers, and reaches verified completion.
The authoritative knowledge, live operating condition, history, and permissions surrounding the work.
Representative tasks, edge cases, failure thresholds, evidence checks, and regression tests.
What software, AI, and people may read, propose, execute, approve, escalate, and change.
Exceptions, corrections, decisions, actions, and downstream results that make the next run better.
WORKED EXAMPLE
A pricing exception should not require replacing the CRM.
Keep the CRM and ERP as reliable systems of record. Build the distinctive layer that assembles deal context, checks policy and margin, coordinates approvals, preserves evidence, and learns which exceptions produce good downstream outcomes.
Accounts, opportunities, contracts, products, price books, orders, invoices, and financial records.
State, context assembly, policy, exception routing, evidence, approvals, outcome verification, and learning.
Understand the request, retrieve precedent, diagnose the exception, explain tradeoffs, and prepare the decision.
Own the customer relationship, accept material margin or policy risk, decide novel exceptions, and govern change.
EXECUTIVE CHECKLIST
Before the renewal, platform purchase, or custom build.
Name the judgment, coordination, customer promise, economics, or operating knowledge that competitors cannot obtain from the same product.
Include workarounds, duplicate entry, handoffs, reconciliation, exceptions, delayed decisions, and lost operating visibility.
Separate replaceable infrastructure from context, procedures, evaluations, authority, evidence, and learning that should survive any vendor.
Assign production ownership, security, evaluation, support, incident response, change approval, and budget before approving the build.
Define modular boundaries, export paths, interfaces, rollback, replacement cost, and the smallest first commitment.
Run a bounded workflow test with real tasks and a finance-approved baseline before making a platform-wide commitment.
EXECUTIVE QUESTIONS
Before committing architecture and capital.
Sometimes, for a bounded and consequential workflow. AI-assisted engineering can reduce implementation effort, but production cost still includes discovery, integration, data, evaluation, security, human review, support, maintenance, and recovery. Compare fully loaded three-year economics, not a prototype build against a SaaS list price.
Buy when the capability is standard, stable, not strategically differentiating, well served by the market, and can be adopted without material workarounds, lock-in, or loss of control over critical operating knowledge.
Build when the workflow contains proprietary judgment, unique coordination, changing rules, material economics, or outcome-linked learning that creates advantage—and when the organization is prepared to own production operation and governance.
Compose means retaining commercial systems for commodity capabilities or records while building an enterprise-owned workflow layer across them. That layer preserves state, applies context, coordinates tools, defines authority, verifies outcomes, and learns from real work.
Not by default. Many enterprises should initially keep reliable systems of record and use agents inside a governed workflow operating system that crosses them. Individual applications can be replaced later when evidence shows that doing so improves economics, control, or adaptability.
Include implementation, licenses, infrastructure, integration, data, evaluations, human review, exceptions, controls, support, change, failure exposure, switching cost, and the value of strategic flexibility. Keep modeled benefits separate from observed results.
The decision should be shared. The workflow owner establishes the business outcome, the CIO owns architecture and operability, the CFO challenges economics and value evidence, security and risk leaders set controls, and operators validate how the work actually behaves.
Facing a software renewal or AI platform decision?
Bring one workflow, the current operating cost, and the decision in front of you. We will map what to buy, configure, compose, build, or retire—and define the evidence needed for a defensible CIO–CFO decision.