For most of the cloud era, enterprise technology leaders followed a practical rule: buy standard capabilities and reserve custom development for the few processes that truly differentiate the business.
AI-assisted development is changing that calculation.
Across industry events and conversations with technology leaders, Inforedge is seeing organizations reconsider software decisions that appeared settled only a few years ago. Capabilities that would previously have been purchased as part of a large software platform are now being evaluated for internal development, configuration or composition using existing platforms, APIs and AI-assisted tools.
The result is a more complicated decision than build versus buy. Organizations must decide what to buy, what to configure, what to compose from existing services and what deserves custom development.
AI has reduced the cost and time required to create software. It has not removed the responsibilities that begin after the software enters production.
Why the decision has changed
Enterprise buyers have traditionally accepted that established software platforms provide the fastest and safest path to a required capability. That assumption is being tested.
Organizations are now challenging expensive renewals, removing underused modules and evaluating whether focused internal applications can provide a better fit. In some cases, a small development team can create a workflow that aligns more closely with the business than a heavily customized commercial product.
McKinsey’s 2026 global AI survey found that nearly one-third of respondents had decided against buying at least one software product or feature because agentic coding tools allowed them to build it internally.[1]
This does not mean that one-third of enterprise applications are disappearing. It means buyers have a credible alternative when a vendor feature is expensive, limited or poorly matched to the organization’s process.
Inforedge is also hearing greater hesitation about long-term contracts for AI-related products. A traditional three-year software agreement can become restrictive when models, capabilities and pricing structures are changing every few months. Organizations increasingly want shorter commitments, portable data and the ability to change models or providers without rebuilding the entire environment.
The strongest build cases are narrow
AI is making focused applications easier to justify before it makes wholesale platform replacement practical.
Internal approvals, reporting interfaces, document processing, employee assistants, data reconciliation and workflow extensions are examples of capabilities that can be clearly defined and measured. They are easier to build, test and support than an attempt to recreate an entire ERP, CRM or HR platform.
The market is therefore moving toward componentization. Organizations want to change the user experience, model or automation layer without replacing the trusted systems underneath.
A company may retain its financial platform as the system of record while developing an AI-assisted forecasting experience. It may keep its CRM database while building a specialized sales workflow around its own processes. It may continue using an established HR platform while replacing an expensive employee-support module with a focused internal application.
Forrester has described build versus buy as an increasingly incomplete framework. The available choices now include customizing, composing and generating software.[2]
This capability-level approach allows organizations to preserve investments in identity, data, security and transactions while applying custom development only where it creates measurable value.
SaaS renewals are receiving greater scrutiny
The changing economics of software development are also affecting vendor negotiations.
Organizations are looking more carefully at paid add-ons, overlapping applications and modules with limited adoption. A feature that once required a specialized vendor may now be reproducible through an existing cloud platform, data environment or AI service.
This does not necessarily lead to complete platform replacement. More often, it gives the buyer leverage to reduce scope, negotiate pricing or avoid purchasing an additional module.
Long contracts are also becoming harder to justify for rapidly evolving capabilities. Buyers want the flexibility to respond when a better model becomes available, usage costs fall or a capability becomes part of a platform they already own.
Vendors that provide strong APIs, portable data and flexible commercial terms will be better positioned than those that depend primarily on lock-in.
Day two remains the hardest part
The ease of producing a prototype can create a misleading comparison.
A vendor’s subscription price includes more than the visible feature. It may also include upgrades, security controls, documentation, regulatory changes, availability commitments, user support and an ecosystem of trained professionals.
An internally developed application needs those capabilities as well.
Before building, organizations must identify who will own security reviews, access controls, testing, monitoring, incident response, documentation, user support and future enhancements. The first version may be inexpensive, but an unmanaged production dependency is not.
Spotify reported that AI-assisted development significantly increased the volume of software changes produced by its teams. The company did not identify AI-authored code as a material direct cause of reviewed incidents, but it found that verification, testing, rollout and observability controls had to keep pace with the higher volume of change.[3]
AI can move the constraint from writing software to verifying and operating it.
Gartner expects more than 40 percent of agentic AI projects to be canceled by the end of 2027 because of rising costs, unclear business value or inadequate risk controls.[4] The same warning applies to internally generated applications. Technical possibility is not a complete business case.
Data readiness can decide the outcome
Data has become one of the most important factors in the build-versus-buy decision.
Across enterprise conversations, Inforedge continues to hear that organizations struggle with unclear data ownership, conflicting definitions, incomplete classification and information distributed across departmental systems or acquired businesses.
An AI agent cannot reliably automate a process if the organization cannot determine which data is authoritative or what that data means.
A custom application may appear simple at the interface level but still require substantial work to establish data lineage, permissions, quality controls and a semantic layer that translates technical data structures into business meaning.
The business case for building must therefore include the cost of preparing and governing the underlying data. Funding only the visible application can produce an impressive demonstration that is not ready for reliable production use.
Governance must become continuous
Traditional annual technology reviews are poorly matched to AI systems whose models, usage patterns and costs change throughout the year.
Organizations are beginning to move toward continuous assurance. This includes discovering AI usage, enforcing policies, monitoring costs, recording model activity and reassessing material changes to the system.
Model flexibility is another important consideration. Organizations can reduce costs by routing simpler tasks to smaller models, avoiding repeated processing and selecting the most appropriate model for each workload.
Those benefits require an architecture that separates the data, orchestration and model layers. An internal application that depends tightly on one model or provider can recreate the same vendor lock-in it was intended to avoid.
A practical decision framework
The right unit of analysis is the business capability, not the application name. Leaders should evaluate each capability using the following questions:
Does the capability differentiate the business
Processes that encode proprietary knowledge, customer experience or competitive advantage provide a stronger reason to build or compose.
Is the process regulated or operationally critical
Capabilities with frequent regulatory changes, significant audit requirements or demanding availability commitments may be better suited to a mature commercial product.
Can the scope remain narrow
A focused workflow is easier to test, govern, support and eventually replace than a new enterprise platform.
Who will own the application after launch
A build decision requires a named owner for security, monitoring, documentation, support and future development.
Can components be changed independently
Portable data, well-designed APIs and model flexibility reduce lock-in and preserve negotiating leverage.
What is the full five-year cost
The comparison should include licensing, development, implementation, integration, model usage, security, governance, operations and eventual replacement.
When buying remains the better decision
Buying remains attractive when the capability is common across organizations, regulations change frequently, vendor accountability carries real value or the company does not have the capacity to support the application.
Payroll, identity, financial accounting, security tooling and other control-heavy systems often fall into this category.
Mature platforms may also contain years of business rules, permissions, integrations and audit history. Reproducing the visible workflow is easier than reproducing these operational obligations.
Buying is also sensible when speed matters more than differentiation. If an established product satisfies most requirements and the process does not create a meaningful competitive advantage, custom development can become an expensive way to reproduce standard functionality.
When building becomes credible
Building deserves serious consideration when the workflow reflects proprietary knowledge, the required data is available, the scope can remain focused and the business has an accountable owner.
It becomes particularly compelling when the organization is repeatedly paying for a narrow vendor feature that can be created using platforms, models and APIs it already owns.
The build decision should still survive a long-term ownership test. Leaders must account for security, data engineering, testing, model consumption, monitoring, incident response, documentation and user support.
A prototype demonstrates that something can be built. It does not prove that building is the best long-term decision.
Where Inforedge adds value
Inforedge helps organizations evaluate the decision at the capability level.
We begin by mapping the business process, data, integrations, controls, contract position and total cost of ownership. This analysis may support buying a platform, configuring what the organization already owns, composing a workflow across existing systems or building a focused application.
When building is justified, Inforedge can support architecture, implementation, integration, testing, governance and operational handoff.
When buying is the stronger choice, we can assist with product evaluation, implementation planning, platform optimization and adoption.
We also help organizations identify the middle path: retaining a trusted system of record while creating a differentiated experience or automation layer around it.
Our vendor-neutral approach keeps the recommendation tied to the business outcome rather than a predetermined product or development model. It also makes ownership explicit before launch, reducing the risk that a fast prototype becomes an unmanaged production dependency.
The durable rule
Buy the capabilities that should be standard.
Build the capabilities that encode meaningful differentiation.
Compose when the advantage comes from connecting trusted platforms, proprietary data and a workflow unique to the organization.
Whatever the decision, evaluate the full operating model—not only the speed or cost of producing the first demonstration.
Sources
1. McKinsey & Company. The State of AI Global Survey 2026.
2. Forrester. Forget About Build Versus Buy Your Choice Is Customize Compose Or Generate. July 9, 2026.
3. Spotify Engineering. AI Changed How Spotify Builds What We Learned and Fixed About Quality at Higher Velocity. September 16, 2026.
4. Gartner. Gartner Predicts Over 40 Percent of Agentic AI Projects Will Be Canceled by End of 2027. June 25, 2025.