Between SaaS, open source, and custom software: Why all three will survive the AI hype

Software leaders are hearing a new story. If AI can generate entire applications, why keep paying for SaaS, why invest in open source, and why commission custom software at all. The promise sounds simple. Ask an assistant for what you need, and it will materialize as code, ready to run and replace half your vendor list overnight.

In reality, the organizations that have already gone through cloud migrations, DevOps transformations, and compliance audits know that software does not live in isolation. It lives on infrastructure, around processes, inside legal obligations, and across many teams who need to understand and operate it over years.

This post looks at the three dominant models of business software today, how AI changes the economics of each, and why a pragmatic mix of SaaS, open source, and custom systems will outlast the current hype cycle.

Three models, three operating realities

In order to identify why SaaS applications, custom software and open source solutions still have merit, we first need to understand the three dimensions that contribute to the software stack of a modern IT workplace:

1. SaaS applications are delivered over the internet from provider-managed environments, usually accessed through a browser or mobile app, with the vendor taking responsibility for hosting, updates, and most of the maintenance. They shine when you need standard capabilities such as email, CRM, collaboration, or accounting, where the problem space is well understood and the main question is adoption rather than invention.
2. Custom software is built specifically around the unique workflows, rules, and integration landscape of a particular organization, either by internal teams or external partners, and the business owns the roadmap, architecture, and long-term evolution of the product. Custom systems excel where differentiation matters, such as proprietary quoting engines, specialized operations platforms, or domain-specific analytics that simply do not fit within a generic product.
3. Open source adds a third dimension. It provides full transparency into code and architecture, often under permissive licenses that allow organizations to self-host, modify, and integrate without vendor lock-in, while drawing on a wider community for improvements and security fixes. For teams that care about infrastructure sovereignty and modularity, open source components often become the building blocks underlying both custom systems and self-hostable alternatives to SaaS.

Why AI does not replace these choices

The argument behind why SaaS is supposedly dead usually rests on one observation. If AI can generate working software much faster than traditional teams, the economic moat of SaaS vendors shrinks because customers can build their own tools instead of renting generic ones. A similar narrative shows up around open source. If AI models are trained on public code, why keep releasing more of it, and why not focus entirely on internal AI-assembled solutions instead.

A useful counterpoint comes from the last major shift in software economics. When open source made entire stacks freely available, from databases to ERP suites, most organizations still decided not to build everything themselves once they looked at the real costs of ownership and maintenance. Open source removed the license price as a barrier, but it did not remove the need for teams, processes, security, and long-term product stewardship, which many businesses preferred to access through vendors or managed solutions.

AI changes how quickly we can prototype and extend software, and it puts pressure on pricing and differentiation, but it does not remove the need for clear ownership of operations, support, and compliance. Instead of wiping out SaaS or custom work, AI is more likely to push both toward more modular architectures, richer APIs, and clearer boundaries between standardized capabilities and business-specific logic.

Operational trade-offs that AI does not erase

A recurring theme in build-versus-buy decisions is time to value. SaaS tends to offer faster initial deployment for standard processes because it arrives as a complete product with onboarding flows, permissions, and integrations already defined, whereas custom software usually needs longer discovery, design, and iterative implementation cycles. AI can compress parts of that custom timeline by generating scaffolding and boilerplate, but it does not automatically provide the surrounding operating model or the organizational change required to adopt it.

Functional fit and ownership form the second axis. SaaS products stay strong for common scenarios but naturally reflect the vendor’s roadmap, which means customers adapt their workflows to the product’s assumptions, while custom solutions are shaped around specific business rules and can evolve with the company’s strategy. AI can help explore options on both sides. It can recommend how to configure an existing SaaS platform and it can accelerate the implementation of bespoke features, yet someone still needs to decide which constraints to accept and which to change.

Maintenance responsibility is the third, and final, point that hype often glosses over. In SaaS, the vendor handles the product’s technical upkeep, so internal teams focus on configuration, data quality, and adoption, while in custom systems the organization remains accountable for security patches, performance tuning, and feature evolution over the entire lifecycle. AI tools can assist with upgrades and refactors, but they do not carry legal responsibility when something breaks during an audit, a security incident, or a regulatory review.

Governance, law, and the real cost of exclusively building with AI

The "I can build anything with AI" mentality becomes risky when it ignores laws, regulations, and the long tail of operational effort. Modern digital transformation roadmaps explicitly call out governance, security, vendor choice, and policy frameworks as central to technology decisions, not as optional afterthoughts. Building more software faster only helps if you can explain how it behaves, how it is controlled, and how it fits into the obligations your organization already carries.

Regulations such as data protection rules, sector-specific compliance regimes, and emerging AI governance expectations thereby all point in the same direction. They require traceability of changes, audit trails, clear accountability, and demonstrable control over who can alter systems and on what basis. Whether you run SaaS, open source, or custom platforms, you still need structured logs, approvals, and incident response practices that bridge engineering, operations, finance, and legal teams.

This is where open source and custom software retain particular strength. Self-hosted stacks on infrastructure you control allow you to embed guardrails that reflect your risk appetite, while pairing them with focused SaaS for areas where specialization and third-party certifications add value. AI coding agents sit inside that governed environment. They generate code, configuration, and documentation, but they operate under policies, blast-radius limits, and human review rather than freehand control over production systems.

A pragmatic mix for modern organizations

For mid-sized B2B organizations, the durable approach is not choosing one model and declaring the others obsolete. It is mapping capabilities to the right level of ownership and differentiation. That means, commoditized processes such as email, generic CRM, or standard HR workflows remain natural fits for SaaS, especially where vendors invest heavily in security, reliability, and integrations, whilst areas that define your competitive advantage, such as specialized planning tools, industry-specific operations platforms, or unique customer portals, often justify custom development, drawing on open source frameworks to avoid reinventing basic plumbing while retaining control over core logic. Supporting infrastructure, observability, and billing can then be assembled from open source and self-hostable SaaS components, keeping critical data and workflows on platforms you can operate on your own infrastructure.

In practice, that mix looks like a governed ecosystem where consulting, products, and internal teams work together. AI coding agents help teams implement and refactor systems more quickly. Billing platforms unify revenue streams across services and subscriptions. Multi-cluster orchestration tools keep the underlying environments predictable. All of this runs with audit trails, policies, and guardrails that make leadership comfortable adopting AI without losing control.

TL;DR

• AI accelerates how we build and adjust software, but it does not remove the trade-offs between SaaS, open source, and custom systems around speed, ownership, and differentiation.
• SaaS continues to make sense for standardized capabilities where vendor scale and maturity matter, while custom software remains essential wherever your workflows and rules are too specific to fit inside a generic product.
• Open source still provides transparency, self-hosting options, and modular building blocks, and AI is more likely to reshape how we contribute to and consume open code than to make it obsolete.
• Legal obligations, security requirements, and governance frameworks mean that the cost of operating software over years remains significant, even if code generation itself becomes cheaper, which keeps all three models relevant.
• The most resilient architecture for modern organizations is a governed mix of SaaS, open source, and custom systems on infrastructure they control, with AI embedded as a well-managed assistant rather than a free-running replacement for their entire stack.