If AI can clone your software in a weekend, what value is left?

Right now, a product manager can have an idea during a client call on Tuesday and a working application by Friday — with authentication, workflows, reporting and integrations tested and deployed by AI agents, and at a cost that makes your $150-per-user-per-month software-as-a-service (SaaS) offering, along with the systems integrator who took six months to implement it, look like artifacts from a different era.

Software delivery is changing as AI transforms how applications are designed, built, tested and modernized, as well as how they’re monitored, supported and continuously improved after release.

And that’s bad news for anyone selling hours, tickets or seats.

The old economics don’t add up anymore

The word “modernization” has been repeated so often in enterprise technology conversations that it has almost lost its meaning. And yet, on banking floors and in insurance back offices and logistics operations around the world, the same 30- and 40-year-old applications continue to run reliably — and entirely beyond the comprehension of most people currently employed to manage them.

The technology industry has failed to address this reality directly. Modernization is typically sold as an all-or-nothing proposition: Rip out the old, replace it with the new and endure whatever risk and disruption lies in between. But if your core operations depend on legacy systems containing tens of millions of lines of code and decades of encoded business logic, that’s a reckless proposition.

AI is also changing the economics of maintaining and progressively updating what you already have. It’s called living with legacy, and it deserves your attention.

Modernization and living with legacy are not the same, and conflating them has cost the industry enormous amounts of time, money and credibility. Modernization is a destination — a reimagined architecture, a cloud-native environment or a decommissioned mainframe; living with legacy is a strategy for the journey ahead. It’s the conversation many organizations should be having right now.

Say goodbye to these three principles of app development

1. The waterfall delivery model: Discovery workshops, sprint zero, quality-assurance cycles measured in quarters: all gone. If going from idea to deployment is now a three-day project for one person using AI agents, the 50-person “digital factory” billing on a time-and-materials basis for boilerplate is indefensible. Clients will ask why they paid for your scrum ceremonies when their intern could ship it with Cursor.
2. The feature-moat SaaS: If your product’s claim to fame is “we have that integration” or “we built that workflow,” well, AI can recreate those capabilities in days. The only moats left are proprietary data, regulatory barriers, network effects, trusted brands and distribution. Everything else is a prompt away from irrelevance, and CFOs are already doing the math.
3. The big, billable team: Entire businesses were built on teams of support engineers handling tickets, watching monitoring dashboards and following runbooks. AI is rapidly automating all three. It detects, triages and, in our experience, remediates 80% of incidents in the blink of an eye, making the hourly billing of skilled resources seem like selling horses after the invention of the car. The value now lies in engineering resilience directly into the platform.

As Carlos Galve, Co-Head of the Global AI Office at NTT DATA, Inc., says: “The ‘SaaS apocalypse’ is unfolding as Anthropic and OpenAI roll out vertical agents tailored for finance, legal and operations, bypassing the SaaS application layer entirely. To survive, SaaS vendors are either retreating to the data layer or risk being relegated to interchangeable backends that agents can seamlessly swap out.”

Faster development demands greater discernment

Speed cuts both ways: The same technology that accelerates delivery also amplifies mistakes.

While AI makes engineering dramatically faster — with requirements, code, tests, documentation and deployment happening in a single, continuous loop — it also makes it far less forgiving.

Poor architecture, weak security, bad product decisions and technical shortcuts don’t disappear behind months of development effort anymore. AI scales and exposes them faster than ever.

Raw code is becoming abundant, but sound judgment still comes at a premium. Your developers must learn how to use AI to enforce sound design, strengthen security, improve quality and focus their effort where it creates real value.

Managed services enter the era of proactive resilience

AI can detect anomalies before they become outages, correlate alerts across systems, pinpoint root causes, generate fixes, test them and deploy them — often before anyone even opens a ticket. Your service desk shrinks to an exception-handling unit, reserved for the small number of problems that are too novel, too ambiguous or too consequential to automate.

Traditional SLAs for managed services are no longer a differentiator. The value has moved upstream to self-healing systems, proactive resilience, continuous cost optimization and outcomes measured in business impact.

“We used to staff teams to watch dashboards,” says Subhashini Desigan, Executive Managing Director: Global Application, Business Process Services and Data (North America) and Global Strategy and Operations Head at NTT DATA, Inc. “Now we deliver AI that fixes the problem before the dashboard turns red. If your managed service isn’t autonomous, it isn’t a service — it’s a liability.”

Who manages what in the new development landscape

Tasks shifting toward AI

AI is taking over repetitive, structured tasks that follow known patterns: code scaffolding, unit tests, CRUD applications, documentation, alert triage, knowledge-base lookups, routine patching and status reporting. If a junior employee could learn to do it, there’s a good chance AI can already do it better and cheaper.

The roles under pressure

The roles most exposed are those built on coordination, repetition and process administration rather than expertise. Routine coders, manual testers, L1 support teams handling common incidents, project managers who only run Jira boards and schedule meetings, and coordinators who forward emails without adding context or judgment — all these intermediaries are in danger of being at least partly replaced by AI.

Where the value goes

The most valuable people in technology are now the ones making the most important decisions:

• Solution architects who can prompt entire systems into existence
• Domain-led engineers who know what not to build
• AI delivery controllers who prevent hallucinations and compliance failures from becoming lawsuits
• Platform owners who turn “anyone can build” into “everyone builds safely” by prioritizing security, compliance and governance
• Site reliability engineers who design for resilience

The second-order effects of AI-native development

1. Procurement is more rigorous: When applications can be built in days, executives start asking questions: “Why are we paying millions for software subscriptions? Why are we funding multiyear implementation programs?” Every vendor will need a far stronger answer than “That’s how we’ve always done it.”
2. Shadow IT accelerates: Marketing can build applications. Finance can build applications. Operations can build applications. And security, compliance and governance teams find themselves chasing hundreds of applications they didn’t know existed. Your attack surface just doubled overnight.
3. The talent model flips: The traditional offshore delivery pyramid — hundreds of people executing tasks defined by a few experts — becomes harder to justify. You need fewer people performing routine work and more people capable of directing AI, planning architecture and making expert decisions.
4. Intellectual property is in flux: When AI can already do so much, what’s left? Proprietary data, trusted brands, customer relationships and the ability to operate safely in regulated environments are now your key differentiators.

3 priorities for leaders in the next 6 months

The end of the road for legacy application development is here. If a competitor, a customer or even one of your own business units can recreate your application in a week, what’s your reason to exist?

It’s time to make some far-reaching decisions about your operating model. Focus on these three priorities in the next six months: 

1. Audit every licensed application against the cost of building an AI-native alternative.
2. Eliminate delivery models that sell hours and tickets. Customers are now buying outcomes.
3. Rebuild for an AI-native future. Invest in architecture, governance, security, platforms and outcome pricing.

We’re advising clients to be a lot more discerning about what they continue to buy — to keep the systems that provide genuine differentiation, regulatory assurance or strategic advantage, and challenge everything else. If a capability can be rebuilt faster, cheaper and more effectively with AI, someone will do exactly that. Will that be you or your competitor?

• ALSO READ: Your bank is now an application — and that changes everything

What to do next

Read more about NTT DATA’s Application Development services to learn how to build software and platforms for greater differentiation.