Step Forward, Not Sideways: The AI-Native Answer to App Sprawl

Two interesting pieces crossed my feed over the past couple of weeks. First, Lovable and Microsoft announced a partnership: apps built in Lovable can now be published straight into a company's Microsoft tenant, powered by Copilot Managed Runtime. Then, not entirely by coincidence, Satya Nadella followed with a LinkedIn post framing Copilot as "a new OS for work, with a governed infinite SaaS factory built in."

Both signal where enterprise software is heading, and a reset of business SaaS is, without a doubt, overdue. We're also seeing similar demand from our Agent360 customers - domain experts want more control to build the tools that fit how they actually work.

But before letting everyone build apps, businesses that want to be truly AI-native should weigh four things we've seen play out with our customers.

1. The hidden cost of maintenance

We have seen this before. Every wave of democratised software has followed the same arc. From Microsoft Access databases to Lotus Notes applications and, more recently, Power Apps and a string of other low-code and no-code platforms, each let the people closest to the work build their own tools. Each delivered real value. Each also left IT with years of orphaned tools that nobody fully owned.

AI app builders will likely repeat that pattern at far greater speed. The new apps land in your tenant, visible to IT and under existing policies, and engineers can review and extend the code. Every one of those apps is a codebase someone has to own, patch and keep correct long after its builder moves on.

For Australian mid-market businesses, this matters most. Many have few in-house engineers and rely on external managed service providers who are still building the skills the agentic era demands. An app factory without the engineering capacity to maintain what it produces quickly becomes a backlog.

2. Capabilities, not apps

When agents become the interface to your systems and processes, the question stops being "which app should we build?" and becomes "what should people do, and what should agents do?"

That's an operating model question, not a tooling one. Instead of people building more apps to work around their workload, AI-native businesses redesign the work itself: agents take on repeatable tasks end to end, and people move to judgement, relationships and exceptions. The unit of value shifts from the app to the capability the business can perform.

Domain experts still shape the work: they bring the knowledge and help design the workflows agents run on, rather than maintaining apps of their own. Boundaries between functions start to blur, and that's where growth and efficiency compound. It also means fewer apps to build, own and maintain.

3. Provenance is the basis of trust

As Satya points out, agents make systems of record more important. But trusting what an agent does takes more than access to the record. Democratised app building is only half the story: every app can read and write the record, but the knowledge and reasoning behind each action stay locked in its own code, with no common trail across them.

Businesses need to know where every answer and action came from, which organisational knowledge shaped it, and who signed it off. In regulated, high-consequence work, that auditability is the difference between a pilot and production.

4. Security designed for agents, not retrofitted

Thinking in capabilities lets businesses design security from the ground up, which means governing people and agents under one policy engine designed for both. Each agent works as a scoped system user inside the same organisational boundaries as its team. It sees only what its role, group and business unit allow, down to individual records and fields, and it can never act with more access than the person it works for.

The alternative is stretching permission models built for humans and legacy apps to cover autonomous agents. That's how AI ends up reaching data a user should never see.

Step forward, not sideways

Letting everyone build apps is a step sideways: the same app-by-app model, rebuilt faster and in more places. For businesses ready to, the bigger opportunity is a step forward, reimagining how the business runs with AI at its core.

That means AI as the connective tissue between your data, systems and processes, not another layer bolted on top. Capabilities replace apps, provenance is built in, and governance is designed for agents from day one.

Our early Agent360 adopters are already making the shift to AI-native without building more apps. One Australian property group gave its staff a governed alternative to shadow AI, built over its own systems: eight weeks in, 88% of users are still active, and a report its implementation partner spent three months on now takes minutes. It's one of a growing number of customers taking this path.

Agent360 by Nimbly is the AI-native business operating system, built so the reasoning is swappable and your org knowledge stays yours. See it on your data, connected to your systems, with no cost to start.

Talk to us: contact@nimbly.au