Why Owning Native Tools is the Future of Agentic Work

"Software is really meant for machines, and humans using software really is kind of an outdated concept."

With this statement, Philipp Holke frames his argument not as a new feature, but as a fundamental challenge to how most software is built today. The core idea is that the next era of autonomous systems requires a new foundation, one that moves beyond the limitations of the current SaaS landscape.

SaaS is a Tool, Not a Solution

For years, the promise of Software-as-a-Service was a fleet of specialized tools to solve every business problem. But as Philipp argues, this has led to a fragmented reality where the actual "job to be done" remains undone by humans. Companies subscribe to platforms like Monday.com or Asana for project management, but the work itself still requires people to log in, manually update tasks, and bridge information gaps between systems.

"SaaS applications are simply tools," Philipp states. "And the next era of autonomous systems is meant to be agents using those tools and not having people required to kind of log into all of these different tools and do these things manually."

The problem is that real work is messy. It's spread across HR software, email, WhatsApp, and documents. The work is not centralized, which makes it incredibly difficult to build an autonomous system because there is no single source of truth for an agent to operate on. This fragmentation makes it nearly impossible to access or reason across data coherently. Philipp's argument forces us to confront the reality that software built for humans to click on is a dead end for software built for agents to operate on.

The AGI Principles for Agent-Native Work

Mission Systems developed a set of product principles for building software designed for machines first. They call it AGI: Autonomous, Generative, and Invisible.

Autonomous means the software is expected to get the job done itself, not just provide a tool for a human to do the work. It takes on the responsibility for the outcome.

Generative means the system is dynamic and adaptive. Instead of a static, one-size-fits-all dashboard for every employee, the information presented is tailored on-demand to the individual user's role and context.

Invisible is about hiding the complexity. While AI engineers may enjoy tinkering with tool calls and workflows, the general workforce should not have to. The automation should happen behind the scenes, without the user needing to see or understand the underlying mechanics.

Owning the Native Stack

To put these principles into practice, you need to connect agents to all the applications a business uses. But wiring together dozens of third-party SaaS apps creates its own problems. An agent trying to process a simple leave request might have to make calls to three different vendors for rostering, HR, and payroll, each with its own database schema and API quirks. Managing this at scale becomes prohibitively difficult.

This creates a fascinating split. In one recent discussion, we explored how to feed existing API documentation to an agent so it could operate within the current SaaS ecosystem. Philipp's position is more radical, suggesting that such efforts are a dead end and that the only path forward is to build a new, unified application layer that agents can own natively.

The alternative, Philipp argues, is to build and own that unified, native application layer. At Mission Systems, this core agentic layer is called Launchpad. It is a single interface where users interact with a suite of in-house applications for functions that, five years ago, would have each been a separate SaaS company. Because the cost and time to develop these simple CRUD applications has fallen dramatically, it is now preferable to build them in-house. This gives builders full control over functionality and makes it far easier to create powerful agentic workflows that can operate across every layer of the business.

The Payoff: Impossible Insights

During the Q&A, an audience member challenged this approach, asking what makes it better than an incumbent like SAP. Philipp's response cuts to the core of the value proposition.

"Our system is built from the ground up to share all of those data sources," he explains. He gives the example of an aviation client in Australia. "They now have an agent through which they can query and ask questions like, Which aircraft tends to always run late? And because we built a rostering and payroll and HR and leave requests, plus the entire engineering layer, plus the flight operations layer, all in one application... those types of insights would be completely impossible to obtain, practically impossible to obtain, if you didn't have a unified application layer."

When data is not siloed, an agent can triangulate across the entire business, connecting crew schedules to maintenance logs to flight operations. It can reveal insights that were previously hidden not because they were complex, but because they were inaccessible.

---

SQ Collective hosts Coworking Fridays for founders, operators, and AI builders working through real product questions in Singapore.

Join an upcoming Coworking Friday, Technical Homebrew or a Masterclass: https://lu.ma/ai-labs

Explore SQ Collective: https://www.sq-collective.com

Michael