12 APPS IN Q1. 9 FROM OUR SCIENTIFIC AND RESEARCH TEAMS.

An astonishing 12 applications shipped at Montai in Q1. All of them were built on Montai's new Agentic AI Development Platform, Magentic.

3 of the 12 came from the Platform Engineering team, including 2 full application migrations. 9 came from our scientific and research teams. That ratio is the most interesting thing about how AI is changing software development inside a biotech, and it has almost nothing to do with engineers being faster.

Most of the conversation around AI-driven development is about the engineer typing less and shipping more. That story is real, but it is the smaller one. The bigger story is who gets to build at all. A scientist who would never have written a deployable application six months ago now has the leverage to design one, get it reviewed, and ship it to a contained environment where their colleagues can use it instantly. That is a different shape of leverage, and it is the one that is going to redraw a lot of biotech engineering org charts.

WHAT THE PLATFORM ACTUALLY GIVES THEM

Magentic does not write code for the domain expert. It defines the framework: the project skeleton, the deploy pipeline, the data access patterns, the permissions, the observability, and the lifecycle hooks. The domain expert, or the AI assistant working alongside them, writes the application logic. Magentic absorbs the operational complexity that has always been the real barrier to non-engineers shipping software.

Concretely, Magentic gives any new application 3 environments out of the box: local for the laptop, dev for iteration, and prod for everyone else. Dev and prod sit behind an Okta-secured load balancer by default, so the application is on company SSO from the moment it exists. Each application deploys as its own ECS cluster with dedicated listener rules in the load balancer, ensuring they are completely isolated. Scaffolds come pre-wired to the data layer the rest of the platform already runs on, including RDS and Athena. Every app deployed through Magentic also gets monitoring out of the box, surfaced through AWS CloudWatch and connected to a shared Grafana dashboard, so usage, errors, and resource consumption are visible from day one. The deployment itself is no-code from the application author's perspective: the domain expert describes what they want, and an agent generates the project, configures the framework, and ships it.

The visual layer is part of the framework too. Every Magentic application starts with Montai's brand colors, typography, and layout already wired in, so a scientist who deploys their first app on Tuesday morning has an interface that looks like every other application across Montai. Polish is the default, not a final-week project. That is what lets a one-week experiment look like something a scientist can show to a colleague without apologizing for it, and it is a quiet but real reason apps from non-engineering teams are actually getting used.

The biggest time savings happens upstream of shipping. Domain experts no longer wait for a development cycle to design what they need. They build a prototype in minutes, iterate until they are happy with the results, and share the working tool with their colleagues the same day. The conversation that used to be “we should add this to the roadmap” is now “I built it this morning, here is the link.”

An iterative cycle that used to take months of back and forth between the stakeholder and the Platform Engineering team now happens between the domain expert and their agent, or team of agents, in hours, with little or no Platform Engineering intervention. The domain expert is the designer, the implementer, and the first user at the same time. That compression is the real productivity story, and it is invisible from the outside.

This is the part that feels obvious in retrospect. The barrier to domain experts shipping software was never the typing. It was the long tail of integration, packaging, deployment, identity, secrets, observability, lifecycle, and visual polish. AI assistants made the first part trivially easy, and a platform that absorbs the rest is what makes the whole thing real. Without the framework, you get more notebooks. With the framework, you get applications other domain experts can rely on.

Domain experts are free to explore, build, and share at the speed of their own ideas. The power to design the tool now sits with the people who use it.

WHAT THE PLATFORM ENGINEERING TEAM'S JOB BECOMES

When more apps are coming from outside the Platform Engineering team than from inside it, the team's job rebalances. Less time writing applications other teams have asked us to build. More time defining the framework, hardening the deploy pipeline, watching what domain experts actually build, and continuously improving the tools and workflows that let agents better support our domain experts.

This is not the Platform Engineering team becoming a service desk. It is the opposite. The team builds the framework that lets others build well, and that is a higher-leverage version of the same job. It is also the only sustainable version inside a small biotech engineering organization. A 5-member Platform Engineering team will not absorb the application backlog of 70+ Montaineers spanning data science, machine learning, multiple areas of chemistry and biology, business development, and legal, to name a few. But we can build the framework and let domain experts and their AI agent teams do the rest.

Tracking what domain experts actually build is part of the new job too. The Platform Engineering team watches usage and feature patterns across the apps they ship, and the most useful tools surface themselves through use. When a feature built by a domain expert belongs in a Platform Engineering developed product, we work with them to fold it back in: the domain expert gets credit for the design, the broader team gets a hardened version of the idea, and the Platform Engineering team gets design input from the people closest to the science. Features designed by domain experts end up shipped in Platform Engineering developed tools, and the people who designed them have something they can be proud to have contributed.

WHAT WE ARE WATCHING

12 apps in a quarter is not a brag. It is a number that comes with 3 new questions, and the answers to those questions are most of next quarter's work.

First, the apps that probably should not exist. Some will turn out to be experiments that did not need to ship to a real environment. Knowing which is which, and giving the builders a graceful path to retire what is not working, is part of the framework too.

Second, governance drift. A domain expert building an app does not automatically respect the same data boundaries an engineer does, so the platform has to enforce those boundaries by default rather than by review. That is an ongoing investment, not a one-time configuration.

Third, the harder case. Some apps built by domain experts become the pillar for certain workflows, and that brings new requirements on top of what every Magentic app already gets: formal SLAs, alerting policies on what matters most, and a real maintenance plan that prevents downtime. The model here is partnership, not handoff. Platform Engineering and the domain expert who built the tool work together to keep the application's reliability standards in place, preventing catastrophic failures in a critical-path application and avoiding delays and frustration in the long run.

AI-driven development is mostly being talked about as a story for engineers. Inside a biotech engineering organization, the more important version of the story is about everyone else. The metric that matters is not commits per engineer. It is the number of people we empower to ship their own apps safely.

Everyone at Montai is now a developer. That is what a cutting-edge AI-driven biotech looks like. The democratization of development is raising the bar of what is possible, and the things that get built next will not look like anything that came before.