A small production team used Claude Opus 5.5 to build one internal app for drafting scripts, attaching media, sharing work with editors and collecting ideas. The point was not to reproduce every feature of its existing software. It was to put the few functions the team relied on in one place.
That distinction matters for teams paying for several software-as-a-service (SaaS) products—subscription tools delivered online—while using only a small part of each one. In this case, the team set out to replace five paid subscriptions, including workflows handled in Notion, mymind, Typefully and Google Drive. Its estimate of roughly $4,000 in annual savings describes this case, not a result other teams can expect.
Build around the work that keeps a tool in place
After two years with several of those products, the team estimated that it used about 10% of their available features. Notion remained important chiefly because editors needed scripts with comments tied to specific passages and supporting images or video. The team found its Google Docs workflow did not handle those media-heavy comments the way it needed. Meanwhile, teaching new colleagues to navigate a larger system of databases, blocks and settings took time.
The custom app kept the script at the center. A writer can attach a comment to a line or paragraph, then add an image or a playable video clip to that comment. An editor sees the material beside the passage it supports, rather than having to match a separate file to the script. The team can also send a read-only link to an outside editor, who can review the script and download its attachments together as a ZIP file, a compressed archive.
▲ Media attached to script comments
Other views bring related work into the same workspace. A media library collects assets uploaded to script comments and links each one back to its original passage; a search for “subscriptions,” for example, finds a graphic associated with that topic. A bookmarks area gathers saved X posts for idea development. The app also handles X post scheduling and sponsorship tracking that previously sat in separate tools and spreadsheets.
This is a narrower aim than rebuilding Notion. The team did not need every database structure, formula or edge case in a general-purpose product. It needed a script, contextual feedback, accessible media and a handoff to editors. That limited scope appears to be the main reason a custom app was practical.
Let the AI work inside the app
The team built the workspace through vibe coding: describing desired behavior in plain language while an AI coding tool writes and revises the software. Its development involved approximately 20 rounds of prompts. Across five days of intensive Claude Opus 5.5 use, the team encountered no usage limit and reported no failed feature request. Those are observations from this build, not guarantees about other projects or usage plans.
The app was also designed to be agent-native, meaning an AI agent can update its working data rather than only suggest changes in a chat. Claude Code uses a configured skill—a task-specific set of actions—to add an idea directly to the app’s database. In one example, that process produced a new script entry with a formatted production template. The writer could open and develop the entry without copying an outline from another workspace.
▲ AI updates to a custom app
Claude’s Projects workspace helped keep coding and design tasks in separate threads. Claude Code also produced a Figma-style board of interface components and three alternatives for a script list: a status board, a column-based Kanban board and a dashboard table. Claude Cowork generated a 14-slide deck describing the app. These outputs did not replace the app’s core editorial workflow, but they let the team explore its layout and document what it had built in the same broader working environment.
What this approach changes—and what it does not
A standard SaaS purchase starts with an existing feature set and asks a team to adapt its work to that product. This build started with the team’s recurring tasks, then assembled only the necessary pieces. Its advantage appears to be less switching between tools and less onboarding around features the team does not use.
The example does not show that every subscription can be replaced quickly or that a custom app needs no upkeep. It shows a focused build for a particular production workflow. The team’s favorable assessment of Claude Opus 5.5 rested on its observed speed, cost and reliability during that work, rather than on a complete comparison of every possible app or AI model.
Start with one recurring workflow
The practical lesson is to define a small target before trying to replace a software suite:
- Identify the feature that keeps your team using an existing tool, such as comments with media attached to exact script passages.
- Map what happens before and after that feature. In this case, ideas became scripts, comments carried assets, and read-only links delivered the result to editors.
- Build and evaluate that connected workflow first. If AI needs to create records in the app, give it a defined way to do so and check that the resulting entries fit the team’s process.
Claude Opus 5.5 helped this team turn a narrow set of needs into one working app. For others, the useful question is not whether AI can recreate an entire SaaS product, but which small, repeated part of the job would be simpler in a tool built for it.