Human Work Focuses on 'Three Approvals': Agentic AI Autonomously Operates iOS App Business — 3 Days from Planning to Review Submission, 13 Apps Released on App Store

AllNew LLC (Headquarters: Minato-ku, Tokyo; Representative Member: Masanori Ueno; hereinafter 'AllNew') has autonomously operated its iOS app business end-to-end, from market research to customer support, using its development platform 'App Factory', which leverages 'Agentic AI' where multiple AI agents collaborate autonomously to execute tasks. As of June 11, 2026, the company has released 13 iOS apps on the App Store. Humans are primarily responsible for three approval gates: 'Launch', 'Pre-release Confirmation', and 'Improvement Adoption'. All other practical work is carried out autonomously by AI. For the most recent app, the entire process from market research to review submission was completed in 3 days.

The apps are primarily in the healthcare and utility sectors, with several more currently being submitted and under review. The lineup will be expanded sequentially.

Additionally, the company officially launched 'POIPOI-STUDIO' on June 11, 2026, a feedback site for soliciting improvement requests and new app ideas from current and potential app users.

① Starting from the 'Launch Gate', Agentic AI cycles through market research to sales. User feedback is initially organized by a reception AI, after which humans make decisions at ② Pre-release and ③ Improvement Adoption gates. Adopted feedback is fed back into market research, creating a cycle.

Background of Development: The Wall of 'Business Continuity and Operation' Beyond Code Generation and New Trends

With the explosive spread of generative AI since 2023, automatic source code generation has rapidly become commonplace. However, to establish and sustain an iOS app as a 'business', it is necessary to consistently manage a vast and fragmented set of processes beyond just writing code: market research, compliance with Apple's various rules and guidelines, real-device testing, store application preparation, marketing (ASO/advertising operations), customer support, and daily improvement decisions.

As the focus in development shifts from 'speed of creation' to 'sustainable operation as a business', AllNew has seamlessly connected these repetitive processes using Agentic AI. The company designed a next-generation operational cycle of the 'Human-in-the-loop' type, where humans only approve at key points, and embodied it in its proprietary platform 'App Factory'.

This represents a shift from the stage of 'prompting AI each time to get deliverables' to the stage where 'humans design and control the loop (Observe → Plan → Execute → Verify → Correct) that AI agents run autonomously'. Among developers worldwide, this approach is called 'Loop Engineering' or 'Harness Engineering' and is rapidly gaining attention as a next-generation software development paradigm. App Factory is a practical commercial application of this concept to the iOS app business. Coincidentally, Apple clarified rules regarding spammy and low-quality apps in its App Store review guideline revision in June 2026. App Factory's operational premise of 'not releasing anything that doesn't meet a certain quality standard' aligns with this trend.

Features of 'App Factory': An Autonomous Operation Platform Backed by Release Track Record

'App Factory' is an in-house operational platform where multiple AI agents, each with specialized expertise, take on roles and pass the baton through the stages of market research, planning, research, development, testing, App Store application preparation, sales, and customer support (Note: not publicly available).

1. Proven Track Record of Being Published on the App Store After Apple's Review Process

A group of iOS apps, including 'WeightSnap' (weight management) and 'GlucoSnap' (blood glucose recording), are already being distributed to the general public on the App Store. The fact that they have passed Apple's review process and reached actual users demonstrates that the AI-powered development process has reached the commercial app publication and operation phase.

Behind the apps currently being distributed, dozens of prototypes have been created. Apps deemed not to meet a certain quality standard are discarded or rebuilt at the App Store application stage, even if completed. The operational premise is to not release low-quality apps into the ecosystem.

2. Measured Speed: 3 Days from Planning to Review Submission, 7 Days to Publication

For the recently released 'AI Medication Reminder - Prescription Reminder', the entire process from market research and planning to implementation, testing, and creation of legal documents and store listing materials was completed in 3 days, leading to review submission. It took 7 days to be published on the App Store after Apple's review. What this platform shortens is not 'code writing time', but the entire process required to reach publication as a business.

3. A 'Sustainable Cycle Structure', Not a One-Off Tool

Unlike conventional code generation tools that are used transiently, this platform automatically passes the deliverables of one process to the next, and furthermore, feeds data obtained from customer support after release back into the initial market research phase. This creates an ecosystem where the business continues to operate autonomously.

The Role of Humans: Focusing on 'Three Key Gates' for Decision-Making and Quality Control

Repetitive practical tasks are handled by Agentic AI, while humans concentrate their resources on the following three gates that require 'irreversible critical decisions'.

① Launch Gate (Idea/Policy Direction): Humans communicate an idea like 'it would be nice to have an app like this' or development policy. This signals the start of Agentic AI. You can either give specific development themes or have the AI conduct market research and formulate the theme itself.

② Pre-release Gate (Final Review): When application preparations are complete, humans perform operational checks using a real-device test environment. They also rigorously review legal documents, privacy policies, store listing materials, and guideline compliance, making the final decision to submit to Apple's review and publish.

③ Improvement Adoption Gate (Go/No-go Decision): AI presents user feedback from 'POIPOI-STUDIO' as a report that has been categorized, summarized, and turned into implementation candidates. Humans decide 'Adopt (Go)', 'Reject (No-go)', or 'Hold'. Requests that receive a 'Go' are fed back into the ① Launch Gate.

[Division of Roles between AI and Humans]

・Scope handled by Agentic AI: Market/demand research, creating project proposals, researching and implementing necessary technologies, testing, App Store application preparation, autonomous operation of Apple Ads (formerly Apple Search Ads) within budget, ASO (App Store Optimization), categorization and summarization of feedback

・Scope handled by Humans (Mandatory): Development start instructions, confirmation of legal/privacy-related documents, decision to submit to App Store review, decision to release/publish, setting budget limits

App group developed and operated by App Factory, including 'WeightSnap' and 'GlucoSnap'. As of June 11, 2026, 13 apps have been released and are being operated on the App Store.