AIToday
AI Coding AssistantsTop Companies' AI MovesAI Business & IndustryTop Companies AIPublished: Nov 7, 2026, 06:30 JST

SMFL taps Replit for enterprise vibe coding, cuts 6-month job to 1 week

Send on LINE
SMFL taps Replit for enterprise vibe coding, cuts 6-month job to 1 week

3 Key Points

  1. What happened

    On November 4, SMFL announced enterprise vibe coding using Replit's technology, adding per-customer customization to its asset management cloud assetforce. It completed a customization expected to take 6 people about six months with 1 person in 1 week.

  2. Why it matters

    SMFL pairs the speed with safeguards — an access gate checking authentication and permissions, deterministic tests, and final review by its own engineers.

WHO IT HITSEnterprise IT and engineering teams that maintain packaged SaaS for finance and accounting will watch whether this speed-with-controls model holds up; SMFL's own engineers and its FDE (forward-deployed engineer) hires are the first to work this way.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, every morning.

Context & Analysis

SMFL is a leasing company with roughly ¥9 trillion in operating assets, and its path to this announcement runs through internal development. A turning point was its 2016 acquisition of the leasing business from GE's Japanese unit, which gave it an in-house development team. That led to a 'customer zero' approach of delivering systems it had honed itself, and assetforce, first sold externally in 2021, has grown to more than 200 paid contracts.

The bottleneck was customization: fine-grained customer requests consumed engineering time. SMFL's answer is vibe coding, a style it attributes to OpenAI co-founder Andrej Karpathy that spread around 2025. President Tetsuro Imae described it at a press conference as telling the AI in words what screen or workflow you want, with no instruction documents for engineers needed.

SMFL built ADP, a development foundation connecting assetforce and Replit, which checks authentication and permissions on both sides and lets through only approved data and operations. It also addressed token costs by splitting context into work units so each developer's AI receives only the information for their scope, and it requires deterministic tests plus final human review rather than leaving judgment to the AI. Replit CEO Amjad Masad argued vibe coding reduces dependence on individuals, since AI-written code tends to be more standardized. Imae acknowledged a risk of over-personalization and said SMFL will propose assetforce's default functions as the standard while absorbing some customization demand, and is training forward-deployed engineers as requirements definition grows more important.

FAQ
What is enterprise vibe coding?
It is SMFL's development method using Replit's technology, where people instruct AI in natural language to write code and produce working apps. SMFL adds an access gate, deterministic tests, and final checks by its own engineers.
How much faster was the assetforce customization?
A customization expected to take 6 people about six months was completed by 1 person in 1 week. For a depreciation calculation system with over 200 screens, SMFL expects the development and testing period after requirements definition to be cut to one-third of the conventional baseline.
What is Replit doing in Japan?
Replit opened a Japan office to strengthen enterprise AI development services, following a London office opened in September, and aims to expand in Asia from Japan. SMFL is preparing to offer Replit licenses and customer referrals, and will also run enterprise training.
Top Companies AIRead Original Article
Send on LINE

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.

Questions and answers are published on this page.

Related Articles

Next articleTanium joins OpenAI's Trusted Access for Cyber program