← Back
Construction & interior Business automation / AI agent
AIStaff — Business automation by AI staff (agents)

Manual paper, handwriting & drawing work is auto-processed by AI staff (agents). Redirecting up to ~1,300 hours/year to judgment and management.

Sanei Kenso Co., Ltd. — Non-value-adding manual work such as input, transcription, aggregation & matching is shifted to AI staff. Without changing on-site practices at all (paper, handwriting, MFP scanning), CS surveys, thank-you cards & drawing takeoff are automated, building a setup where people can focus on exception checks and judgment.

1,300h
Annual hours saved (max)
Three workflows combined: ~675–1,339 hours/year
90%
Drawing-takeoff effort reduced
Auto-extract floor, room & area with VLM
97%
CS survey processing reduced
Auto-read & aggregate, paper as-is
Challenges
Before

Non-value-adding manual work such as input, transcription & aggregation squeezed time. Drawing takeoff became person-dependent and time-consuming, and existing SaaS didn't fit the work and failed to take hold.

  • We want to shift routine tasks to AI staff and focus on judgment work
    Up to ~1,300 hours/year spent on non-value-adding work
  • We want to auto-read & aggregate paper surveys as-is
    514 forms × 15 min ≈ ~193 hours/year of manual entry
  • We want to auto-extract floor, room, area & spec from drawings
    4–10 hours per project · up to ~1,000 hours/year, person-dependent
  • We want efficiency gains without changing the field (paper, phone, proxy entry)
    Existing SaaS didn't fit the work and failed to take hold
Solution
After

Input, aggregation, matching & sorting are handled by AI staff (VLM/handwriting OCR + Dify), while people focus on exceptions, judgment & improvement. Three workflows are automated into one structure via a shared processing pipeline (classify → read → structure → rule-based decision → exception extraction). Standard tasks run on Dify (no-code); high-precision drawing processing & core-system integration use a custom-development hybrid for phased rollout.

Input Paper / PDF / Excel AI reading VLM / handwriting OCR Structure & decide Rules / exceptions Output Excel / Teams Human involvement only on exception flags (a few%–10–20%) Three AI staff (individually optimized on a shared processing base) CS survey Read → aggregate → notify Thank-you card Decide → award points Drawing takeoff Extract floor · room · area
aistaff.example.jp
Room A Room B Bath WC Hall LDK Takeoff result 3F · Room · m² LDK16.2 Room A8.4 Room B9.1 Bath3.0 WC1.6 Total area 38.3m² OK Review
Drawing-takeoff AI staff: auto-listing of floor × room × area
Tech stack Dify (LLM Orchestration) VLM / handwriting OCR RAG Python Teams / Excel integration
3+2AI staff
NewAutomation
Considering adoption at your company too? → info@amela.vn