Legacy AI
Visualize the core systems that "no one fully understands" with AI, and move toward safe migration. See it first, then you can move it. From there, our people and yours work side by side, all the way through to safe migration.
What is Legacy AI
Legacy AI is a service that visualizes existing systems through AI reverse engineering, supporting safe migration and modernization.
Departures, reliance on specific individuals, no documentation. AI turns a black box that has been running for 10+ years into a "map anyone can use to make decisions," in as little as 2 days (up to 5 business days depending on scale). No tools to buy, no dedicated team to assemble (done-for-you).
Core system migrations undertaken without first getting an accurate picture of the current state have led to major losses even at large enterprises. "Visualizing" the system before moving it is essential.
Does any of this sound familiar
- Only one specific person truly understands the whole system. When they leave, the internal knowledge goes with them
- Design documents no longer match reality. Every modification means guesswork, and judging the scope of impact takes time
- Just re-investigating the current state used to take 3 people x 2 weeks. The cost and time keep pushing the modernization decision further out
Moving forward "without seeing" carries a heavy cost
Core system migrations undertaken without first getting an accurate picture of the current state have led to major losses even at large enterprises. "Visualizing" the system before moving it is essential.
Shipments were halted for more than 2 months, with an impact reported at about ¥15 billion scale.
Delivery delays on some products were reported.
* Both are reference cases based on published materials and news reports. They do not indicate our involvement.
AI reads through the "inside" of the system and turns it into a map
AI reverse engineering handles everything from visualization to documentation on your behalf.
Reading multiple sources at once
Beyond source code and databases, AI also cross-reads existing design documents and exchanges from Slack, Jira, Backlog, and more.
Cross-checking code against design documents
Detects gaps such as "in the design document but not in the code" or "in the code but never reflected in the documentation."
Turning the full picture into a "diagram," then human review
Module relationships, business flows, and screen structures are reconstructed as diagrams. AI output is verified by experienced engineers before delivery.
For end-to-end migration and modernization on a contracted basis, see our Legacy modernization and migration →
What we receive, and what we deliver
- Source code
- Databases
- Existing design documents (even old ones)
- Exchanges from Slack / Jira / Backlog, etc.
- Module map / business flow diagram
- Screen inventory / screen reproduction
- Function inventory / data and ER overview
- Non-functional requirements / current risk inventory
- Migration decision report
* In the initial visualization (Phase 1), your existing design documents are used as reference. Formatting to your specified template is handled in the design and development phase, after the migration decision has been made.
How it's different from other options
We compared ways of grasping the current state across three dimensions: speed, team structure, and deliverables.
| Our service | Typical AI development | Traditional manual investigation | |
|---|---|---|---|
| Visualization speed | As little as 2 days (up to 5 business days depending on scale) | Several weeks | 3 people x 2 weeks or more |
| AI-driven analysis | Yes | Yes | No |
| Verification by engineers | Yes | No | Yes |
| Specialized in visualization | Yes | No | No |
| Deliverables usable for internal approval | Yes | No | Depends on the person |
| Coverage (core / legacy systems) | Yes | Depends on the person | Depends on the person |
| Continuity and documentation after migration | Yes | No | No |
| Cost guide | Quoted based on scale | High | High |
AI visualization isn't the goal, it's the starting point
Using visualization as the entry point, we support you continuously from grasping the current state through to executing the migration.
Resolving "AI makes me uneasy" through a built-in mechanism
We never hand you raw AI output as-is. Every analysis comes with its "basis" attached, is corroborated across multiple sources, and is verified by humans. That's how "AI's inference" and "confirmed fact" stay clearly distinguishable at a glance.
Basis always attached
Every analysis result clearly states which file and line of code it is based on. Your team can cross-check it on the spot as well.
"Confirmed" and "estimated" kept distinct
Each conclusion is assigned one of 4 confidence levels: confirmed (multiple sources agree) / estimated (single source) / needs verification / unknown.
Corroborated across multiple sources
Code, databases, design documents, Slack/Jira/Backlog, your materials, and discovery sessions are all cross-checked, and something is marked "confirmed" only once 3 or more sources agree.
We don't take AI output at face value
In our source-priority order, AI-generated content is weighted lowest. AI is a first draft. Confirmation comes from people and primary sources.
Contradictions aren't hidden, they're checked
Where sources disagree, we log it in a Q&A log, confirm with you, and then reflect the resolution.
In an investigation of one core system (about 12 business categories / 31 functions scale), we raised the confidence of the business flow documentation to over 90% by moving from a code-analysis-only stage through cross-validation with your materials and discovery sessions across multiple sources.
How we handle the source code you entrust to us is explained through our approach to security.
See our approach to securityHow to apply
From your inquiry to receiving the deliverables, as little as 2 days (up to 5 business days depending on scale). You can start small, with low risk.
Get in touch
CONTACTJust describe your current situation briefly through the form. No need to prepare detailed materials in advance.
Discovery and scoping
HEARINGWe decide together which system to visualize and how to hand it over safely. An NDA can also be signed if you'd like.
AI analysis and engineer verification
ANALYZEAI analyzes it in as little as 2 days (up to 5 business days depending on scale), and experienced engineers verify it by eye.
Delivery and explanation of deliverables
DELIVERWe report back with diagrams and figures. Whether to proceed further is a decision you make after seeing the deliverables.
* Information we receive is managed under a scope of work, retention period, and access rights agreed upon in advance. We can only take on a limited number of projects per year, and we handle each company with care, one at a time.
6 deliverables turn the full picture into your "next decision material"
Current-state investigation summary
A single overview of the system's scale, structure, and technologies used. The overall premise becomes clear at a glance.
Module map
Visualizes functional dependencies as a diagram. Makes clear the scope of impact when making modifications.
Business flow overview diagram
Reconstructs key processes from a business perspective. Makes the daily-to-monthly big picture clear.
Screen inventory / screen reproduction
Identifies existing screens and reproduces them as an inventory and flow. Makes the migration scope concrete.
Risk / technical debt inventory
Organizes aging and knowledge-siloed areas by priority. Makes clear where to start.
Migration decision report
Presents the key considerations and recommendations for modernization and migration, along with an effort estimate. Becomes material for your next decision.
For details on contract types and cost estimates, see Engagement models & team structure.
Request an actual deliverable sample
We send a real sample of the 6 deliverables above (8 sections in total) individually to interested companies. We prepare an example close to your current state, at a level of detail you can use directly for internal approval and migration decisions.
* All samples use fictional data. You can request one through our contact form.
We won't let "migrated, but now harder to use" happen
The goal of migration is not simply to make things new. We carry forward every function you rely on today without loss, and make it even easier to use. And we leave behind documentation so it can keep growing in the future.
- 01
Functionality, at minimum on par with today
We identify the functions used in your current operations and confirm nothing is lost after migration.
- 02
Then, we improve it
Technical debt and inefficiencies revealed during visualization are resolved at the point of migration.
- 03
Documented for the future
We record the design and specifications as documentation, so future upgrades can be decided in-house as well.
Frequently asked questions
Is it safe to hand over our source code? Is it used to train AI?
It is safe. Under enterprise-grade privacy settings, the source code you entrust to us is not used for training beyond the scope defined in our contract. We analyze it in a managed environment with data storage and training disabled, and proceed under an agreed scope of work, retention period, and access rights. An NDA (non-disclosure agreement) can also be signed.
Is it a problem if we don't have design documents anymore?
Not at all. Beyond code and databases, AI also cross-reads existing design documents (even old ones) and exchanges from Slack, Jira, Backlog, and more, building the migration plan from the current state. We do not rely on interviewing specific individuals.
Can you handle older languages and environments (VB.NET, legacy Java, etc.)?
Yes. We have a track record analyzing languages that have been in long-term use, including VB.NET, older versions of Java, .NET, and PHP. All you need to prepare is the source code and, if available, existing documentation. Even at the stage where you're not sure it can even be read, please reach out first.
Can screens (UI) also be reproduced as a deliverable?
Yes. We identify the screens in your existing system and reproduce them as a screen inventory and flow. We have past experience reconstructing a large number of screens from a legacy ERP (VB.NET).
Is it safe to simply trust the AI output as-is?
No, and we don't ask you to. AI output is weighted lowest in priority, corroborated across multiple sources, and verified by experienced engineers. Each conclusion comes with a confidence level (confirmed / estimated / needs verification / unknown) and its basis (file and line), delivered in a form where confirmed fact and inference stay clearly distinguishable.
How long until we receive the deliverables?
The initial full visualization is delivered in as little as 2 days (up to 5 business days depending on scale). Only the step where AI reads through large volumes of code is sped up; we don't skip quality steps such as corroboration, human verification, and confirmation with you. You can start small, narrowing the scope first.
Can you match our specified design document format?
In the initial visualization (Phase 1), we deliver in our standard format, using your existing design documents as reference. Formatting to your specified template is handled in the design and development phase, after the migration decision has been made.
Will our current functionality still work after migration? Can it be maintained and modified?
We identify the functions used in your current operations and confirm nothing is lost after migration. The new system aims for, at minimum, parity with the current one, and we improve the inefficiencies revealed during visualization. Because we document the design and specifications, future modifications become easier to decide on in-house as well.
Can you give us a sense of cost?
The initial full visualization (initial diagnosis) starts from several hundred thousand yen, quoted based on scale. You can start small, and the initial consultation and quote are free. Pricing for full-scale implementation is presented based on the visualization results.
Why not start by visualizing your current state?
You don't need a big investment right away. First visualize the current state, get decision-making material you can use for internal approval, and then decide on the next step. The initial consultation and quote are free.
* "As little as 2 days" is the target for starting the initial visualization. It may extend to around 5 business days depending on the scale and complexity of the target system.
* "Initial diagnosis from several hundred thousand yen" is a quote based on scale. Pricing for full-scale implementation is presented based on the visualization results.
* "200+ projects" is Miichisoft's own development track record (our company record).