Buyer guidePlatform and AIVersion 1.025 October 2026
Modernize your platform and put AI into production
AI stalls when the platform under it cannot carry it: data locked in a legacy core, integrations nobody dares to change, no record of what an agent did and no way to stop it. We modernize that platform one component at a time, with a rollback at every step, and put the AI operating layer on top, so AI runs in production while the business keeps running.
On this page
One engagement: modernize the platform, then run AI on it under control.
Platform modernization with AI is one engagement that replaces a legacy platform component by component, with a rollback at every step, and puts the AI operating layer on top, so each AI workflow runs with a named owner, a full record and a way to stop.
In this page, “we” means Hikari Blue, which engineers and operates the AI operating layer that connects models, data, and business workflows.
A platform that cannot carry AI stops the AI, whatever the model.
Surveys of the middle market name data, skills, legal clarity and legacy systems as the obstacles, not access to models.
Data locked in the core
The data an AI workflow needs sits in a system of record with no clean interface.
Integrations nobody dares to change
Undocumented logic lives in legacy code; any change puts operations at risk, so the AI waits.
No record, no stop
Nothing records what happens on the old platform, and nothing halts an agent without halting the workflow.
Two programs that never meet
Modernization and AI run as separate programs, each waiting for the other.
Modernize under risk control, then put the AI operating layer on top.
Both halves run as one engagement under one partner, so the first AI workflow does not wait for the migration to end.
Modernization under risk control
Map what the legacy really does
We surface the business logic, undocumented parts included, and instrument the legacy before touching it.
Replace one component at a time
New components are tested on a copy of live traffic, then take over step by step, each cutover with a rollback.
Build what packages no longer fit
Where packages no longer fit, we build module by module on a blueprint signed by a senior architect.
The AI operating layer on top
Policy and governance
Rules on which task may run, on which model and data, under whose authority, evaluated at runtime.
Routing and runtime
Each task goes to a model chosen by risk, cost and jurisdiction, with a declared fallback: Claude, Gemini, GPT or open weights.
Evidence and audit
Every prompt, output, tool call and operator action is recorded when it happens, append-only.
Control
Each agent acts under its own identity and tool scope, with a stop path.
One week to decide, then phases that each end with a go/no-go.
Each step gives you what you need to decide on the next one, before you commit to it.
- 01
First exchange · free
Describe the platform and the workflow that matters most. A named senior partner replies quickly, in writing.
- 02
Operating Assessment · 1 week, from $18,000
1 week on one critical workflow and the platform under it: baseline, value, risk and the controls production needs.
Deliverables: Time to Stop established, the workflow’s value quantified, a named owner, a plan to put it into production.
- 03
Modernization and AI · in phases
Legacy Modernization: typically 24 to 48 weeks, from $300,000. AI-Enabled Operations, one workflow from pilot to governed production: typically 12 to 24 weeks, from $180,000, fixed at the Assessment.
Payment: in phases, each ending with a go/no-go.
- 04
Transfer · on a dated plan
Operation moves to your team on a dated plan written into the contract from day one. Run support is an option you contract, not a dependency you inherit.
Scope, pricing and duration depend on the agreed engagement.
At exit, the platform, the code and the runbooks are yours.
Ownership transfer is written into the contract from day one, for the platform as much as the AI.
Code in your repositories
Custom code, tests and pipelines live in your repositories from the build phase.
Prompts and models
Prompt libraries, custom models and fine-tunes belong to you at exit.
Runbooks and operating documentation
Handed to trained owners on your team, with the stop procedure and its drill history.
No surprise licence
The reusable components we retain are listed exhaustively in the MSA.
What we can show, and how.
Cases are anonymized at each client’s request, and no performance figure is published. Each indicator is measured by the client; values are not disclosed. Walk-throughs and demonstrations take place under NDA, led by a named partner.
What CIOs ask before they modernize with AI.
Can you modernize without a rebuild?
Yes, and we prefer it. We replace the legacy one component at a time behind a routing facade, test each new one on a copy of live traffic and keep a rollback for every cutover. Your current system runs the business until each part retires.
Do we keep our ERP?
Yes, if it still runs your business. An SAP, Oracle or Microsoft Dynamics core that cannot be ripped out is modernized through interface extraction, peripheral replacement and progressive cutover. AI agents then act in it under their own identity and scope, and your operator approves each consequential step.
How do you avoid vendor lock-in?
By architecture and by contract. Switching model provider is a routing change, not a rebuild, and the platform is built in modules with no single-vendor cloud lock-in. At exit, code, prompts, models and runbooks are yours; what we retain is listed in the MSA.
Who is accountable?
A named partner. Xavier de Maillard, partner of the Engineering Practice, signs the engagement and answers for its delivery from week one to exit. On your side, each AI workflow has a named owner.
What is the first step?
A free first exchange. Describe the platform and the workflow that matters most; a named senior partner replies quickly, in writing. If the brief fits, the next step is the Operating Assessment: 1 week, from $18,000.
Bring the platform that holds your AI back.
30 minutes with a named partner; the first exchange is free. We look at your platform and the workflow that matters most, and tell you in writing what to modernize first and what the first phase would cost.
- The first component to replace, and its rollback
- The first AI workflow it can carry
- What the first phase would prove and cost