- Fast, but without control
- The code becomes a black box
- No traceability, no documentation
- Risky in production
Intelligent Software Factory
Softwares que think, delivered with governance.
Buying an AI licence buys you an engine. We deliver the factory: people, platform and the CIT method orchestrated to build software with AI, and with intelligence inside the product itself.
Multi-LLM
No model lock-in · cloud or on premise
IA no core
Not only to build: inside the product
100%
Traceable end to end
Operating modeAI Code Studio
Three ways to code with AI. The CIT method decides which one to use.
◀ freedomcontrol ▶
Vibe
Intent drives.
Prototype
Agent
The goal drives.
Features
Harness
The method drives.
Production
The decision-maker's question
Why is buying a licence not enough?
Because a licence gives you speed without control. The factory gives you both.
- Fast and governed
- Auditable from day one
- Traceable end to end
- Ready for critical production
The model
Three layers. One orchestration.
The advantage is not ter AI, it is orchestrating it. Technology executes, people judge, and the process decides when each one steps in.
Executes
Technology
Our own Citsmart platform: the AI does the heavy lifting, and ships inside the product.
Judge
People
Seniority where the decision matters. AI lifts the team, it does not replace it.
Govern
Process
The CIT method and governance: the layer an AI licence will never deliver.
01 · Technology
Our own platform, and what holds it up
What sets Central IT apart from a consultancy assembling loose tools: the Citsmart platform, modular, on premise or SaaS.
Flagship product
AI Code Studio
The environment where the team builds, with every coding mode: Vibe, Agent and Harness. Where seniority meets the machine.
Agent pipeline
HyperAgents
Specialised agents with shared memory: one agent's output is the next one's input (agent chaining).
Delivery management
Agility
Project control: pipeline, sprints and management. Governance becoming visible operations.
Intelligence at the core
AI does not stop at the pipeline: it ships inside the product. Hence “Intelligent Software”, in the plural: the intelligence stays in what you deliver.
Under the pipeline: what holds the factory up
Multi-LLM with no lock-in
Each task goes to the right model. The AI market shifts every quarter; the pipeline keeps standing.
RAG and enterprise context
The agents read your documentation, business rules and existing code. They build inside your reality.
Integration with systems and APIs
The software is born connected to what you already have, not on an island.
Governance end to end
Every action is traceable: which agent, which model, which decision, which outcome. Human approval where it matters.
Nativo no CITSmart X²
The pipeline is not a side project. It lives at the heart of the platform.
What the factory delivers
Build from scratch or modernise what already exists.
Two delivery models, the same agent pipeline and the same governance method. What changes is the starting point.
AI-orchestrated applications
Coordinates the entire development cycle with intelligent agents that integrate, automate and synchronise every phase of the project. The AI orchestrates flows, distributes tasks and connects systems to deliver complex applications with precision and consistency.
More integration, smoother delivery and digital products built end to end to operate intelligently.
Application modernisation with AI
Turns legacy systems into agile, scalable platforms ready for what comes next. AI accelerates refactoring, optimises architecture and supports cloud migration securely and efficiently, and embeds intelligence as the results engine in the new software.
Better performance, lower cost and applications that evolve at the pace of your business.
02 · HyperAgents · The pipeline
One agent per role, end to end.
One agent's output is the next one's input. Backend knows the DBA's schema, Frontend knows Backend's API, QA knows the Architect's rule. That shared memory is what separates “AI that writes a snippet” from “AI that builds a system”.
01
Architect
defines the blueprint
▸
02
DBA
creates the schema
▸
03
Backend
exposes the API
▸
04
Frontend
consumes the API
▸
05-07
QA · Docs · Deploy
tests, documents, delivers
Shared memory (agent chaining): each agent works on the previous one's blueprint, not on a guess.
The seven roles in the pipeline
01Autonomous
Architect
Coordinates the project and defines the blueprint. Translates the business objective into technical decisions: domain model, layers, contracts between modules and the standards the others follow.
02Autonomous
Database
Builds the data foundation. Modelling, schema, indexes, migrations and referential integrity, on the Architect's blueprint rather than on a guess.
03Autonomous
Backend
Implements the logic and the APIs. Business rules, services, endpoints, authentication and error handling, to the project's standard rather than a tutorial's.
04Autonomous
Frontend
Implements the interface and the consumption. Screens, components and integration with the APIs Backend has just exposed. No mismatch between who publishes and who consumes.
05Assisted
Quality
Tests what was built. Coverage generated alongside the feature, not three sprints later. Unit, integration and edge cases.
06Assisted
Documentation
Records what exists. Documentation born with the code and updated when the code changes. No gap between what is written and what runs.
07Assisted
Delivery
Puts it into production. Integration with the CI/CD pipeline, impact analysis and deploy. And from there the cycle starts again.
The proof
An entire ERP, in production.
We have already built a complete ERP with this pipeline, among the most complex software there is. If it handled that, it will handle your product.
03 · The factory flow
Who does each step: AI, process and the dev team.
The AI drives execution; the dev team co-acts at the points of judgement; and everything runs inside a process with decision gates.
IA
Processo
Dev team
Return / rework
04 · People
AI lifts the team. It does not replace it.
People still deliver: AI gives each of them a superpower. No new headcount, no swapping people for AI.
The existing team, amplified
Each professional produces more and decides better. AI is a layer, not an isolated department.
The developer becomes production manager
From executor to orchestrator: they command the pipeline, review and approve what the AI produces, and decide the architecture.
The promise to your team
Nobody is swapped for AI.
Every person is multiplied by it.
The squad: every role amplified by AI
Product Architect Experience Designer Architect / Tech Lead AI Engineer / Orchestrator Platform / DevOps QA / Evaluator
● with the client · ● delivery engine
05 · Process
Governance of the system, not only of the delivery.
Legacy systems became black boxes because they were born without a method: no documentation, no traceability, no governance. The CIT method makes sure nothing is born that way again.
1
Discovery
where the value is
2
Design
how it will work
3
Build
builds and integrates
4
Delivery
safe production
5
Operations
maintains and evolves
Living memory: knowledge is not wiped between deliveries. Each phase consults the base before acting and updates it at the end. What one delivery learns feeds the next.
The line that justifies the investment
This is what an AI licence will never deliver on its own.
Frequently asked questions
Common questions about the Factory.
Is the generated code mine?
Yes. The pipeline produces real source code, in your stack and your repository. No proprietary runtime: the software keeps running without depending on the platform to exist.
Does this replace my development team?
No. AI lifts your team: each person moves to orchestrating, reviewing and deciding, producing far more. Nobody is swapped for AI; the existing team is amplified.
How do I know the software will work?
Quality and governance from day one: tests generated alongside the feature, traceability for every decision and human approval at critical points. And we have proven the pipeline on an ERP in production.
Have you actually used this?
Yes. We built a complete ERP, in production, with the agent pipeline. This is not a pilot or a weekend proof of concept.
Do I need to build an AI team to run it?
No. You hire the service: our team leads, the agents build, you receive working software. No prompt engineering, no learning curve on your side.
Does it run on my infrastructure?
Yes. Modular, on premise or SaaS, with whichever LLM you choose, cloud or open. The data does not have to leave your environment.
What is an AI powered software factory?
It is a development pipeline where AI agents carry out the build steps, from requirement to test, under human direction and review. It is not an assistant suggesting snippets: it is the whole delivery cycle instrumented, with traceability for each decision and a human approval point wherever risk demands one.
Is it faster than traditional development?
Yes, and the gain is not in typing faster. It is in removing waiting: specifications that turned into ambiguity, tests postponed, documentation that never happened. The pipeline produces those artefacts alongside the feature, so what usually delays delivery stops being a queue.
How much does it cost to build a system with AI?
Cost is sized by scope and integration, not by hours of typing: number of modules and business rules, systems to integrate, compliance requirements and the quality level contracted. The initial design defines that boundary before any commitment.
How does contracting work, per project or per squad?
Both exist and serve different needs. Fixed scope fits when the outcome is clear and requirements are stable. A dedicated squad fits when the product will evolve continuously and priorities shift each cycle. The criterion is how predictable the scope is, not commercial preference.
Can you modernise a legacy system?
Yes, and it is one of the uses where the pipeline pays off most. Agents read the existing code, recover business rules nobody documented and help rewrite it in parts, with the legacy system still running. Modernisation by stages, without the full rewrite that so often sinks halfway.
Do you maintain the software after delivery?
Yes. Sustainment can stay with Central IT or move to your team, and both exits are planned from the start. Since the code, the tests and the documentation are yours, there is no forced dependency: continuity is a choice, not a contractual lock in.
How do I buy this?
Through four routes: direct sales, AWS Marketplace, SERPRO and TELEBRAS. Public bodies often find the shortest path through the state owned companies they already contract with. The detail of each route is at centralit.com.br/en/how-to-buy.
Next step
Start small. Scale with method.
1
Modular
Engage it in parts, on premise or SaaS, with whichever LLM you want.
2
Pilot
One squad and one project to prove the model with real metrics.
3
Scale
Replicate the method across the other teams, keeping governance in place.
Intelligent software,
delivered with governance.
Bring a real project. We set up the pipeline and show you the first module running.
Start with a pilot ▸