Products
Products
Two products, two architectures, one loop. Agenticos designs the agent team, CompanyBrain orchestrates it every day — and both run either in the cloud or on your own infrastructure. As the loop runs, a living digital twin of the organization emerges.
CompanyBrain in 97 seconds
Agenticos
From process to governed AI agents, in a single session.
Turns your process document into a workflow, a role-based agent team, EU AI Act-referenced governance notes and a test plan. The design stage of every new company brain happens here. In the cloud architecture it runs in the browser; in the local architecture it is installed on your own infrastructure.
Who it is for: Process owners, enterprise architects, governance and compliance teams.
CompanyBrain
Turns your organization into a company brain — whatever your sector.
A multi-agent decision-intelligence layer that sits on top of your existing systems. The agent team sees, understands, reasons, acts according to risk, verifies the outcome and converses in plain language — and a living digital twin of the organization emerges. It runs in the cloud or on your own infrastructure; final approval stays with a human.
Who it is for: Any organization whose data is scattered across systems, and new companies building their brain from scratch. Sector-specific company brains: ScholarLens (higher education) and Maarifetname (K-12 education).
A multi-agent platform
CompanyBrain is not a single assistant but a team of agents with separate roles. Agenticos designs the team — which agents exist, what each does, what it may not do, to whom it hands off — and CompanyBrain orchestrates it.
| Role | What it does | Authority tier |
|---|---|---|
| Query and analyse | Queries data in plain language and answers with real numbers | Flexible — read-only |
| Create and update | Opens and updates records; asks follow-up questions when information is missing | Constrained — every write needs human approval |
| Guide | Explains the platform and processes, onboards new team members | Flexible — informational only |
| Model management | Decides which model runs where, cloud or local | Governed — administrator setting |
| Organization-specific specialists | By sector: maintenance, quality, climate control, SLA risk, root cause | According to the autonomy envelope designed in Agenticos |
Agent-to-agent: one agent’s finding becomes another’s input — the agent that finds the root cause hands the evidence to the agent that proposes the action. Human-to-agent: every agent has a human owner; approval gates and escalation paths are part of the team.
The create-and-update agent submitting every write for approval is Risk-Proportional Determinism running in code.
Role names come from the first instance — in Indigonix’s own operation these agents run as Cornflower, Indigo, Arctic and Midnight. In every organization the team is redesigned in Agenticos around that organization’s processes.
A living digital twin of the organization
As the loop runs, a living digital twin of the organization — or of the system — emerges. A twin of a production line, a farm or an energy grid; and at the same time a twin of its processes, roles and decisions.
Most digital twins model an asset’s state. This one models decisions too: which decision was made, on what evidence, when, and under whose approval. It is a twin not only of the organization’s state but of its reasoning. Because every verified outcome is written back, it does not stay static; it learns with the organization.
The twin is the ground on which an action is compared with alternatives and tested in shadow mode before it touches the real world. The workflow and agent architecture produced by Agenticos form the twin’s process skeleton.
How they work together
- Design — AgenticosFrom a process document: workflow, agent architecture, authority limits, governance notes and a test plan.
- Operate — CompanyBrainThe designed agents run on top of the organization’s existing systems, through the six-stage loop.
- Improve — feedbackVerified outcomes and operational data flow back into the next design round; processes are redesigned as the organization learns.
The two products are designed to work together in either architecture. Where process documents are sensitive too, design and operation are installed together in the local architecture.
Two architectures: Cloud and Local
Both products come in two separate architectures. Which one fits is decided not by the product but by how sensitive the data is and which regulation the organization falls under.
| Cloud architecture | Local architecture | |
|---|---|---|
| Agenticos | Runs in the browser, no installation | Installed on the organization’s own infrastructure |
| CompanyBrain | Runs as a hosted platform | Model and engine run on the organization’s own hardware; fully air-gapped if required |
| Where data is processed | The service providers listed in each product’s privacy notice | The organization’s own infrastructure; the organization decides which outbound connections exist |
| Getting started | Immediately | Through an installation project |
| Who it suits | Fast starts, pilots, non-sensitive data | Regulated or sensitive data; critical operations that must not depend on an outside provider |
The principle of the local architecture: bring the model to your data — not your data to the model.
In the local architecture data is not transferred abroad, so KVKK’s cross-border transfer questions do not arise. Other obligations — purpose limitation, retention, access control — still have to be met by the organization’s governance in either architecture. CompanyBrain makes them traceable; it does not make them disappear.
The six-stage decision loop
| Stage | What happens |
|---|---|
| SEE | Data is collected from existing systems, sensors, documents and historical records. No system is replaced. |
| UNDERSTAND | The current state of assets and processes is inferred in a shared time-and-entity context. |
| REASON | An event and knowledge graph plus specialist agents produce the cross-domain root cause and alternative decisions. |
| ACT | By risk level: a recommendation, a human-approved action, shadow mode, or constrained automation within safe limits. |
| VERIFY | The effect of the action is measured against data and the outcome is written to organizational memory. |
| CONVERSE | Managers and teams ask in plain language — and can see which evidence each recommendation rests on. |
“What evidence is this recommendation based on?” Most AI assistants cannot answer that question. CompanyBrain answers it by design: every recommendation arrives with the data and events behind it.
Design principle: Risk-Proportional Determinism
We do not try to make intelligence deterministic. We make its operational consequences deterministic, bounded and auditable.
Probabilistic intelligence inside, deterministic boundaries outside. The agent’s reasoning space is kept wide: it searches for root causes, generates alternatives, builds plans. Where it touches the world, rules apply: which command it may call, within which parameter range, under which condition approval is required, which actions are forbidden, when an action is rolled back, and what the deadline is.
| Consequence severity | Autonomy | What enforces the boundary |
|---|---|---|
| Low | Flexible autonomy — the agent explores, recommends, executes | Monitoring and logging |
| Medium | Governed autonomy | Policy checks and continuous monitoring |
| High | Constrained autonomy | Approval gate — nothing executes without human approval |
| Safety-critical | Deterministic execution | Real-time controller and hardware interlocks |
Make the rules too tight and the system stops being an agent and becomes a workflow engine; make them too loose and it becomes unpredictable. The balance fits in one sentence: adaptivity decreases as consequence severity increases.
Risk-Proportional Determinism is the design principle defined for agentic and autonomous systems by Indigonix founder Dr. Damla Aslan; Agenticos and CompanyBrain are built on it. Full definition: What is Risk-Proportional Determinism?
The agent never touches the system directly: a five-layer execution stack
| Layer | What it does | Who owns it |
|---|---|---|
| Agentic intelligence | Reasons, plans, generates alternatives | CompanyBrain — REASON |
| Governance and safety | Policy, authority limits, approval gate, risk checks | Designed in Agenticos, enforced in CompanyBrain |
| Execution orchestrator | Dispatches only permitted commands, within permitted ranges | CompanyBrain — ACT |
| Deterministic execution boundary | Timing guarantees, interlocks, emergency stop, rollback | Depends on the system — see the table below |
| System | Enterprise software or a physical agent | ERP, CRM, field systems — or a cobot, humanoid, AMR, drone |
The rule: an agent never writes directly to a system of record or to an actuator. Every action passes through the safe execution layer, is logged, and can be rolled back where needed.
Who provides the deterministic boundary?
The deterministic execution boundary is a role, not a technology. Not every system has a real-time operating system (RTOS); depending on the system, a different layer fills the role:
| System type | What provides the boundary |
|---|---|
| Embedded controllers running an RTOS (e.g. drone flight controllers, motor drivers) | The RTOS — timing guarantees at hardware level |
| PLC-based industrial systems | The PLC scan cycle and safety interlocks |
| General-purpose operating systems (e.g. ROS 2 on Linux) | No timing guarantee: the boundary lives in software — command allow-list, parameter ranges, timeouts, rollback — plus a hardware emergency stop |
| Software-only agents | The boundary sits entirely in the policy layer — API allow-list, approval gate, rollback |
The strength of the boundary sets the ceiling on autonomy. A system without a real-time deterministic controller does not execute a safety-critical action on its own; that action stays at the approval gate. Two things set the autonomy level together: how severe the consequence is, and how much determinism the execution boundary actually provides. Agenticos designs that ceiling for every agent.
Embodiment-agnostic: from software agent to humanoid
The loop does not care what body the agent has. A software agent, a cobot, a humanoid, an AMR or a drone all connect through the same five-point contract:
- What can it sense? — SEE
- What can it do? — ACT
- What is its autonomy envelope? — what it does alone, what needs approval
- Who stops it, and how? — the override path
- How is the outcome verified? — VERIFY
Agenticos designs the contract; CompanyBrain operates it. Only the content of the contract changes.
Physical agents run in the local architecture, at the edge: on a production line or in a barn they must keep operating safely when connectivity drops, and a safety decision cannot wait for the cloud. For controllers whose inner workings cannot be inspected, governance shifts from explaining the decision to constraining the action and verifying the outcome — which is exactly what Risk-Proportional Determinism is for.
Enterprise-grade: traceability, access, integration
- End-to-end traceability — in the first instance a request is traced from ticket to backlog, to code and to release; which customer request shipped in which release is known.
- Role-based access and data scoping — separate role tiers for internal staff, customers and partners; every organization and user sees only the data they are entitled to.
- Audit trail — every login, change and administrative action is recorded with who did what and when.
- Integrations — email to records, GitHub issues and releases, marketplace orders, evidence-grounded answers over a knowledge base; field systems and sensors depending on the sector.
AI capabilities — roadmap
The capabilities below are on the roadmap. Each carries a risk rule — Risk-Proportional Determinism at the level of a single capability.
| Capability | Risk rule |
|---|---|
| Automatic classification and prioritisation | Human confirmation required |
| Evidence-grounded answers from similar records and the knowledge base | Sensitive customer data is redacted |
| SLA breach risk prediction | Escalation not fully automated at first |
| Defectiveness prediction | Model limitations and bias documented |
| Root-cause clustering | Human validation required |
| Governance tagging (EU AI Act, ISO/IEC 42001) | Explainability and auditability mandatory |
The same loop for every sector — real instances
The loop does not depend on the sector; what changes is which systems are connected and where the organization’s risk limits sit. Products such as ScholarLens and Maarifetname are not separate software but sector-specific company brains built on CompanyBrain. The same method is turning organizations into company brains in these sectors today:
| Sector | Company brain | What it does | Status |
|---|---|---|---|
| Customer service and delivery | Indigonix’s own operation | Brings tickets, SLA, backlog, GitHub and email together; sees SLA risk, proposes replies and escalations — writes are human-approved | Live |
| Higher education | ScholarLens — research intelligence | On top of the institution’s own bibliometric subscriptions (e.g. Scopus, SciVal): researcher identity, publication analytics, benchmarking, collaboration and opportunity discovery; plain-language reporting for leadership | First deployment: a private university |
| K-12 education | Maarifetname — edtech | Reads handwritten exams with OCR and proposes scores against a rubric — the teacher approves the grade; turns textbooks into a structured knowledge base | In development |
| Agriculture and livestock | A robotic dairy farm | Brings milking robots, herd management, climate control, cameras and sensors together; cross-domain root cause and guarded action | R&D project |
| Logistics | With a real-time modelling & simulation engine | A company brain that brings logistics operations data into a shared context, running with a real-time modelling & simulation engine | In development |
| Civil aviation | With a real-time modelling & simulation engine | AI works at the recommendation and approval tiers; safety-critical flight functions stay in certified, deterministic systems | In development |
| Production | Risk/non-conformity detection & root cause analysis | Brings production, quality and sensor data together; detects risks and non-conformities early, analyses the root cause across domains and proposes corrective action — approval stays with a human | In development |
| Defense | Governed C4 & decision support | Brings data from command, control, communications and computer systems into a shared operational picture; presents options and risks with their reasoning — decisions and authority always stay with humans | In development |
Institution names are withheld for confidentiality. The same method applies to sectors such as energy, healthcare operations and finance; in every organization the loop is redesigned in Agenticos around that organization’s own systems, risks and authority limits.
The line behind it
CompanyBrain did not start as a startup idea but as the outcome of a long research line: from software defectiveness prediction research in 2012, to a product management information system in 2020–2021, to agile software engineering methodology, and to an AI-native platform in 2026. Its first instance was built for Indigonix’s own customer service and delivery operation, where a request is traceable from ticket to backlog, to code and to release.
Where to start?
If the scope of agentic transformation is not yet clear, start with Agenticos: it turns a single process document into a working design and makes the scope concrete. If the scope is clear and you want day-to-day operations supported by decision intelligence, move to CompanyBrain. The architecture is decided by the data: not sensitive, cloud; sensitive, local.
Frequently asked questions
Agenticos or CompanyBrain — which should I start with?
If the scope of agentic transformation is not yet clear, start with Agenticos: it turns a single process document into a working design and a role-based agent team, and makes the scope concrete. If the scope is clear and you want day-to-day operations supported by multi-agent decision intelligence, move to CompanyBrain. In most organizations the two are sequential: design first, then operation.
Should I choose the cloud or the local architecture?
The data decides. For organizations with non-sensitive data, and for pilots, the cloud architecture lets you start immediately; data is processed by the service providers listed in each product’s privacy notice. Where data is regulated or must not leave the organization, the local architecture — model and engine on your own hardware — is recommended. Physical agents such as robots and field systems run in the local architecture, at the edge, so they keep operating safely when connectivity drops.
Can the two products be used independently?
Yes. Agenticos can be used on its own as a design tool, and CompanyBrain on its own as an operating layer. Used together, they close the design–operate–improve loop: the agent team designed in Agenticos runs in CompanyBrain, and verified outcomes flow back into the next design round.
Tell us briefly about your systems and where your data is allowed to go; we will work out together where to start and with which architecture.
Send an email Services