GIIP FDE Ops — Frequently asked questions
Straight answers to the operations, security and contract questions enterprise buyers actually ask.
Product & service
Is GIIP FDE Ops a SaaS or a managed service?+
It is a managed operations service. AI agents (the FDE Ops Engine) and human engineers (the FDE Squad) run development, infrastructure and operations as one continuous system under enterprise governance — not a pure self-service SaaS.
Is GIIP FDE Box a GPU / AI server?+
No. FDE Box is an execution appliance installed inside the customer environment. It connects securely to internal servers, databases, cloud accounts and dev tools and executes work under policy, approval, audit and rollback control — it is not a GPU box that runs AI models itself.
What is the GIIP FDE Squad?+
A team of human experts responsible for high-risk judgement, architecture decisions and emergency escalation — the people side of the responsibility model.
Does GIIP build its own AI model?+
No. GIIP uses existing models (Claude, GPT, Gemini and others) and puts them to work in real enterprise operations. Its core asset is not the model but the closed operating loop: observe → assess → approve → execute → verify → rollback → audit.
How is this different from general AI coding tools (ChatGPT, Claude, Copilot)?+
General tools stop at generating code or advice. GIIP takes that output, deploys it to real infrastructure and operates it — monitoring, incident response, performance and cost optimization included.
Models & data
Which models does GIIP use — Claude, GPT or Gemini?+
GIIP uses a default base model and routes to different models per task. Multiple providers (Claude, GPT, Gemini) are supported.
Can the customer choose or change the model?+
Yes. Customers can specify or switch the model used.
Is customer data sent to external LLM APIs?+
Data sent externally is masked by default, and an on-premise local-inference option is available for customers who require it.
Can it run air-gapped with local-only models?+
Local-only model operation is possible. For best results GIIP recommends Claude as the base model.
Is customer data used to train models?+
No. Customer data is not used for training.
Where is data stored (data residency)?+
Data is stored in a customer-specified location.
Security, control & responsibility
Does the AI access Production directly?+
High-risk Production actions require prior approval. The AI automatically handles state collection, analysis, testing, approved-runbook execution and monitoring, but Production deployment goes through a pre-approval gate.
Which actions require human pre-approval?+
Production deployment, data deletion, permission and firewall changes, exceeding cost limits, and any irreversible change.
Which actions does the AI execute automatically?+
State collection and analysis, testing, execution of approved runbooks, monitoring, and documentation and reporting.
Which decisions do human experts make?+
Diagnosing new incidents, architecture decisions, reviewing high-risk changes, priority and business judgement, and emergency escalation.
Is every action logged and auditable? What about rollback?+
Each action is retained in auditable form together with its policy, approval, execution record and rollback information. High-risk or irreversible actions proceed only with a rollback plan secured in advance.
Who is accountable when an incident happens? What about SLAs?+
Accountability and SLA terms are agreed with each customer and operated accordingly; new-incident diagnosis and emergency escalation are handled by the human FDE Squad.
How are credentials (secrets) and least privilege handled?+
Credentials stay within the FDE Box — the customer’s credentials do not leave the box (optionally kept in AWS Secrets Manager and fetched on user confirmation). Access is governed by audit logs and command logs; BYOK is not currently supported.
Is there an emergency stop (kill switch)?+
Yes. Operation can be stopped remotely, and the FDE Box can also be powered off physically.
Integrations & environment
Can it connect to our existing AWS / Azure / GCP accounts?+
Yes. It integrates with (rather than replaces) your existing accounts. AWS, Azure, GCP and on-premise are all verified.
Does it work with Zabbix / CloudWatch / other monitoring?+
Yes. Verified monitoring integrations include Zabbix, New Relic and CloudWatch.
What OS, on-premise and Kubernetes environments are supported?+
Verified environments include Windows, Linux, Solaris, FreeBSD, HP-UX and Kubernetes — physical and on-premise servers included.
Does it integrate with GitHub / GitLab and Slack / Teams?+
GitHub, GitLab and Slack are verified. Requests are typically driven through Slack.
What is the full list of supported, verified integrations?+
Cloud: AWS / Azure / GCP / on-premise. OS & orchestration: Windows / Linux / Solaris / FreeBSD / HP-UX / Kubernetes. Databases: SQL Server / MySQL / PostgreSQL / Oracle. VCS: GitHub / GitLab. ChatOps: Slack. Monitoring: CloudWatch / Zabbix / New Relic. IaC: Terraform. AI models: Claude / GPT / Gemini. (Verified across istyle and multiple customer environments.)
Adoption & contract
How long does adoption take?+
Standard adoption takes about two months.
How does a PoC work, and how does it become production?+
GIIP takes in the customer’s requirements and proceeds directly without a separate PoC phase, with a refund available if it does not meet expectations.
How is pricing determined?+
Pricing starts from KRW 5,000,000 / month and is based on the number of managed servers, cloud accounts and operated services, support hours, automation scope and FDE Box specification.
What is the minimum contract unit and term?+
The minimum contract term is one month.
If we end the contract, what happens to our data and automation assets?+
Deliverables belong to the customer. In the case of a refund, all deliverables are returned to the provider.
Can GIIP work alongside our existing team or outsourcing vendor?+
Yes. GIIP can collaborate with an existing team and run in parallel with your current setup.