Welcome to the VINCISGPT Help Center.
Here you can learn the basics of using VINCISGPT, account and permission rules, enterprise knowledge management, AI Employee collaboration, the research and evidence mechanism, and how to resolve common issues.
VINCISGPT is under continuous development. The features actually available may differ depending on your account, version, deployment model and system configuration.
01Quick Start
Create an account and sign in
Sign in to VINCISGPT using a valid login method. Keep your credentials safe and never share verification codes, access tokens or account information with others.
After signing in, the system provides features based on your verified identity and permission records. An email address, display name or role label shown in the interface does not by itself create administrative rights.
Start your first conversation
In Desk, describe what you would like VINCISGPT to help you with, for example:
- research an industry or market;
- organise enterprise materials;
- analyse a document;
- draft a plan or a piece of content;
- create a Work item;
- query existing enterprise knowledge.
For better results, state your goal, background, constraints and the deliverable format you expect.
Verify AI output
AI output may contain errors, omissions or inferences. For material business, legal, financial, medical, safety or other high-risk matters, verify independently and consult a suitably qualified professional.
A statement by AI that a task is “completed” does not mean the task is actually complete. Where proof of execution is required, rely on the recorded Evidence, external read-back results and Outcome.
02Core Features
Desk
Your main entry point for working with VINCISGPT — ask questions, state goals, add materials and track progress.
Console
View system status, work progress and authorised administrative information.
Knowledge
Store and manage enterprise knowledge, separating confirmed facts from unverified information.
Content
Organise, prepare and manage content work, including pre-publication confirmation and approval.
Desk
Desk is the primary place to communicate with VINCISGPT and initiate work. You can raise questions, state goals, supply materials or check task progress. Depending on the task type, the system may decompose the question, retrieve information, organise evidence or plan the work.
Console
Console is used to view system status, work progress and authorised administrative information.
Sensitive functions in Console are available only to identities verified by the system as authorised. Where identity status cannot be confirmed, the system may deny access or temporarily return service-unavailable in order to avoid incorrect authorisation.
Knowledge
Knowledge is used to store and manage enterprise knowledge. Please note:
- conversation content does not automatically become an enterprise fact;
- an AI inference is not formal knowledge;
- model-generated content cannot prove its own accuracy;
- a formal fact should have a clear source, context and an appropriate basis for confirmation;
- information that cannot be confirmed should remain in an unknown state and must not be auto-completed into a fact.
Content
Content is used to organise, prepare and manage content work.
Certain publishing, sending or external write operations may require explicit user confirmation or approval by an authorised person. Without valid confirmation, VINCISGPT does not treat an AI suggestion as authorisation to execute.
AI Employee
An AI Employee participates in work according to the responsibilities and capabilities assigned to it.
Can
- receive and analyse tasks
- propose plans or recommendations
- invoke tools within permitted scope
- execute work that has valid authorisation
- record the execution process and results
Cannot
- impersonate a real person
- acquire Owner rights on its own
- approve tasks for itself
- write model output directly into enterprise facts
- bypass user confirmation or system permissions
- declare a task successful without valid evidence
03Work, Evidence and Outcome
VINCISGPT uses three core objects to manage the truthfulness of work.
Work
Work describes what needs to be done, including the objective, responsible party, status and relevant constraints.
Evidence
Evidence records material capable of proving an operation or result — system records, files, tool receipts or read-backs from an external service.
A model’s own statement cannot on its own serve as proof of successful execution.
Outcome
Outcome is the system’s determination of a work result based on valid evidence. Common outcomes include:
“No data found” and “the system failed to read” are two different states. Where a read error occurs, the system must not interpret it as an absence of problems.
04Identity, Permissions and Confirmation
VINCISGPT distinguishes between the following concepts:
Execution capability
Which tools an AI Employee or execution agent may use, and what types of task it may handle. Having execution capability does not mean having approval authority.
Request permission
Whether an authenticated user may raise a given request. Making a request is not the same as approving it.
Owner rights
Owner rights derive only from an identity record formally appointed and verified by the system. An email address, role label, organisation name or client-submitted field cannot create Owner rights on its own.
Approval decisions
An approval must come from a real, verified person who has authority to make that decision.
AI, models, system programs and AI Employees cannot substitute for a real person where a human-level approval is required.
05Research and Evidence
When conducting research, VINCISGPT may:
- understand the question;
- break down the information required;
- form a research plan;
- obtain observable material;
- resolve entities and ambiguity;
- organise the evidence;
- reason from that evidence;
- form a supportable claim;
- produce the answer.
Citing a web snippet does not necessarily mean the full page has been verified. For material conclusions, review the evidence source, how it was obtained, and its evidence grade.
Where evidence is insufficient, sources conflict, or the relevant entity cannot be reliably identified, the system may return UNKNOWN, MISSING or CONFLICT rather than forcing a definitive answer.
06Files and Enterprise Materials
Before uploading a file, confirm that you have the right to use and process its content.
Do not upload:
- confidential material you are not authorised to disclose;
- personal data you have no authorisation to process;
- passwords, private keys or complete access tokens;
- malicious code;
- unlawful, infringing or prohibited content.
Where files are stored and how they are processed may differ by deployment model, feature and third-party service. Unless a feature is expressly marked as local processing only, do not assume that all material stays on your device.
07Third-Party Services and Connectors
Some features may connect to model, search, storage, payment, email or other third-party services.
Before using a connector, confirm:
- that the connected account is correct;
- that the authorisation scope is no wider than necessary;
- that the operation target is accurate;
- whether external side effects will be produced;
- whether user confirmation or Owner approval is required.
Where a third-party service fails, times out or returns an indeterminate status, VINCISGPT may halt execution and require a status check to be completed first.
08Frequently Asked Questions
Why was my operation refused?
Common reasons include:
- you are not signed in, or the session has expired;
- your current identity lacks the relevant permission;
- Owner identity status cannot be confirmed at present;
- the request involves an unauthorised external operation;
- user confirmation or Owner approval is missing;
- the relevant capability is not yet connected;
- the system detected a cross-organisation, privilege-escalation or security risk.
Why does it show UNKNOWN?
UNKNOWN means the available material is insufficient to determine the result reliably. It does not mean the task succeeded, and it does not necessarily mean it failed.
Where an external service times out, evidence cannot be read, or a result cannot be independently verified, the system keeps the state as UNKNOWN to avoid duplicate execution or a false declaration of success.
Why does a feature show NOT_CONNECTED?
This means the relevant data source, tool, connector or execution chain is not yet formally connected. The system does not use simulated data in place of real results.
Why did the AI not execute directly?
Some operations require:
- explicit user confirmation;
- approval by an authorised person;
- a configured connector;
- valid execution permissions;
- a recordable and verifiable execution path.
Where these conditions are not met, the system refuses or pauses execution.
Can AI-generated content be used as-is?
You should review it for your specific purpose. AI-generated content may be similar to content received by others and may carry factual, copyright, brand or compliance risk.
How do I report a bug or security issue?
Please provide:
- the time the issue occurred;
- the feature you were using;
- the page or task identifier;
- the error message;
- screenshots where relevant;
- reproduction steps that are safe to share.
Do not send passwords, private keys, complete access tokens or other sensitive credentials through ordinary support channels.
09Account and Data Requests
Subject to the features provided and applicable law, you may request to:
- review your account information;
- correct inaccurate information;
- obtain a copy of your data;
- delete your account or related personal data;
- withdraw a specific authorisation;
- cancel a subscription;
- report unauthorised account activity.
Certain transaction, security, evidence and audit records may be retained for legal, security or dispute-handling reasons.
When submitting a request, use the contact details associated with your account. Note that an email address is used only for contact and identity verification, and does not by itself create Owner or administrative rights.
10Service Status
During maintenance, upgrades, security incidents or third-party service failures, some features may be temporarily unavailable.
VINCISGPT does not label data as live where there is no real connection, and does not label simulated, forecast or test results as formal results that have already occurred.
11Related Documents
Before using VINCISGPT, please also read: