Discover Agent
- Capability
- Discover
- Produces
- Scope report, gap/fit analysis, Blueprint draft
- Boundary
- Produces the scope and gap/fit draft; approval is the consultant’s.
MindDX PlatformAI and human control
AI agents carry out the work; a human makes the decision and carries the responsibility.
AI speeds up the work; a human makes the decision and carries the responsibility. This page explains, together, where in the work AI operates, what it produces, and whose hands that output passes through before it reaches you.
01The path of an output
Until an output reaches you
- 01
AI produces the draft
Information is structured and a first draft is produced. At this stage the output is a proposal, not a decision.
- 02
A human reviews it
A consultant or your authorised user reviews the content, corrects it and validates it in context. Who reviews depends on the type of output.
- 03
Approval is recorded on the document, with who and when
Where an output needs approval, the document records who gave it and when.
- 04
A change reaches your system only after approval
The only path that writes to your system follows a preview, a written approval and a transfer sequence; every step is recorded, and steps that cannot be undone are stated before approval.
The level of review depends on the type of output, and we write it down product by product: measurement reports go through expert consultant review, while some documents your authorised user produces themselves.
02Who does the work
Five execution roles, each with a written boundary
These are the execution roles MindDX publishes. Each card says which capability the role serves, what it produces and where it stops.
Audit Agent
- Capability
- Audit
- Produces
- Findings with their evidence shown; it deliberately produces no score
- Boundary
- It changes none of your settings. On Odoo, it works with read-only methods when access has been separately authorised and established. On Zoho and abas ERP no direct live-system connection is used; the input is structured questionnaires and customer documents, and the consultant makes the evaluation.
Design Agent
- Capability
- Design
- Produces
- Configuration guide, a technical design where needed, acceptance scenarios
- Boundary
- The consultant’s approval, when given, is recorded on the document as a stamp with who and when.
Apply Agent
- Capability
- Design
- Produces
- Applied configuration and the project skeleton
- Boundary
- No change reaches your system without your approval: first a preview, then a written approval. Every step is recorded, and steps that cannot be undone are stated before approval.
Support Assistant
- Capability
- Support
- Produces
- Answers from the knowledge base of the product you bought
- Boundary
- It opens no ticket; it redirects. It is an assistant, not an agent: there is no human review per answer.
Checkup, Train and Process are not presented as agents. Checkup scoring is deterministic and the report narrative goes through expert approval; Train documents are produced by the authorised user; Process is run by a consultant with ProcessMind technology.
03Where approval is required
Where approval is required, and where it is not
The same word can mean three different things. So we split approval into three kinds and assign every output to one.
Approval gate
Until approval is given, the output is not final, or the change does not reach your system.
- Checkup report — an expert consultant reviews and approves
- Audit findings — a consultant reviews and approves
- Validated Scope — validated by a human decision
- A change that writes to your system — preview and written approval
- Testing and acceptance — acceptance is given by human approval
Approval stamp
When approval is given, it is recorded on the document with who and when; it does not stop the document being produced.
- Design documents — an authorised user can produce them; the consultant’s approval is recorded as a stamp
- Discover draft — approving the draft is the consultant’s
No human approval
These outputs do not go through consultant review, and we say so plainly.
- Support Assistant answers — from the knowledge base; when it cannot answer, it points to a support agreement
- Train documents and exam — produced by an authorised user; the exam is scored deterministically
04What agentic means
«Agentic» does not mean fully autonomous.
As a category, MindDX Platform is an agentic digital transformation platform: part of the work is carried out by AI agents under governance. That does not mean unsupervised execution, or a change made in your system on its own.
Agents carry out the work; a human makes the decision and carries the responsibility. We neither promise nor deliver a fully autonomous ERP implementation.
The agents on this page carry out MindDX’s own transformation work. Agents that run in your own systems are a separate solution area and are not offered yet.
Agentic AI solution05Which reasoning
Models do not make decisions.
The agents work with models from leading foundation-model providers. Model choice, parameters and cost are under MindDX control; every AI call is recorded with its owner and capability.
FAQQuestions
Frequently asked questions
No. Agents carry out the work; a human makes the decision and carries the responsibility. The only path that writes to your system follows a preview, a written approval and a transfer sequence, and Audit does not write to your system. We neither promise nor deliver a fully autonomous ERP implementation.
No. No change reaches your system without your approval: first a preview, then a written approval. Every step is recorded, and steps that cannot be undone are stated before approval. Audit makes no change in your system: on Odoo it works with read-only methods; on Zoho and abas ERP no direct live-system connection is used.
No. The level of review depends on the type of output. Checkup and Audit reports go through consultant review and approval; on Design documents the consultant’s approval is recorded as a stamp; Train documents and Support Assistant answers do not go through consultant approval.
No. What you buy is a commercial product: Assess, Implement or Transform in the core transformation journey; Checkup can also be bought on its own, and Process and Support on their own as annual subscriptions. Agents carry out part of this work; the commercial product, the capability and the execution role are separate layers.
Models from leading foundation-model providers. Model choice, parameters and cost are under MindDX control; models do not make decisions.
Let us talk about the Platform on your own transformation.
Validate the scope. See risk early. Clarify the decisions. Start with a validated scope.