A research prototype from Hochschule Harz

Clarity across
your business information.

Meetings, emails, and company knowledge—connected through an agent-based AI Executive Assistant designed to help managers find what matters and move work forward.

Designed for managerial support.
Built transparently as a thesis prototype.
THESIS PROTOTYPEThree core modules in focusOne broader intelligence visionSee the research scope
01THE CHALLENGE

Business context is everywhere.
Finding it shouldn’t be.

Decisions draw on conversations, documents, and the history behind them. Yet that context often sits in separate tools, leaving managers to search, summarize, and reconnect the pieces themselves.

My Intelligence explores an assistant that can bring relevant information together—while keeping people in control of what they do with it.

A

Scattered context

Important details live across messages, meeting records, and files.

B

Follow-ups take effort

Actions, decisions, and deadlines can be hard to keep visible.

C

Organizational memory fades

Past information is valuable only when people can retrieve it.

02THE INTELLIGENCE SUITE

Four lenses on
business information.

The first three form the thesis prototype’s core focus. Executive Insights & Business Intelligence describes the wider direction of the system.

02Fictional sample · no transcript ingestion

Meeting Intelligence

The dashboard shows fictional meeting notes. Transcript ingestion and AI extraction of decisions or action items are not implemented.

INPUT Meeting transcripts → OUTPUT Decisions & follow-ups
03Sample screen · keyword search exists

Knowledge Intelligence

Workspaces can store selected files and search extracted text by keyword. The demo screen uses fictional sources; vector retrieval and generated answers are not implemented.

INPUT Documents & history → OUTPUT Relevant answers
04Broader direction

Executive Insights & Business Intelligence

A future direction: connect information across the organization to support managerial understanding and decisions.

INPUT Connected business context → OUTPUT Executive insights

i These labels separate existing code paths, fictional dashboard samples, and proposed research architecture. Google OAuth configuration and approval have not been verified in this review.

03HOW IT WORKS

From separate sources
to shared context.

The complete flow below is proposed architecture. Live code paths are summarized separately so the concept is not mistaken for a deployed RAG or multi-agent system.

PROPOSED RESEARCH ARCHITECTURE

  1. 01 · DATA SOURCESGmail · meeting transcripts · company documentsOpt-in, workspace-selected information
  2. 02 · PROCESSINGParsing and information extractionNormalize content; identify entities, decisions, dates, and actions
  3. 03 · ORCHESTRATIONAgent coordinationProposed planning and tool-routing layer
  4. 04 · KNOWLEDGERepository and RAGProposed retrieval with source evidence and workspace boundaries
  5. 05 · INTERFACEExecutive assistant dashboardContext, citations, and human review
  6. 06 · OUTPUTSSummaries · answers · decisionsAction items and follow-ups for people to review
CURRENT SOURCE-CODE STATUSOperational provider configuration not verified

Implemented paths

  • Account email verification and password sign-in
  • Gmail and Outlook email analysis after consent
  • Google Calendar read-only route for up to 25 events in the next 90 days
  • Workspace Action Hub, encrypted file storage, and on-demand keyword search
  • Telegram bot webhook for messages delivered to that bot

Not implemented

  • Meeting transcript ingestion and analysis
  • Agent orchestration and vector-based RAG
  • AI-generated answers across email, meetings, and documents
  • Yahoo, WhatsApp Business, Slack, and Teams connections

Interpretation: The current dashboard’s Meeting, Knowledge, and cross-source panels use fictional sample data. The code paths above still require the relevant provider configuration and the customer’s own authorization.

04THE PROTOTYPE TODAY

Start with one
useful signal.

The deployed dashboard contains account and source-connection routes. Whether Gmail or Calendar can connect depends on server OAuth configuration, provider approval, and your own authorization. Meeting Intelligence uses fictional notes. Private workspaces support keyword search over stored documents and show a direct roll-up of connected sources.

Open the dashboard
05PRIVACY, ACCESS & LIMITS

People stay
in control.

These are the controls visible in the current code. They are not a security certification or a claim of regulatory compliance.

01 · ACCOUNT

Website sign-in

Sign in with your My Intelligence account email and password. Provider accounts are separate. The account API verifies email and uses hashed passwords and secure, HTTP-only session cookies.

02 · SOURCES

Provider permissions

Gmail analysis requests read-only access. Sending a Gmail reply requires separate authorization and approval of the specific draft. Calendar access is read-only.

03 · DATA USE

Explicit email consent

After consent, recent Gmail or Outlook messages may be sent to the configured AI service for analysis. The public demo uses selected analyses published by the owner of the demonstration workspace.

04 · OVERSIGHT

Human review

Messages are analyzed read-only. You choose whether to add extracted email actions to the Action Hub. No autonomous email sending or calendar changes are implemented.

Prototype boundary: Some legacy endpoints still use a single-user session and global storage, and the account service has not had a full production security review. Do not interpret this site as GDPR compliance, enterprise-grade security, or complete tenant isolation.

06THE RESEARCH

A thesis about
useful intelligence.

A prototype-based Master's thesis at Hochschule Harz, developed in the context of collaboration with New Digital Intelligence (NDI).

The thesis explores whether a context-aware assistant can help managers find relevant organizational information and follow through on work. The website separates the current prototype from the broader proposed architecture.

PRIMARY RESEARCH QUESTION

How can an assistant help managers retrieve business context and follow through on work—while making its benefits and limitations visible?

RESEARCH OBJECTIVE

Develop a bounded prototype that demonstrates selected information-access and follow-up workflows, then assess its usefulness and limitations for managerial support.

PROTOTYPE DEVELOPMENT

The codebase contains a deployed website, account/workspace flows, selected provider routes, an Action Hub, and explicitly fictional meeting and knowledge examples. It does not contain a complete multi-agent or RAG implementation.

METHODOLOGY STATUS

The reviewed project files do not document an approved study design, participant sample, or completed evaluation. No user-study results or effectiveness claims are reported here.

CANDIDATE EVALUATION CRITERIA · TO BE CONFIRMED
  • Relevance and source-grounding of retrieved context
  • Completeness and correctness of extracted decisions and actions
  • Task success and effort for finding context or follow-ups
  • Perceived clarity, trust, and user control
These are proposed evaluation dimensions, not completed measurements.
ACADEMIC CONTRIBUTION

A transparent prototype and research framing for connecting business context with managerial follow-through. The contribution is a design artifact and evaluation opportunity, not validated evidence of organizational impact.

LIMITATIONS
  • Meeting transcript processing is not implemented.
  • Document search is keyword-based; vector retrieval and RAG are not implemented.
  • Executive Insights summarizes available workspace data. AI-generated cross-source conclusions are a research direction.
  • OAuth readiness and full workspace isolation have not been verified.
HOCHSCHULE HARZ · MASTER'S THESISBack to top ↑

MY INTELLIGENCE

Bring the pieces
into perspective.

Explore the four intelligences