Pretoria, South Africa – May 2026

CohesionX’s flagship AI platform, VectorMind, continues to evolve from a powerful GenAI platform, into a fully integrated enterprise intelligence ecosystem; combining secure AI orchestration, advanced automation, and human-centric governance at scale.

Built on VectorMind’s AI Trust, Risk & Security Management (TRiSM) principles and enterprise-grade architecture, the latest platform release introduces major upgrades across usability, governance, observability, and AI capability.

This release places special focus on two platform products: Atlas 2.0 and BusinessGPT 2.0.

Find line / Governed knowledge backboneAtlasMaps your enterprise information into a searchable, governed knowledge backbone.

Atlas 2.0: Enterprise knowledge navigation, reinvented

Atlas has always been more than an AI search function. It orchestrates semantic, metadata, keyword, and contextual discovery across enterprise knowledge ecosystems using graph-based AI reasoning.

With Atlas 2.0, the platform now delivers a dramatically more scalable and manageable enterprise discovery experience.

Key Enhancements in Atlas 2.0

Atlas Easy Config

Organisations can now rapidly deploy AI-powered knowledge bots across multiple repositories with minimal configuration.

New capabilities include:

  • Multi-site SharePoint bot creation
  • Microsoft Graph-powered site resolution
  • Dropbox integration within the same onboarding flow
  • Namespace-scoped project lookups to eliminate cross-tenant collisions
  • Improved monitoring through asynchronous bot creation tracking and processor nudge logic

The release also resolves several operational bottlenecks, including:

  • SharePoint duplicate validation issues
  • Filter date training inconsistencies
  • Background service display name preservation
Create Atlas AI Service screen with SharePoint and Dropbox source options
Figure 1 – Atlas new easy config

Atlas Edit: Live Bot Management

Atlas 2.0 introduces a fully-fledged edit and management experience for deployed bots.

Administrators can now:

  • Add or remove SharePoint sites dynamically
  • Rotate Dropbox credentials without rebuilding bots
  • Resolve new sites directly from the edit flow
  • Monitor bot progress directly from the platform
  • Test bots through the new Office taskpane integration

The result is a more agile enterprise AI deployment model designed for operational scale.

Most importantly, Atlas remains governed by VectorMind’s embedded AI guardrails, ensuring enterprise knowledge discovery stays secure, explainable, compliant, and grounded in trusted organisational data.

BusinessGPT 2.0: A complete rebuild of the chat experience

BusinessGPT 2.0 is an enterprise chat experience designed around one powerful idea: you should be able to shape your assistant, not the other way around.

Act line / Closed-loop gen AIBusinessGPTA closed-loop gen-AI app for data sovereignty, compliance and control.
BusinessGPT 2.0 chat screen with suggested prompts for chart, image and web search tools
Figure 2 – Business GPT 2.0

At the centre of BGPT 2.0 is the new Assistant Builder, a guided, visual setup experience, helping users to configure an assistant’s model, tools, document processing pipeline, vector store, and embedding model in one place.

Document intelligence has also been rebuilt. BGPT 2.0 now supports smarter extraction strategies for different document types, including optional vision-capable models for scans, images, diagrams, and complex PDFs. Users can see live extraction progress as documents are classified, extracted, embedded, and prepared for querying.

The platform also introduces Tools such as web search, chart generation, image generation, and image analysis with suggested prompts that automatically match the enabled toolset. For supported models, reasoning can now be surfaced inline, helping users understand how an assistant reached its answer.

Tool Registry screen listing active Chart Gen, Image Gen and Web Search tools
Figure 3 – Business GPT Tool Registry

Power users can switch between chats across multiple assistants, swap models per message, move between Builder and Chat in one click, and rely on auto-generated thread names to keep conversations organised.

Most importantly, BGPT 2.0 is now hot-swappable end to end. Teams can configure the document processing model, vector store, embedding model and dimensions, LLM provider, max tokens, and reasoning effort per assistant.

Visual workflow linking a BusinessGPT assistant to tools, LLM providers and a RAG document store
Figure 4 – Business GPT AI visualised workflow

This means organisations are no longer locked into a single AI stack. They can curate the exact assistant experience their teams need: all while operating within VectorMind’s enterprise guardrails for security, observability, compliance, and control.

VectorMind proudly helps organisations adopt AI and operationalise trusted, enterprise-grade intelligence at scale.

Want to take control of AI in your business? Mail us: info@cohesionx.co.za

Are you a user and need support? Log a ticket: support@cohesionx.co.za