From hype to help: When AI gets down to business (includes Breaking change, v.26) – ERP.BG

Artificial intelligence is no longer a distant promise or a passing buzzword — it’s becoming a practical, reliable tool inside ERP.net. With version 26, we’re moving past the hype and focusing on helping people get real work done.

Through our “Make AI Practically Usable” initiative, we’ve set a clear mission: to transform AI in ERP.net from an experimental advisor into a hands-on productivity partner. This means AI that doesn’t just suggest what to do, but actually helps you do it — from capturing invoices and extracting lead details to supporting conversations and automating tasks.

In this post, we’ll walk you through the key AI improvements in v26 — including a major infrastructure shift that lays the foundation for the future: preparing our systems for MCP-based AI servers. This isn’t just about new features; it’s about making sure AI in ERP.net is ready to scale and support your business for years to come.

“Just Snap the Invoice” – INGEST AI Import

AI-Powered Document Import: INGEST can “swallow” any document – invoices, PDFs, business cards – automatically reading and importing it into ERP.net.
Seamless Mobile Capture: Employees on the go can snap a photo of an invoice; it’s instantly imported and routed to the assigned accountant with the original image attached for easy verification.
Effortless Desktop Use: Accountants can drag and drop or select a file, and the invoice appears in the system, ready for review and posting.
Adaptive Intelligence: Learns from how your team processed previous documents from the same supplier and applies the same workflow, requiring no extra training or configuration.
Part of the “Invisible ERP” Vision: Reduces manual data entry so you can focus on running your business, while the system quietly handles the rest.

(Case 39147)

Boost Lead Creation with AI-Powered Data Extraction

We’ve just made creating new Leads faster and smarter. With our new “Fill data from text or Email (AI)” feature, you can generate a complete Lead record in seconds.

The function is available through the business app Mail, under the name "Create Lead from this Email (AI)".

Here’s how it works: when you’re creating a new Lead, instead of manually filling in all the details, you can choose the AI option.

The system will scan the text you provide—or the content of an email—and automatically recognize key attributes such as:

Lead’s name
e-mail address
Company
Telephone
Title and position
Any other information relevant to form fields

On top of that, the AI creates a clear summary of the text or email and places it in the Notes section. That way, you always have the context handy without having to read through the entire original message.

The result?

Less manual work, fewer errors, and much more time saved. Your team can focus on building relationships and closing deals instead of copying and pasting data.

Simply add the required field Campaign - and you are all set up!

This functionality opens up exciting opportunities:

Efficiency boost – capture Leads from emails with just a click.
Better organization – no more missing details; important data is recognized instantly.
Smarter workflows – your sales team starts conversations equipped with context-rich Notes.

In short, this AI-powered feature turns messy emails and text into structured Leads—helping you work faster, smarter, and more effectively.

(Case 39048)

Create а ToDo (AI) from notifications

Another similar productivity feature: Create ToDo (AI) directly from your Notifications list (under the 🔔 bell icon).

This option works similarly to the Create Task feature found in chat comments. However, with the help of our integrated AI model, this version automatically generates a ToDo task using the notification class of the most recent notification as context.

Create a ToDo (AI) from a comment in the chat

With this feature, the users can instantly turn any comment into a Task with a right-click, thus avoiding context switching—no need to copy-paste or navigate away.

How it works:

Right‑click any comment in a chat group
Choose “Create Task” from the menu

A ToDo modal window opens with prefilled:

- Title – AI-generated, context‑aware, and auto‑translated into the user’s working language, regardless of the language used in the original comment
  
  The chosen AI model must be marked as default.

Other relevant fields also have values

Assigned User and Owner user – defaults to the task creator
Data Object – connected to the original topic/item
Social Group – auto-linked if the comment is in a group

3. Click the button Create to confirm the values

Benefits:

Speeds task creation
Reduces manual effort
Keeps workflow within context
Ensures multilingual consistency with AI-powered title generation

After this light exercise, the assigned ToDo is promptly announced in the group chat and enters the list, visible from the tab ToDo. More to this, the group itself is focused, and a series of notifications is triggered so that all members never miss an important group event.

(Case 38911)

Ensuring privacy for аssistant conversations via the security function

Аssistant conversations in ERP.net are now fully secured per user. The system ensures that each conversation is visible only to its creator. This change enhances data privacy and aligns assistant functionality with user-specific work contexts.

(Case 38677)

AI Assistant & Chat Arbiter: System Message when the AI can not respond

When a request to the AI fails due to a technical issue or timeout, the system now shows a clear, user-friendly message instead of leaving the user waiting without feedback.

Immediate Feedback: Alerts the user that something went wrong and suggests retrying or contacting the system administrator.
Access to Error Details: A “Click here” link lets authorized users (or admins) open Informational Messages to view the technical error details for troubleshooting.
Better User Experience: Prevents confusion when the AI appears to “stay silent” and helps users know what to do next.
Improved Support & Maintenance: Stores the error information so admins can investigate issues even after the incident.

(Case 38785)

AI Server Availability Check for AI Assistant & Arbiter (Breaking change, v.26)

When opening the Discussion panel (with a virtual user in the group) or the AI Assistant panel, the system now verifies if an AI Server site is running.

This change is more than a technical safeguard — it’s a strategic step towards the future of AI infrastructure. MCP (Model-Control Protocol) servers are emerging as the standard backbone for modern AI systems. By enforcing AI Server availability now, we are:

Laying the groundwork for a seamless transition to MCP-based services.
Ensuring compatibility with upcoming AI features that rely on more advanced server orchestration.
Reducing future technical debt, as systems will already be configured to align with MCP-driven architectures.

In short, this is not just a patch for v26; it’s future-proofing your environment for next-generation AI capabilities.

What You Need to Do After Upgrading to v26

Create an AI Server site.
This ensures the AI Assistant and Arbiter can operate without interruption.
Contact your system administrator if you encounter any issues creating or starting the AI Server site.

(Case 38924)

AI Chat Arbiter: Extended Context to 100 Messages per Day
The chat arbiter now considers up to 100 previous messages from the same day when processing a new user query (previously limited to 10).

Richer Contextual Understanding: Enables the AI to deliver more relevant and coherent answers by leveraging a broader conversation history.
Configurable via System Key: Controlled by the new key /AI/Chat/ContextDepth, set to 100 by default, allowing adjustments if needed.
Backward-Compatible Update: Existing custom keys have been standardized with a leading “/” and migrated automatically, minimizing setup effort.
Improved Conversation Flow: Reduces repetitive clarifications and keeps the discussion consistent across longer same-day sessions.

(Case 38751)

AI Assistant: Use User-Visible Field Names Instead of System Names

The AI Assistant now understands and responds using the same field names users see in the ERP.net interface, improving clarity and usability.

User-Friendly Language: When exporting data for the current record to the AI Assistant, it now uses the translated field names as shown in the UI, instead of internal system names.
Language-Aware: Automatically picks the translations of the active ERP.net language (e.g., Bulgarian, English, etc.), ensuring consistent communication.
Respects Custom Labels: If a field has a custom label defined in the current view, that personalized name is sent to the AI, taking priority over standard translations.
Better Understanding: Makes the AI’s answers more intuitive and relevant for end-users without requiring them to know technical or backend terminology.

(Case 38744)

AI Assistant: Improved Debug Mode with Raw Data Logging & Clear Toggle

The Debug Mode for the AI Assistant has been significantly enhanced to make diagnostics easier and more transparent.

Raw Request & Response Logging: When Debug Mode is enabled, the log now includes the full raw request and response content exchanged with OpenAI, alongside the token usage statistics.
Restored Token Summary: The total token usage is displayed at the top for quick reference.
Cleaner, User-Friendly Presentation: Reduced heading font sizes and removed unnecessary icons for a more readable log layout.
Switch-Style Toggle: The Debug Mode button now includes an intuitive ON/OFF switch, making it immediately clear whether the mode is active, instead of relying on subtle indicators like bold text.
Consistent UI & Better UX: Uses the same switch component as the “Customize Form” panel in the web client, improving consistency and preventing confusion.
Enhanced Diagnostics: Captures detailed request/response data at the lowest operational level, supporting efficient troubleshooting and issue resolution.

(Case 38746 & Case 38761)

dataurlimage (1).png

AI Assistant: Distinct Colored Background

To reduce confusion between the AI Assistant panel and the regular Discussions panel, the AI Assistant’s side panel now has a subtle bluish primary background instead of plain white.

Clear Visual Separation: Makes it obvious when users are interacting with the AI Assistant versus standard discussions.
Theme-Friendly Design: Works well in both light and dark modes, ensuring readability and a consistent look.
Enhanced Readability: Added subtle borders around comments and avatars so text remains clear even on the colored background.
User-Focused UI: Prevents misclicks and improves overall navigation experience in the ERP.net web client.

(Case 38762)

AI Configuration: Dropdown for Base Model Selection in Providers

Configuring AI definitions in ERP.net is now simpler and less error-prone thanks to an improved Base Model Name field.

Auto-Populated Dropdown: The Base Model Name field in the Providers definition now offers a dropdown list of available OpenAI models, eliminating the need to type the model string manually.
Reduced Errors: Minimizes typos and mismatched names by providing predefined, valid options.
Flexible Input: Still allows manual entry of a custom value if needed, supporting advanced or future use cases.
Streamlined Setup: Makes configuring AI providers faster and more reliable for administrators and advanced users.

(Case 39071)