HEGLA‑HANIC’s new glass365 platform brings a bold pivot in how flat‑glass manufacturers capture and process customer orders: a Microsoft‑based ERP built on Dynamics 365 Business Central that embeds a customised
AI Copilot tuned specifically for glass workflows to convert emails, PDFs and even
handwritten sketches into draft sales orders, check inventory in real time, and generate quotations — while keeping humans firmly in the approval loop.
Background
The glass fabrication and glazing sector has long been hampered by a classic digital‑operations mismatch: production and nesting systems are highly automated, yet order intake remains fragmented and manual. Orders arrive as plain emails, PDF drawings, spreadsheets, photos of shop sketches, or phone calls transcribed by staff. That fragmentation creates delays, transcription errors, and missed production windows — especially painful where custom shapes, lamination options, edge finishes and tight tolerances are involved.
HEGLA‑HANIC’s glass365 is positioned as a purpose‑built ERP for that environment. Rather than a generic cloud ERP, glass365 combines the modern platform and integrations of
Microsoft Dynamics 365 Business Central with an industry‑tuned Copilot agent that claims to perform document intelligence and order drafting tasks tailored to glass processing. The result is marketed as faster, more accurate order entry and a smoother handoff from sales to MES and cutting schedulers.
Overview: what glass365 promises
- A Microsoft‑based ERP foundation (Dynamics 365 Business Central) for finance, inventory, and sales modules with Office 365 integration.
- An embedded, customised Microsoft Copilot (an agent-style AI) trained and configured for glass‑industry terminology, processes and document types.
- Document intelligence that ingests emails, PDF attachments, images and handwritten sketches to extract order data and line items automatically.
- Automated real‑time stock checks and reservation logic before draft orders are created.
- Automated quotation generation and a review/approval workflow that leaves control with users rather than fully autonomous order posting.
- Integration points for MES, nesting/CAM systems, and workshop scheduling to close the loop from order intake to production.
This collection of capabilities is notable because it places AI at the critical upstream point of the manufacturing workflow: order acquisition. If executed well, it reduces manual touchpoints and brings order data into downstream systems faster and with fewer transcription errors.
How the Copilot‑driven order intake works (typical flow)
- A customer email, PDF order, scanned drawing or photo arrives in a monitored company mailbox.
- The Copilot/document intelligence stack extracts relevant fields: customer identity, item descriptions, dimensions, cut shapes, processing steps (e.g., temper, edgework, lamination), quantities and requested delivery dates.
- Extracted text is matched to internal items and routing logic (product codes, production operations).
- The ERP checks real‑time inventory and stock levels and flags shortages, suggested substitutions or lead times.
- A draft quotation or sales order is assembled and presented to a human operator for review, edit and approval.
- Once approved, the order is posted to Business Central and — via connectors — pushed to MES/nesting and scheduling systems.
The vendor message emphasises that users “stay fully in control” and can review, adjust and approve every step before confirmation. That design choice mitigates the risk of silent automation errors and supports regulatory or contractual requirements where explicit human sign‑off is required.
Technical underpinnings (what makes this plausible)
- Dynamics 365 Business Central provides the ERP backbone for inventory, finance and sales logic and already supports Copilot/agent features in recent Microsoft releases. Those built‑in agents (for example, the Sales Order Agent) are explicitly designed to work with shared mailboxes and email processing for order intake.
- Document intelligence functions (OCR, PDF parsing, handwriting recognition) are mature capabilities in the Microsoft cloud stack (Azure Cognitive Services, Form Recognizer, Computer Vision) and can be combined with Business Central agents to extract structured data from unstructured inputs.
- Industry verticalisation is achieved by mapping extraction outputs to the vendor’s product catalogues, form‑to‑internal code mappings, and rule engines that understand glass‑specific operations (e.g., edgework vs. tempering sequences).
- Standard integration options (REST APIs, OData, and Business Central extension points) make it possible to pass approved orders to MES, nesting software and downstream automation.
These building blocks are available today and make the overall solution technically credible. The true differentiator will be how well HEGLA‑HANIC has embedded glass domain knowledge into the Copilot and how robustly the connectors to shop‑floor systems perform in live production contexts.
Strengths and opportunities
- Domain specificity. A Copilot trained on glass‑industry terms, drawings and standard processing sequences reduces the amount of human correction required after automatic extraction. The more domain knowledge an agent has, the fewer mapping errors will occur when converting “laminated IGU” or “edge cut radius X” into ERP line items.
- Faster order‑to‑production lead times. By automating extraction and initial drafting, orders can be validated and routed to nesting systems much sooner — reducing idle time between sales acceptance and production.
- Tighter integration with Microsoft ecosystem. For shops already using Microsoft 365 and Business Central, the solution promises a more seamless identity, security and licensing model than piecing together disparate tools.
- Human‑in‑the‑loop design. The ability to review and approve automatically created drafts keeps responsibility explicit and provides a comfortable migration path for teams nervous about full automation.
- Scalability and updates. Building on Business Central and Azure offers scalability and continuous updates, easing the burden of patching and platform maintenance for smaller vendors.
- Potential for data quality improvement. Structured capture of orders and consistent product code mapping improves analytics, costing and re‑use of designs.
These advantages make glass365 attractive to mid‑sized fabricators who get many custom or semi‑standard orders and want to reduce the repetitive admin load on their sales clerks.
Risks, limitations and what to verify before you buy
- Accuracy of handwriting/sketch interpretation. Handwritten dimensions and free‑form sketches are notoriously inconsistent. While modern OCR and handwriting models can achieve impressive accuracy, results vary widely with image quality, handwriting style and the diversity of formats used by customers. Any vendor claim of handling “handwritten sketches” should be validated against your actual customer documents.
- Document variability. Customers send orders in dozens of formats. Success at scale depends on the system’s ability to map many different supplier/customer templates and phrasing into your ERP item master. Expect an initial period of supervised corrections and training.
- Data residency and model training. Where extracted documents and derived prompts are stored, and whether vendor telemetry or model training uses your data, must be contractually clarified. Organisations with strict data residency or GDPR constraints need explicit guarantees about where AI processing occurs and how long text/images are retained.
- Integration gaps with bespoke shop‑floor systems. Many fabricators use custom MES/PPS or dated CAM tooling. The smoothness of automation depends on the quality of connectors, middleware and event sequences between Business Central and those systems.
- Change management. Adopting an AI‑driven intake requires new acceptance workflows, staff retraining and explicit escalation procedures when the Copilot’s suggestion is incorrect.
- Vendor lock‑in and licensing complexity. Embedding Copilot in Business Central can create tighter ties to Microsoft licensing, and any bespoke add‑ons can increase the cost and complexity of future migrations.
- Unclear performance metrics. Vendors frequently tout “faster, more accurate” order entry without publishing baseline metrics (error rates, percentage of orders drafted correctly). Ask for pilot data or run your own proof‑of‑concept with your documents.
Flag any vendor claims that include precise accuracy percentages, time‑savings, or ROI without demonstrable pilot evidence. Those figures are useful but must be validated using representative company data.
Deployment checklist: how to pilot glass365 responsibly
- Select a narrow pilot scope: choose one customer or order channel (e.g., shared email from a top 5 customer) to limit document variability.
- Collect representative documents: PDFs, scanned sketches, Excel templates and photos. Provide these to the vendor for initial mapping and training.
- Define acceptance criteria: set measurable KPIs such as percentage of orders auto‑drafted accurately, average time saved per order, and human correction rate.
- Audit data flow and residency: confirm where documents are stored, whether any text is used for vendor model training, and retention windows.
- Test integration points: produce and verify the end‑to‑end sequence from email → draft order → user approval → posted order → MES/nesting ingestion.
- Monitor exception handling: log all Copilot rejections, corrections and manual edits to improve mappings iteratively.
- Establish governance: specify who can override Copilot, who reviews exceptions, and how audit trails are maintained.
- Define rollback criteria: if error rates or production disruptions exceed thresholds, have a plan to revert to manual intake for the pilot channel.
A pilot of 4–8 weeks with clear KPIs is the minimal practical approach to judge viability. Successful pilots should demonstrate both time savings and reduced transcription errors before expanding.
Security, compliance and operational governance
- Permissions and audit trails. Ensure the Copilot and any agents inherit Business Central’s role‑based access controls and produce auditable logs that show which user accepted or edited a draft order.
- Data movement and consent. Confirm whether Copilot features require cross‑region data movement or use Azure OpenAI endpoints. If so, document consent and the vendor’s commitments regarding telemetry and training data.
- Data retention policies. Demand explicit retention periods for uploaded attachments, parsed text and any intermediate artifacts.
- Model transparency. Ask whether the Copilot uses third‑party foundational models, vendor‑hosted fine‑tuning, or on‑tenant inference, and whether prompts or extracted data are stored to improve the service.
- Third‑party assessments. For high‑value or regulated customers, request a security assessment or SOC/ISO attestation from the vendor that covers the Copilot and integration components.
Attention to these controls is particularly important for firms with EU or sectoral privacy obligations: vague vendor statements about “data used to improve the product” are a red flag without contractual limits.
Integration with shop floor systems — practical realities
- MES and nesting connectivity. Successful automation requires deterministic, tested connectors between Business Central and downstream MES or nesting solutions. These connectors must preserve production parameters (cuts, kerfs, nesting orientation, edge processes) and maintain consistent part identifiers.
- Versioning and change control. Where the Copilot or mapping logic updates item codes or tolerances, a robust change‑control process must exist to avoid silently routing orders with incorrect processing steps.
- Fallback and manual overrides. Operators must be able to pause automation, correct draft orders and rerun mappings. The system should produce human‑readable discrepancy reports when mapping fails.
- Cutting list and nesting implications. Misinterpreted dimensions or part counts compound when passed into nesting — scrap rates can rise quickly if extraction errors aren’t caught early. Include nesting validation as part of the pilot.
Integration complexity is frequently the practical barrier to achieving “end‑to‑end” automation; plan time and budget for connector development and testing.
Competitive landscape and strategic positioning
Automated order intake is becoming mainstream in industry software. Several ERP and industry‑specific vendors are rolling out AI‑backed order entry assistants, each with different approaches to model hosting, domain tuning and integrations. The strategic tradeoffs are:
- Build on a major cloud ERP (Microsoft): faster integration with Microsoft 365, familiar identity and security model, but potential vendor lock‑in and tight coupling with Microsoft licensing.
- Use a specialist vertical provider with bespoke connectors: deeper out‑of‑the‑box domain logic for glass but potentially higher integration effort for non‑standard office tooling.
- Hybrid approach: combine a cloud ERP backbone with best‑of‑breed document intelligence and middleware to preserve flexibility.
For buyers, the practical question is which approach reduces total cost of ownership while preserving future agility. Vertical specialists like HEGLA‑HANIC offer product knowledge that generalist vendors cannot replicate; conversely, BigTech‑backed Copilot integrations ease the path for organisations already standardised on Microsoft.
Business case: realistic ROI expectations
AI‑assisted order entry can deliver substantial operational savings, but the magnitude depends on order mix and initial manual burdens. Conservative, actionable assumptions for ROI modelling:
- Time per manual order entry today: measure baseline (for example, 10–30 minutes depending on complexity).
- Expected reduction in manual time per order after pilot: conservative estimate 30–50% for standard orders that match trained templates; bespoke or highly free‑form orders may see less benefit initially.
- Error reduction: tracked as fewer production reworks, lower scrap, and fewer customer returns — quantify historic rework costs and apply conservative improvements (e.g., 10–30% reduction).
- Implementation cost: include Business Central licensing (if not already in place), HEGLA‑HANIC glass365 licensing/customisation, integration work for MES/nesting, and pilot professional services.
- Ongoing costs: monthly cloud services, support, and periodic retraining or mapping updates.
Do not accept generic vendor ROI claims without getting a tailored model built against your historic order volume and error rates. Use pilot results to revise assumptions and scale rollout decisions.
Questions to put to HEGLA‑HANIC (and any vendor) before purchase
- Can you run a pilot with our real customer documents and share anonymised accuracy metrics (extraction accuracy, order‑to‑order error rate) for those pilots?
- Where are documents and extracted data stored, and will any of it be used for third‑party model training?
- Can the Copilot run in‑tenant or within a private tenant, or does it require multitenant services?
- What audit trails and role‑based permissions are provided for automated agent actions?
- How do you handle mappings when a customer uses non‑standard terminology or creates ambiguous sketches?
- What connectors exist for our MES, nesting/CAM and scheduling systems, and can you supply sequence diagrams of the end‑to‑end flow?
- What SLAs and escalation paths do you offer if automated orders lead to production disruptions?
Insist on documented answers and include them in contractual terms where data residency, auditability and availability are important.
Practical recommendations for glass shops considering glass365
- Begin with a tightly scoped pilot that focuses on one or two repeat customers or a single channel (email attachments). This reduces variability while demonstrating the capability.
- Prepare a labelled dataset of past orders and sketches to accelerate mapping and fine‑tuning during the pilot.
- Include your MES/nesting team early; automation benefits are realised only when the downstream chain is robust.
- Keep human oversight in the loop and treat the Copilot as a productivity multiplier rather than a replacement for domain expertise.
- Negotiate data and model governance clauses, including explicit language on data retention, telemetry opt‑out and model training restrictions if you have strict privacy obligations.
- Measure and publish pilot KPIs internally (time saved, error reduction, order cycle time) to secure buy‑in for rollout.
Conclusion
glass365 is a credible and timely product for glass fabricators ready to digitise the first, often‑messy step of manufacturing: order intake. By combining
Dynamics 365 Business Central with a customised Copilot agent and document‑intelligence tooling, HEGLA‑HANIC is offering a tightly focused solution that addresses a long‑standing operational bottleneck in glass processing.
The potential upside is significant: faster order processing, fewer transcription errors, and a cleaner feed into nesting and MES systems. The pragmatic design choice to keep humans in control of final approvals is sensible, and Microsoft’s agent architecture provides a scalable, enterprise‑grade foundation.
However, buyers must be clear‑eyed about limitations. Handwritten sketches, extreme document variability and bespoke shop‑floor software create real implementation risk. The most important determinants of success are the quality of vendor mapping for your documents, the robustness of MES connectors, and contractual clarity about data use and model training.
For glass manufacturers, the smart path is a measured pilot with real documents, tight KPIs, and transparent governance. If the pilot demonstrates reliable extraction accuracy and smooth downstream integration, glass365 — and the wider class of ERP‑embedded Copilot agents — can materially reduce administrative friction and accelerate the path from inquiry to cut list. The future of glass order entry is clearly more automated; the task now is to ensure that automation is accurate, auditable and aligned to the realities of glass production.
Source: glassonweb.com
AI Meets Glass ERP: Smarter Order Entry with glass365