Talk:ADempiere Integration With Alfresco
I've done some work to extend this work for the zkwebui - I'll publish it on google code
The use case I'm interested in mostly is the bulk upload of documents and then indexing/scanning and attachment and/or creation (hopefully automatically) to documents. To elaborate around the PO'ed Invoice use case:
1. We want a highest level of expenditure to be purchase ordered because this provides the opportunity to both authorise the expenditure before legally contracting with another business and control the expenditure (also against budget) when the supplier invoices us.
2. The use case would be to issue the PO as normal with the appropriate authorisation rules and then when we receive invoices we scan (and index) these. (Note: Possible integration with Ephesoft or components - main issue to date is finding the right config options on the tesseract OCR engine v recostar windows only) So let's say a batch of images are scanned (for one client) and we ocr the fields invoice number, supplier name/code, invoice date, po number. If the PO number field is missing this is a non-PO invoice and must be authorised via a different route/process.
We would want to load the scanned documents to a folder in alfresco (or another cmis compliant repository) in to a document specific directory under a "site" - and name these documents probably by invoice number and supplier code. Either way, I envisage extending the (afresco) document model to hold ADempiere fields such as AD_CLIENT_ID, AD_TABLE_ID, RECORD_ID. I think this is effectively the reverse of the use case modelled in the POC. We would want to sync the documents back the the AD_ATTACHMENT from Alfresco and match the record in the invoice table based on the meta data scanned (or input) in to Alfresco.
3. Where an invoice does not already exist (should be the case as we are scanning new invoices), this should be created. Duplicate invoices should be detected based on supplier and invoice number and some action performed. (Note: some suppliers do this and we may need to accommodate this bad practise. One idea would be to flag the document in some invalid/duplicate state which could be overridden in some manual workflow). To achieve this import to ADempiere, we could scan for unmatched invoices (status when first imported) and then try to match these.
4. An unmatched report would be helpful as well as tracking of the matching process (batch, date or process ID?)
I think the biggest use case here is invoice (AP) processing. Benchmarks in the UK indicate that manual processing of invoices is around £10 per invoice. Electronic gateways (like OB10) reduce this cost to around £4 per invoice but create some noise around invalid data which can be accommodated (and adds cost). Manual indexing offshore is around the same price (e.g. Accenture Bangalore). Automated indexing (e.g. Kofax) brings the cost down to around £1.30-£1.40 per invoice (multi page) for low volumes (30k invoices p.a.) and to around £0.70 per invoice for volumes greater than 200k. Note: Actual experience in a business
The business case could be made between the different options and with benchmark annual invoices to calculate the return.
--Juddm 08:22, 10 July 2013 (UTC)
