<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by AssetSolve on Medium]]></title>
        <description><![CDATA[Stories by AssetSolve on Medium]]></description>
        <link>https://medium.com/@contact_56156?source=rss-b4981e0ddcc6------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*dmbNkD5D-u45r44go_cf0g.png</url>
            <title>Stories by AssetSolve on Medium</title>
            <link>https://medium.com/@contact_56156?source=rss-b4981e0ddcc6------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 21:16:07 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@contact_56156/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[When does Oracle Fusion cash application need a custom agent?]]></title>
            <link>https://medium.com/@contact_56156/when-does-oracle-fusion-cash-application-need-a-custom-agent-54167c12b86b?source=rss-b4981e0ddcc6------2</link>
            <guid isPermaLink="false">https://medium.com/p/54167c12b86b</guid>
            <category><![CDATA[cash-application]]></category>
            <category><![CDATA[erp]]></category>
            <category><![CDATA[oracle-fusion]]></category>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[accounts-receivable]]></category>
            <dc:creator><![CDATA[AssetSolve]]></dc:creator>
            <pubDate>Tue, 17 Nov 2026 16:01:02 GMT</pubDate>
            <atom:updated>2026-11-17T16:01:02.115Z</atom:updated>
            <content:encoded><![CDATA[<p><em>AssetSolve editorial. Research and drafting were assisted by AI. This is an illustrative test design.</em></p><p>A customer sends a $10,000 payment. Two invoices are open for exactly that amount, and the bank memo contains a purchase-order number. The customer intended to settle one invoice and leave the other in dispute. An exact amount match is easy to demonstrate. A correct decision requires evidence that identifies what the customer meant to pay.</p><p>This is a useful case for deciding whether to configure embedded Oracle Fusion AI or build an extension. Give both approaches the same receipt, the same invoice history and the same approval rules. Agree in advance on when the right answer is to leave the payment unapplied and route it to an analyst. A fast but wrong match can make the collection queue look better while leaving the customer account wrong.</p><h4>Test the native workflow before pricing the extension</h4><p>Oracle’s ERP AI catalog lists a Cash Processing Agent for receipt processing and cash application. Check the version applicable to the target environment and what evidence the configured workflow can access. Use the normal receivables role, including its actual business-unit access. Record which references the agent can use, what an analyst can see before accepting the suggestion, and which system status confirms the allocation.</p><p>Only then write down what an extension would add. A remittance file from an external processor may identify the payer and invoices more reliably than the bank memo. If that is the only gap, an extension could fetch and validate the remittance, then hand it back to the existing receivables workflow. The custom scope should explain why configuration or a supported integration alone is insufficient. Oracle’s AI Agent Studio documentation covers external tools and human approval steps. Verify both the tool’s access and the person’s authority before using either in a transaction path.</p><h4>Make ambiguous receipts part of the test</h4><p>Use a set of labeled receipts covering an exact match, equal-amount invoices, a one-to-many payment, a partial payment, an unrelated payer, and a repeated bank transmission. Specify the expected result per case. For an approved allocation, read back the amount and invoice reference from the receivables record. For an ambiguous case, verify that the workflow leaves it unapplied with the reason and owner intact. A replay should cause zero additional allocations.</p><h4>Count the cost of a verified allocation</h4><p>For each approach, measure correct allocations among all eligible receipts, manual minutes including rework, and the number and age of cases left unapplied. Compare false matches separately. Include the cost of supporting external data and repeating tests after releases. The extension earns its place if the added evidence measurably improves accepted results for the target case mix at a defensible operating cost. If the native workflow meets that bar, keep the scope small.</p><p><a href="https://docs.oracle.com/en/cloud/saas/fusion-ai/aiafl/ai-erp.html">Oracle Fusion ERP features with AI</a>, <a href="https://docs.oracle.com/en/cloud/saas/fusion-ai/26d/aiaas/toc.htm">Oracle AI Agent Studio documentation</a>, <a href="https://www.assetsolve.com/products/working-capital-finance">AssetSolve working capital finance workflows</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=54167c12b86b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The first AI workflow in a greenfield ERP program should be an exception]]></title>
            <link>https://medium.com/@contact_56156/the-first-ai-workflow-in-a-greenfield-erp-program-should-be-an-exception-1b974370211d?source=rss-b4981e0ddcc6------2</link>
            <guid isPermaLink="false">https://medium.com/p/1b974370211d</guid>
            <category><![CDATA[sap]]></category>
            <category><![CDATA[erp]]></category>
            <category><![CDATA[order-to-cash]]></category>
            <category><![CDATA[oracle-fusion]]></category>
            <category><![CDATA[ai-agent]]></category>
            <dc:creator><![CDATA[AssetSolve]]></dc:creator>
            <pubDate>Tue, 03 Nov 2026 21:58:26 GMT</pubDate>
            <atom:updated>2026-11-03T21:58:26.167Z</atom:updated>
            <content:encoded><![CDATA[<p><em>AssetSolve editorial. Research and drafting were assisted by AI. The example below is illustrative.</em></p><p>An operations manager sees that an order shipped on Monday. The billing analyst sees an invoice number on Wednesday. On Friday, receivables still has nothing to collect. A conversation with an AI assistant might save a few minutes of searching. The harder question is whether anyone can establish which business event failed and who has authority to put it right.</p><p>That is the kind of case I would take into a greenfield design workshop. Start with one order, one delivery and one account. Walk the records in sequence. Agree which document establishes physical delivery, which status establishes billing readiness, and which posting establishes an open receivable. Write down the three most plausible reasons for disagreement. An unposted invoice, an intentional hold and a late interface should lead to different actions.</p><h4>Try the standard first</h4><p>An apparent gap sometimes disappears after the team fixes master data or configuration. An account determination error belongs with the people who own the accounting setup. A carrier proof-of-delivery dispute may still need work across systems. The gap register should show which of those situations actually exists before a team proposes an agent.</p><p>On SAP, evaluate the applicable Joule capability in the target environment and confirm that it works for the intended user. On Oracle, check the documented Fusion release and AI feature type, then demonstrate the same case with the right business-unit access. Both tests should use the buyer’s own process and sample records under appropriate controls. A vendor demo answers a different question.</p><h4>Find out what happens after a timeout</h4><p>Suppose the analyst authorizes a correction. The tool sends the request to ERP, the ERP commits, and the response never reaches the assistant. Automatic retry might be safe for that interface. It might also create a duplicate business effect. The team needs to know which before go-live.</p><p>My preferred test gives the request a stable reference. The workflow looks up the outcome using that reference before retrying. If it cannot establish whether ERP acted, it leaves an open exception for someone who can investigate. Prove the test by reading back the resulting document and reconciling the accounting impact. Repeat the exercise with a stale approval and a user who lacks access to the company.</p><h4>Measure the handoff</h4><p>Compare the same class of exceptions under the standard workflow and the assisted workflow. Count time to a verified result, human review effort, reopened cases and control failures. Keep the difficult cases in view. An assistant can make the average look better by routing away anything ambiguous, leaving the analyst with a smaller but harder queue.</p><p>The lesson is useful even if the pilot stops at read-only assistance. The process team leaves with clearer ownership, a better test set and a baseline against which a future extension can be judged.</p><p><a href="https://help.sap.com/docs/joule/capabilities-guide/about-this-document">SAP Joule capabilities guide</a>, <a href="https://docs.oracle.com/en/cloud/saas/fusion-ai/aiafl/ai-erp.html">Oracle Fusion ERP features with AI</a>, <a href="https://www.assetsolve.com/capabilities/autonomous-erp">AssetSolve Autonomous ERP approach</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1b974370211d" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>