Assessing the Applicability of Fine-Tuning LargeLanguage Models for Designing and Deploying 24/7 Context-Aware Multichannel CRM

Assessing the Applicability of Fine-Tuning LargeLanguage Models for Designing and Deploying 24/7 Context-Aware Multichannel CRM | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Assessing the Applicability of Fine-Tuning LargeLanguage Models for Designing and Deploying 24/7 Context-Aware Multichannel CRM Naoudouwel Fulbert, Maria Vinitha, Kanagasabai Thiruthanigesan This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7693997/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Modern Customer Relationship Management (CRM) systems must provide continuous, personalized support across diverse communication channels, including voice, text, and audio platforms. Traditional CRM solutions suffer from fragmented customer experiences due to their optimization for single-channel interactions, leading to operational inefficiencies and customer dissatisfaction. This research addresses the challenge of achieving consistent and efficient multichannel interactions using fine-tuned Large Language Models (LLMs), specifically the LLaMA 3.1 model. The model was fine-tuned using Low-Rank Adaptation(LoRA) techniques on proprietary datasets sourced from Mic & Mac Solutions. The dataset underwent comprehensive collection, cleaning, and structuring processes to ensure suitability for training. Our approach integrates Speech-to-Text(STT) and Text-to-Speech (TTS) pipelines alongside API-driven communication to ensure seamless transitions between channels. Initial training with basic hyperparameters (single epoch, learning rate of 1e-3) achieved moderate performance with a final training loss of 0.7-0.8, but exhibited optimization instabilities evidenced by gradient norm spikes exceeding 4.0. Following systematic hyper-parameter optimization—including extended training to 10 epochs, reduced learning rate (1e-4), and cosine annealing scheduling—the system demonstrated significantly improved stability with gradient norms maintained below 1.5 and reduced final training loss to 0.3-0.4. Experimental evaluation indicates that the optimized fine-tuned model delivers context-aware and coherent responses across all communication modalities while maintaining 24/7 availability. The findings highlight the transformative potential of fine-tuned LLMs in unifying multi-channel CRM environments, demonstrating substantial improvements in training stability and performance metrics, thereby setting a new standard for continuous, consistent, and efficient customer engagement in contemporary business operations. Large Language Models (LLMs) Customer Relationship Management (CRM) LLaMA 3.1 Multichannel Communication AI-driven CRM Voice and Text Channels Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7693997","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":521386669,"identity":"7a1ed67f-b505-4a75-9195-2aed4fb8ab22","order_by":0,"name":"Naoudouwel Fulbert","email":"","orcid":"","institution":"Loyola College","correspondingAuthor":false,"prefix":"","firstName":"Naoudouwel","middleName":"","lastName":"Fulbert","suffix":""},{"id":521386670,"identity":"c6f67a04-0d5e-45d5-90bb-544c7329e9fe","order_by":1,"name":"Maria Vinitha","email":"","orcid":"","institution":"Loyola College","correspondingAuthor":false,"prefix":"","firstName":"Maria","middleName":"","lastName":"Vinitha","suffix":""},{"id":521386671,"identity":"672e673e-ccbe-457e-a8cb-7ad6e4c10621","order_by":2,"name":"Kanagasabai Thiruthanigesan","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABD0lEQVRIiWNgGAWjYDACCcYGIHmAgYEZJsDewHggwYCBgY8Huw4esJYEZC08BxiAWgwY2HBqAZEgLXBrJRJAluLWYi/d3PaZ98cdeXN29osPf9TcYZCc+fzBgQcFf4BaDmDVwiNzsHk2T8Izw53NPMXGPMeeMUhL5xhAHMbbgMNhic3MPAmHGTcc5kmTZmA7zCAnnQP1Cz8uv0C02AO1pP/88Q+oRfL4A6K0JG44zH6MgbftMIO0BAMBh91IbGack3Y4GWgLszRv32EeyR6wX4x5cHmffUb6Y4Y3NodtN5w//vDjj2+H5SSOH3/48McfOTl+ngTsLkOy0ABMwrmE1IMsfECEolEwCkbBKBiJAAANm1vN184FjwAAAABJRU5ErkJggg==","orcid":"","institution":"Sri Lanka Institute of Information Technology","correspondingAuthor":true,"prefix":"","firstName":"Kanagasabai","middleName":"","lastName":"Thiruthanigesan","suffix":""}],"badges":[],"createdAt":"2025-09-23 11:54:03","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7693997/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7693997/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":92456763,"identity":"ab830e6a-c7df-4c70-8726-8d9f67b7b508","added_by":"auto","created_at":"2025-09-30 02:28:01","extension":"json","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5327,"visible":true,"origin":"","legend":"","description":"","filename":"4003c2a118ea4741ae6dbbac740a73e6.json","url":"https://assets-eu.researchsquare.com/files/rs-7693997/v1/8487edea4696e2bdafdf3035.json"},{"id":92700249,"identity":"c0632867-4d8d-4bcd-91d0-e88ad67c5b64","added_by":"auto","created_at":"2025-10-03 08:09:00","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1350339,"visible":true,"origin":"","legend":"","description":"","filename":"LLMforCRMFinalWithoutAuthorList.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7693997/v1_covered_56b905f2-6851-4e9a-8a4a-d27c3b6f00b1.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"\u003cp\u003eAssessing the Applicability of Fine-Tuning LargeLanguage Models for Designing and Deploying 24/7 Context-Aware Multichannel CRM\u003c/p\u003e","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Large Language Models (LLMs), Customer Relationship Management (CRM), LLaMA 3.1, Multichannel Communication, AI-driven CRM, Voice and Text Channels","lastPublishedDoi":"10.21203/rs.3.rs-7693997/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7693997/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Modern Customer Relationship Management (CRM) systems must provide continuous, personalized support across diverse communication channels, including voice, text, and audio platforms. Traditional CRM solutions suffer from fragmented customer experiences due to their optimization for single-channel interactions, leading to operational inefficiencies and customer dissatisfaction. This research addresses the challenge of achieving consistent and efficient multichannel interactions using fine-tuned Large Language Models (LLMs), specifically the LLaMA 3.1 model. The model was fine-tuned using Low-Rank Adaptation(LoRA) techniques on proprietary datasets sourced from Mic \u0026 Mac Solutions. The dataset underwent comprehensive collection, cleaning, and structuring processes to ensure suitability for training. Our approach integrates Speech-to-Text(STT) and Text-to-Speech (TTS) pipelines alongside API-driven communication to ensure seamless transitions between channels. Initial training with basic hyperparameters (single epoch, learning rate of 1e-3) achieved moderate performance with a final training loss of 0.7-0.8, but exhibited optimization instabilities evidenced by gradient norm spikes exceeding 4.0. Following systematic hyper-parameter optimization—including extended training to 10 epochs, reduced learning rate (1e-4), and cosine annealing scheduling—the system demonstrated significantly improved stability with gradient norms maintained below 1.5 and reduced final training loss to 0.3-0.4. Experimental evaluation indicates that the optimized fine-tuned model delivers context-aware and coherent responses across all communication modalities while maintaining 24/7 availability. The findings highlight the transformative potential of fine-tuned LLMs in unifying multi-channel CRM environments, demonstrating substantial improvements in training stability and performance metrics, thereby setting a new standard for continuous, consistent, and efficient customer engagement in contemporary business operations.","manuscriptTitle":"Assessing the Applicability of Fine-Tuning LargeLanguage Models for Designing and Deploying 24/7 Context-Aware Multichannel CRM","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-30 02:27:57","doi":"10.21203/rs.3.rs-7693997/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"03ebb112-c9fd-4821-905e-f4183a440d2d","owner":[],"postedDate":"September 30th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-10-03T08:08:38+00:00","versionOfRecord":[],"versionCreatedAt":"2025-09-30 02:27:57","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7693997","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7693997","identity":"rs-7693997","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}