This submit is cowritten by Nikhil Mathugar, Marc Breslow and Sudhanshu Sinha from Totogi.
This weblog submit describes how Totogi automates change request processing. Totogi is an AI firm centered on serving to serving to telecom (telco) firms innovate, speed up progress and undertake AI at scale. BSS Magic, Totogi’s flagship product, connects and fashions telco enterprise operations, overlaying legacy methods with an AI layer. With BSS Magic, telcos can prolong, customise, and modernize their methods with out vendor dependencies or prolonged implementations. By partnering with the AWS Generative AI Innovation Middle and utilizing the fast innovation capabilities of Amazon Bedrock, we accelerated the event of BSS Magic, serving to Totogi’s clients innovate quicker and acquire extra management over their tech stack.
On this submit, we discover the challenges related to the normal enterprise help system (BSS), and the progressive options offered by Totogi BSS Magic. We introduce intricacies of telco ontologies and the multi-agent framework that powers automated change request processing. Moreover, the submit will define the orchestration of AI brokers and the advantages of this strategy for telecom operators and past.
Challenges with BSS
BSS are notoriously tough to handle. A typical BSS stack consists of a whole lot of various purposes from varied distributors. However these BSS purposes are tough to combine, both proscribing telcos to the seller’s ecosystem or requiring them to spend money on expensive customizations. Such customizations are gradual and resource-intensive due to their reliance on specialised engineering expertise.
Every change request necessitates an intensive evaluation of potential impacts throughout interconnected modules, consuming vital effort and time. Even small updates can contain a number of rounds of coding, testing, and reconfiguration to realize stability. For telecom operators, the place system reliability is crucial, these safeguards are non-negotiable, however they arrive at a steep value. This course of is additional sophisticated by the shortage of engineers with the mandatory experience, driving up prices and elongating timelines. Because of this, growth cycles for brand new options or providers usually take months to finish, leaving operators struggling to satisfy the calls for of a fast-moving market.
Initiatives like TM Discussion board’s Open Digital Structure (ODA) purpose to resolve this, but most distributors are gradual to undertake such open requirements. This dynamic amplifies technical debt and inflates operational bills.
BSS Magic answer overview
Totogi BSS Magic reduces the complexity utilizing AI-generated interoperability, which helps simplify integrations, customizations, and utility growth. BSS Magic has two key facets:
- A telco ontology that understands the semantic meanings of information buildings and the relationships between them, linking disparate knowledge right into a coherent community of information.
- Multi-agent framework for absolutely automated change requests (CR), which reduces CR processing time from 7 days to some hours.
Telco ontology: The important thing to interoperability
Ontologies function semantic blueprints that element ideas, relationships, and area information. In telecom, this implies translating the BSS panorama into a transparent, reusable, and interoperable ecosystem. Totogi’s telco ontology facilitates a deep understanding of information interplay and seamless integration throughout any vendor or system.
By adopting FAIR ideas (Findability, Accessibility, Interoperability, and Reusability), the ontology-driven structure turns static, siloed knowledge into dynamic, interconnected information property—unlocking trapped knowledge and accelerating innovation. An outline diagram of the ontology is offered within the following determine.
Multi-agent framework for automated change request processing
AI brokers are superior software program purposes skilled to carry out particular duties autonomously. Totogi’s BSS Magic AI brokers have in depth area information and use this understanding to handle advanced knowledge interactions throughout a number of vendor methods. These brokers robotically generate and take a look at telco-grade code, changing conventional integrations and customizations with clever, AI generated purposes. At its core, BSS Magic makes use of a multi-agent AI strategy with suggestions loops to automate all the software program growth pipeline. Every agent is designed to satisfy a selected position within the growth pipeline:
- Enterprise evaluation agent interprets unstructured necessities into formal enterprise specs.
- Technical architect agent takes these enterprise specs and defines technical architectures, APIs, and dependencies.
- Developer agent generates high-quality, deployable code, full with modular designs and optimizations.
- QA agent validates the code for adherence to greatest practices, bettering high quality and safety. It gives suggestions which is utilized by the developer agent to replace the code.
- Tester agent generates sturdy unit take a look at circumstances, streamlining validation and deployment. The results of the take a look at circumstances is utilized by the developer agent to enhance the code.
An outline of the system is offered within the following determine.
This built-in pipeline reduces the time to finish a change request from 7 days to some hours, with minimal human intervention. The stipulations for implementing the system embrace an AWS account with entry to Amazon Bedrock, AWS Step Features, AWS Lambda, and configured Amazon credentials. The AI brokers are carried out utilizing Anthropic Claude massive language fashions (LLMs) via Amazon Bedrock. State administration and workflow coordination are dealt with by Step Features for dependable development via every stage. The AWS infrastructure gives the enterprise-grade reliability, safety, and scalability important for telco-grade options.
To construct the framework, Totogi collaborated with the AWS Generative AI Innovation Middle (GenAIIC). GenAIIC supplied entry to AI experience, industry-leading expertise, and a rigorous iterative course of to optimize the AI brokers and code-generation workflows. It additionally offered steering on immediate engineering, Retrieval Augmented Era (RAG), mannequin choice, automated code assessment, suggestions loops, sturdy efficiency metrics for evaluating AI-generated outputs, and so forth. The collaboration helped set up strategies for sustaining reliability whereas scaling automation throughout the platform. The answer orchestrates a number of specialised AI brokers to deal with the whole software program growth lifecycle, from necessities evaluation to check execution. The small print of the AI brokers are given within the following sections.
Multi-agent orchestration layer
The orchestration layer coordinates specialised AI brokers via a mixture of Step Features and Lambda capabilities. Every agent maintains context via RAG and few-shot prompting strategies to generate correct domain-specific outputs. The system manages agent communication and state transitions whereas sustaining a complete audit path of choices and actions.
Enterprise evaluation technology
The Enterprise Analyst agent makes use of Claude’s pure language understanding capabilities to course of assertion of labor (SOW) paperwork and acceptance standards. It extracts key necessities utilizing customized immediate templates optimized for telecom BSS area information. The agent generates structured specs for downstream processing whereas sustaining traceability between enterprise necessities and technical implementations.
Technical structure technology
The Technical Architect agent transforms enterprise necessities into concrete AWS service configurations and architectural patterns. It generates complete API specs and knowledge fashions and incorporates AWS Properly-Architected ideas. The agent validates architectural choices in opposition to established patterns and greatest practices, producing infrastructure-as-code templates for automated deployment.
Code technology pipeline
The Developer agent converts technical specs into implementation code utilizing Claude’s superior code technology capabilities. It produces sturdy, production-ready code that features correct error dealing with and logging mechanisms. The pipeline incorporates suggestions from validation steps to iteratively enhance code high quality and preserve consistency with AWS greatest practices.
Automated high quality assurance
The QA agent is constructed utilizing Claude to carry out complete code evaluation and validation. It evaluates code high quality and identifies potential efficiency points. The system maintains steady suggestions loops with the event stage, facilitating fast iteration and enchancment of generated code primarily based on high quality metrics and greatest practices adherence. The QA course of consists of fastidiously crafted prompts.
QA code evaluation immediate:
This immediate helps the QA agent carry out thorough code evaluation, consider high quality metrics, and preserve steady suggestions loops with growth levels.
Take a look at automation framework
The Tester agent creates complete take a look at suites that confirm each purposeful and non-functional necessities. It makes use of Claude to know take a look at contexts and generate acceptable take a look at situations. The framework manages take a look at refinement via analysis cycles, reaching full protection of enterprise necessities whereas sustaining take a look at code high quality and reliability. The testing framework makes use of a multi-stage immediate strategy.
Preliminary take a look at construction immediate:
Take a look at implementation immediate:
Take a look at outcomes evaluation immediate:
This structured strategy results in complete take a look at protection whereas sustaining top quality requirements. The framework at the moment achieves 76% code protection and efficiently validates each purposeful and non-functional necessities.
The Tester agent gives a suggestions loop to the Growth agent to enhance the code.
Conclusion
The combination of Totogi BSS Magic with Amazon Bedrock presents a complete answer for contemporary telecom operators. Some takeaways so that you can contemplate:
- Finish-to-end automation: BSS Magic automates all the growth lifecycle—from concept to deployment. AI brokers deal with every little thing from necessities, structure, and code technology to testing and validation.
- Outcomes: The agentic framework considerably boosted effectivity, lowering change request processing from seven days to some hours. The automated testing framework achieved 76% code protection, persistently delivering high-quality telecom-grade code.
- Distinctive worth for telecom operators: By utilizing Totogi BSS Magic, telecom operators can speed up time-to-market and cut back operational prices. BSS Magic makes use of autonomous AI, independently managing advanced duties so telecom operators can think about strategic innovation. The answer is supported by Amazon Bedrock, which affords scalable AI fashions and infrastructure, high-level safety and reliability crucial for telecom.
- Impression to different industries: Whereas BSS Magic is geared in direction of the telecom {industry}, the multi-agent framework will be repurposed for common software program growth throughout different industries.
- Future work: Future enhancements will give attention to increasing the mannequin’s area information in telecom and different domains. One other attainable extension is to combine an AI mannequin to foretell potential points in change requests primarily based on historic knowledge, thereby preemptively addressing widespread pitfalls.
Any suggestions and questions are welcome within the feedback beneath. Contact us to interact AWS Generative AI Innovation Middle or to be taught extra.
Concerning the authors
Nikhil Mathugar is a Presales Full Stack Engineer at Totogi, the place he designs and implements scalable AWS-based proofs-of-concept throughout Python and trendy JavaScript frameworks. He has over a decade of expertise in architecting and sustaining large-scale methods—together with internet purposes, multi-region streaming infrastructures and high-throughput automation pipelines. Constructing on that basis, he’s deeply invested in AI—specializing in generative AI, agentic workflows and integrating large-language fashions to evolve Totogi’s BSS Magic platform.
Marc Breslow is Subject CTO of Totogi, the place he’s using AI to revolutionize the telecommunications {industry}. A veteran of Accenture, Lehman Brothers, and Citibank, Marc has a confirmed observe report of constructing scalable, high-performance methods. At Totogi, he leads the event of AI-powered options that drive tangible outcomes for telcos: lowering churn, rising Common Income Per person (ARPU), and streamlining enterprise processes. Marc is accountable for buyer proof factors demonstrating these capabilities. When not participating with clients, Marc leads groups constructing Totogi’s BSS Magic expertise, producing purposes and bettering effectivity utilizing AI brokers and workflows.
Sudhanshu Sinha is Chief Know-how Officer and a founding staff member at Totogi, the place he works alongside Performing CEO Danielle Rios to drive the telecom {industry}’s shift to AI-native software program. As the important thing strategist behind BSS Magic, he formed its structure, go-to-market, and early adoption—translating AI-native ideas into measurable worth for operators. He additionally helped outline Totogi’s Telco Ontology, enabling interoperability and automation throughout advanced BSS landscapes. With over twenty years in telecommunications, Sudhanshu blends deep technical perception with industrial acumen to make AI-driven transformation sensible and worthwhile for telcos worldwide.
Parth Patwa is a Knowledge Scientist on the AWS Generative AI Innovation Middle, the place he works on buyer initiatives utilizing Generative AI and LLMs. He has an MS from College of California Los Angeles. He has printed papers in top-tier ML and NLP venues, and has over 1000 citations.
Mofijul Islam is an Utilized Scientist II and Tech Lead on the AWS Generative AI Innovation Middle, the place he helps clients deal with customer-centric analysis and enterprise challenges utilizing generative AI, massive language fashions (LLM), multi-agent studying, code technology, and multimodal studying. He holds a PhD in machine studying from the College of Virginia, the place his work centered on multimodal machine studying, multilingual NLP, and multitask studying. His analysis has been printed in top-tier conferences like NeurIPS, ICLR, AISTATS, and AAAI, in addition to IEEE and ACM Transactions.
Andrew Ang is a Senior ML Engineer with the AWS Generative AI Innovation Middle, the place he helps clients ideate and implement generative AI proof of idea initiatives. Exterior of labor, he enjoys taking part in squash and watching aggressive cooking reveals.
Shinan Zhang is an Utilized Science Supervisor on the AWS Generative AI Innovation Middle. With over a decade of expertise in ML and NLP, he has labored with massive organizations from numerous industries to resolve enterprise issues with progressive AI options, and bridge the hole between analysis and {industry} purposes.
