Telecom operators face intense competition and pressure to transform. Many seek emerging technologies like generative AI (GenAI) to increase productivity, boost revenue growth, and enhance customer experience. However, realizing the full potential of GenAI in telecom requires capabilities that address challenges unique to our industry. We see GenAI models enriched with telco BSS/OSS data using platforms like Netcracker GenAI Telco Solution as a sensible, effective, and safer way to adopt cutting-edge technology that provides fast time-to-value.
The Potential Benefits of GenAI for Telecoms
GenAI models, including the large language model (LLM), productized as ChatGPT, demonstrate extraordinary natural language skills. Other machine learning/deep learning models for images, audio, and video generate output nearly indistinguishable from those of human creators.
GenAI brings the promise of innovation and transformation to telecom operators. Applied appropriately, GenAI can transform multiple areas of a telecom operator’s business:
- Customer Service: Digital assistants powered by GenAI can resolve common customer queries faster without involving a human agent. Such assistants can reduce calls to the contact center. Even when GenAI is used in an assistive fashion, it can help human agents resolve issues faster by surfacing the most likely solutions to customer issues.
- Sales: Personalized recommendations and messaging driven by GenAI can enable more effective upsell and cross-sell, increasing sales productivity.
- Marketing: GenAI can help create marketing collateral, messaging, and text/image/video content for online and personalized marketing. The capabilities here are less telco industry-specific, but adding knowledge of telco-specific offerings for B2B and B2C allows specialized messaging.
- Legal and Finance: As in other industries, GenAI can be used for telco contract analysis and summarization to accelerate contract editing and review and to speed up financial data analytics (direct or code-generation supported like ChatGPT’s Advance Data Analysis).
- Procurement and Purchasing: GenAI has been used in other industries to generate and process requests for proposals/information (RFPs/RFIs) and purchase contracts. If adapted with telco-specific information (through retrieval augmentation, fine-tuning, or other non-GenAI ML capabilities), GenAI can accelerate RFP evaluation and approval processes.
- Network Operations: Automating monitoring, predictive incident response, and network planning with GenAI can cut critical network incidents and reduce the time to design and provision new network services.
- Business Operations: GenAI can help automate or accelerate processes like service configuration, content creation, and data analysis, boosting skilled employee productivity.
- Software Development: GenAI can dramatically increase coders’ productivity (as has been shown by Microsoft Copilot, Amazon’s Codewhisperer, and other assistive code-generation technologies). Likewise, telcos can use GenAI and related ML technologies (generative adversarial networks) to generate test cases and test data during unit testing and integration testing to ensure more robust systems.
The collective impact across these critical functions could be enormous. Early studies peg such improvements at between 14-67% across numerous tasks (e.g., 14% for customer service, between 55% to 67% for software development). In aggregate, the consultancy McKinsey expects 30-45% gains in productivity for workers. But telecoms can only realize these benefits with GenAI models tailored to their needs.
The Need for a Custom Telco GenAI
Many of the earliest studies in productivity gain stem from using publicly available GenAI services like OpenAI ChatGPT, Google Bard, or Anthropic Claude 2. However, these services pose several challenges for usage in telecom environments:
- Even the most advanced public models (commercial, free, and open-source offerings) lack the specialized telecom domain knowledge around billing, services, network technologies, and promotions. They cannot drive practical telco-specific applications without this expertise. While ChatGPT can be used for specific marketing and sales functions — writing blog posts, summarizing information, generating marketing copy, and crafting social media posts — more substantive use cases require detailed telco domain information.
- Public models without telecom business knowledge can’t be trained or augmented with sensitive operator data. Strict data privacy laws prohibit exposing customers’ confidential account information or sensitive network data to external models for prompt context, few-shot learning, or retrieval augmentation. This can restrict the ability to use LLM-as-a-service offerings, mainly if the LLM is not hosted in the region or the country. Even with private LLMs, the hosting may have to be in a region or a country due to sovereignty or compliance requirements.
- Building sufficiently capable private models requires massive computing resources for training. Some of the largest LLMs were trained at price tags exceeding $100 million. This puts private models out of reach for many operators. Besides, it’s unclear if that premise even makes sense — remember that LLMs only showed value after training with large corpora of text.
- Pulling together the relevant telco data into repositories for fine-tuning or building retrieval augmentation systems can take significant effort and know-how, and pulling real-time BSS/OSS context for queries can be challenging.
- Even fine-tuning or augmenting a model with telecom data for specific use cases requires deep telco expertise, ML expertise, and private computing resources to ensure no leakage of sensitive customer and network data.
- Building the proper telco workflow requires understanding the use cases, the appropriate models and pipeline to use, prompt management, and output validation and sanitization.
- GenAI models will need tight real-time integration into OSS and BSS systems to unlock the most value. 90% of relevant telco use cases rely on data from these backend operator systems. Public models have no access to this.
A viable path forward for GenAI in telecom requires addressing these data privacy, security, cost, and integration challenges with a tailored approach.
Netcracker’s “GenAI Inside” Go-to-Market Model
The BSS/OSS are core to how telcos run their businesses. By integrating GenAI with these systems, telcos can gain the benefit of GenAI within operational business systems. This is similar to Microsoft building Copilot into their Office 365 suite or Google offering Duet AI as part of Google Workspace and Google Cloud, enabling helpful GenAI assistants to render support for application or use-case-specific tasks. Other popular enterprise SaaS applications like Salesforce, Workday, and Zendesk do the same by building GenAI capabilities into their workflows.
Image source: Netcracker
Netcracker’s GenAI offering takes this integrative approach. The Netcracker GenAI solution provides:
- Comprehensive scenarios based on Netcracker’s extensive telecom Knowledge Bases and expertise which provides the data, context and instructions to enrich GenAI models with precise, personalized prompts for better results.
- Sophisticated techniques, including data obfuscation and ML-based confidential data detection using context to ensure no sensitive customer or network data is exposed to public models, overcoming security limitations.
- Flexibility to use different private and public GenAI models across multiple cloud providers (AWS, Microsoft Azure, Google Cloud), from different leading vendors (Anthropic, Cohere, OpenAI, and others), and free/open-source (Hugging Face, Meta), with Netcracker’s guidance on the combinations that best fit the task.
- Tight integration with Netcracker’s BSS/OSS portfolio and multivendor systems enables GenAI models to access and transform private telco data in real-time.
- A pre-built collection of over 40 sophisticated scenarios, including digital care assistant, B2B sales adviser and digital operations technician to help unlock GenAI’s potential for complex telco use cases.
- The Netcracker Trust Gateway helps ensure security, relevance, and accuracy when applying GenAI models. The Trust Gateway’s role is to ensure that sensitive data stays protected, responses match user intent, and content aligns with operator preferences.
Specific use case examples quoted by Netcracker as part of their GenAI launch include:
- Digital assistants explaining billing details, including usage history, subscription plans, and roaming charges.
- Sales advisers guiding B2B customers through solutions for office moves or upgrades with personalized recommendations.
- Digital operations technicians assisting with fixing network issues and making new installations..
- Automated redesign of underperforming network services by analyzing health metrics and past issues.
By removing the barriers around security, data integration, and domain knowledge, Netcracker aims to help telecom operators realize GenAI’s immense benefits.
AvidThink’s Take on Netcracker’s GenAI Solution
We’ve been engaged with telcos, hyperscalers, data center operators, and ISVs on GenAI for months now (since ChatGPT made its debut). While we had previously consulted on strategies around AI/ML and analytics for telecom operators in the past few years, the impact of GenAI was dramatic. Telco boards and executives wanted to know how to use GenAI, even if they didn’t understand it yet.
Further, we’ve suggested that any public (or even internal) use of GenAI needs to have strong guardrails in place to protect against IP and copyright infringement, data leakage, and toxic and biased content. These protections are in addition to data privacy and sovereignty compliance. Likewise, many telcos with sustainability efforts must acknowledge that increased use of GenAI could impact their carbon footprint.
We’ve advised all our clients, telco and otherwise, that the fastest way to achieve value with GenAI is to integrate it into existing software applications. These software ISVs are trusted vendors with pre-vetted terms of service, confidentiality, data governance, and privacy policies. Many of these vendors (like Microsoft) are extending copyright liability protection to generated content.
Netcracker’s approach falls in line with those recommendations. Further, Netcracker’s strategy in providing a Trust Gateway as guardrails is laudable, as is creating a valuable catalog of ready-to-go scenarios, high quality prompt enrichment that works with any GenAI model, and integration with crucial telco BSS and OSS data.
Image source: Netcracker
While telcos should continue to pursue parallel GenAI initiatives in network and infrastructure automation and continue evaluating non-GenAI ML techniques on RAN and other systems, it doesn’t take a super intelligent AI to know that leveraging GenAI integrated with OSS/BSS in a safe and controlled way, as Netcracker has done, is a smart choice.
Note: Sponsors do not have any editorial control over article content, and the views represented herein are solely those of AvidThink LLC.