Tencent Marvis transforms into a team of AI butlers for managing PCs, files, browsers, and software

Automation has become a standard approach in the field of Customer Experience (CX). According to the CallMiner CX Landscape Report for 2026, 99% of organizations involved in CX use automation in some form when interacting with customers. This figure indicates that the industry has largely addressed the need to implement automation. However, current research findings point to a more complex issue: whether this automation is actually delivering results.

In less than one in four organizations (23%), customers describe their interactions as both highly automated and continuously optimized using CX data. Simply put, the pace of automation adoption has exceeded the speed of developing the intelligence required to manage it. This leads to a widening gap between companies that are implementing automation and those that are continuously refining it.

CX Automation Adoption Outpaces Optimization

Almost every surveyed organization has implemented automation at some stage of the customer journey. Nevertheless, only 24% report that their automated customer experience is 'Very Positive,' indicating that implementation alone does not guarantee improvements for customers.

Organizations do not lack data to bridge this gap. Every surveyed company reports collecting CX data, and 66% now rely on automated processes for its analysis, up from 53% last year. However, data collection and analysis are not equivalent to action based on that data. More than two-thirds (68%) state that they still do not fully utilize CX data, which is higher than the 62% reported in 2025. When CX data is used, it is more often applied to understanding customer satisfaction rather than improving the automation itself. Only 30% use the insights gained to refine the performance of automated interactions, and only 22% use them to determine what should be automated in the first place.

The problem is exacerbated by cross-departmental misalignment. Almost all organizations (94%) face difficulties aligning CX data and feedback across different teams. This results in customer information remaining locked within the department that generated it, instead of influencing automation decisions in other parts of the business.

What Distinguishes the Most Effective Automated Experience

The report draws a clear line between organizations reporting a 'Very Positive' automated experience and those reporting 'Mixed' or 'Negative.' Companies in the 'Very Positive' group much more consistently optimize automation using CX data—49% versus only 8% among companies with 'Mixed' or 'Negative' experiences. They also use richer sources of customer information during and after interactions, including automation performance data from chatbots, voice bots, and Interactive Voice Response (IVR) systems, as well as omnichannel communication transcripts.

Knowing when to involve a human is equally important. Almost all organizations (95%) agree that human agents provide more value than automation or AI in at least one type of interaction, especially when dealing with complex problem-solving, significant business implications, customer vulnerability, or empathy. Organizations demonstrating a 'Very Positive' automated experience are more likely to hand off an interaction to a human (76% versus 68% for those reporting 'Mixed' or 'Negative' experiences), suggesting that the most effective automation strategies are designed with inherent limitations in mind.

Artificial intelligence reinforces this balance, rather than replacing it: 96% of organizations are implementing AI in their CX initiatives, and 85% use it for at least one human-involved scenario—most often for real-time assistance, boosting agent productivity, and for training and coaching. Concerns have not disappeared: 43% of organizations believe customers still prefer human interaction, and 40% think AI struggles with complex, emotional, or risky interactions. These concerns decrease slightly year over year, but they remain a reminder that trust is earned gradually through automation.

Huawei has released the source code for the pretraining, Supervised Fine-Tuning (SFT), and Reinforcement Learning (RL) processes for the openPangu-2.0 model. This release occurred on September 28, 2026, according to information from TMT Post and coverage related to the training stack optimized for Ascend.

The official branding is used as Huawei / openPangu. Previously, the weights for openPangu-2.0-Pro and openPangu-2.0-Flash were released on the Hugging Face platform; however, the current release pertains specifically to the training code path that underpins these MoE models trained on Ascend, rather than releasing new parameters.

The model cards on Hugging Face describe openPangu as Huawei's brand of open artificial intelligence models designed for training and inference on Ascend. The openPangu-2.0-Pro model features approximately 505 billion total parameters with about 18 billion active per token, a 512K context window, and a training budget of around 34 trillion tokens. The openPangu-2.0-Flash model has about 92 billion total parameters, approximately 6 billion active, the same 512K context size, and a comparable budget of ~34 trillion tokens.

Additional details following the training of both models mention combined fast/slow SFT, multi-specialized RL, and Online Distillation (OPD). Architectural features common to Pro and Flash include multi-head latent attention, DSA-plus-SWA layered mix (approximately 1:2), an mHC residual topology with four branches, three-head multi-token prediction, and training using the Muon optimizer.

The September 28th release provides the pretraining, SFT, and RL components as Ascend-specific tools, allowing developers to reproduce and extend this stack instead of merely loading weights for inference. It is important to distinguish between open weights, inference code, and this training code set: the weights were already publicly available; the news is that Huawei is opening up the openPangu-2.0 training pipeline on the Ascend side for the pretrain/SFT/RL components.

For teams working on Ascend, the concrete outcome is an OSS training stack corresponding to the confirmed Pro family models (505B / ~18B active, 512K) and Flash (92B / ~6B active, 512K).

Cricket legend Sachin Tendulkar has officially entered into a multi-year partnership with OpenAI. The goal of this collaboration is to simplify the concepts of artificial intelligence and demonstrate to ordinary users across India how ChatGPT can be applied in daily life.

This interaction aims to make advanced technologies accessible, focusing on practical and understandable examples of use. Tendulkar noted that his learning often begins with questions, and with ChatGPT, he can ask them, develop ideas, and study topics of interest at his own pace, whether preparing for a speech or learning new material. He hopes this will give more people a starting point to independently test the tool based on what matters to them.

The campaign will launch with a humorous video in which Tendulkar challenges ChatGPT to guess his identity in just ten questions. Further plans include launching interactive content, conducting event campaigns, and joint projects with creators, including videos with tech entrepreneur Varun Maiya.

According to Prabhjit Singh, OpenAI's Managing Director in India, engaging with a figure who resonates with different generations helps bridge the gap between technology and trust. This partnership comes amid a sharp rise in AI adoption in the country. OpenAI reported that the ChatGPT user base in India has doubled in the last year, with most users employing the tool for household tasks, writing, and educational purposes.