PEX Network’s weekly news bulletin rounds up the latest research, reports and publications in operational excellence (OPEX), digital transformation, artificial intelligence (AI) and automation, business process management (BPM), process mining and process intelligence and more.
This week includes:
- Researchers find AI model “strategically” misleading its creators
- Businesses turn to open-source AI tools to enhance ROI
- Sales orgs struggle to drive commercial success during transformation
- Paper examines process mining as catalyst of digital twins for production systems
- PEX Network publishes intelligent document processing report
Researchers find AI model “strategically” misleading its creators
A new study discovered an AI model strategically misleading its creators during a training process to avoid being modified. A paper from Anthropic’s Alignment Science team, in collaboration with Redwood Research, provided the first empirical example of a large language model (LLM) engaging in “alignment faking” – pretending to be aligned with trained principles but in fact being driven by its initial, contradictory preferences – without having been explicitly trained or instructed to do so.
Anthropic and Redwood Research carried out most of their experiments on Claude 3 Opus, an advanced model which had already been trained to be “helpful, honest and harmless.” Researchers informed Claude that it was now undergoing supplementary training, via a process called reinforcement learning, that sought to make it comply with any query.
The experiment indicated that reinforcement learning is insufficient as a technique for creating reliably safe models. The researchers also found evidence suggesting the capacity of AIs to deceive human creators increases as they become more powerful. Ultimately, if AI models can engage in alignment faking, it makes it harder to trust the outcomes of safety training. “Our demonstration of sophisticated alignment faking in a large language model should be seen as a spur for the AI research community to study this behavior in more depth, and to work on the appropriate safety measures,” the researchers added.
Watch Tyrone Smith Jr., program faculty member/adjunct associate professor at University of Southern California (USC) and founding member of Society of People Analytics (SPA), discuss building trust in AI
Businesses turn to open-source AI tools to enhance ROI
Separate research from IBM indicated that companies are turning to open-source AI tools to unlock return on investment (ROI). The technology giant surveyed more than 2,400 IT decision makers finding that 85 percent of respondents report making progress in executing their 2024 AI strategy with nearly half (47 percent) already seeing positive ROI from their AI investments. The use of open-source tools for AI solutions could correlate to greater financial viability, with 51 percent of surveyed companies currently utilizing open-source AI tools report seeing positive ROI compared to just 41 percent of those not using open-source, according to IBM.
Nearly two-thirds (62 percent) of all respondents indicate they will increase their AI investments in 2025, with 48 percent planning to leverage open-source ecosystems to optimize their AI implementations. Two-fifths of companies not currently utilizing open-source for AI implementation plan to do so in 2025.
“As organizations begin to implement AI at scale, many are placing greater stock in success metrics such as productivity gains, in part because traditional hard dollar ROI benefits have yet to show up on the balance sheets,” commented Maribel Lopez of Lopez Research. “Yet, companies continue to rapidly advance their AI strategies, with no sign of slowing down. Companies now recognize the value of defining specific use cases and optimizing AI projects. They are leveraging hybrid cloud strategies and open-source to drive AI innovation and deliver financial returns.”
Watch Lisa Williams, global operations talent strategy and employee experience director, manufacturing and engineering operations at Dow, reflect on embracing new technology
Sales orgs struggle to drive commercial success during transformation
Only 11 percent of sales organizations have been able to drive commercial success while executing a transformation, according to new research from Gartner. The analyst firm surveyed 234 heads of sales and senior sales leaders to identify productivity levers that are most effective for chief sales officers (CSOs) to elevate sales performance. They found that top performing sales organizations focus on action-centered insight and design, radical role simplification and adaptivity by design.
“In the past two years, sales organizations have undergone an average of four transformations,” said Greg Hessong, senior director, analyst in the Gartner Sales Practice. “While CSOs have a variety of mechanisms at their disposal to drive productivity, it’s often unclear which are most effective, especially in a rapidly shifting landscape.” That ambiguity, combined with the relentless need to hit quarterly targets, leads most CSOs to default to seller-centric solutions that have worked to drive productivity in the past. “Unfortunately, these efforts are yielding diminishing returns as many sellers are feeling overwhelmed, unable to absorb the volume of change required.”
Watch Nao Anthony, Commonwealth Bank, and Tariq Munir, PepsiCo, explore the complexities and challenges of digital transformation
Paper examines process mining as catalyst of digital twins for production systems
A new research paper examined a selection of topics uniting the fields of process mining and digital twins for manufacturing systems, providing a reflection on current challenges and opportunities that can be seized in the near future.
“The advantages in terms of productivity from recent investments toward higher automation come along with an escalating complexity in manufacturing systems,” the author wrote. This prompts for leveraging digital support assistants, notably digital twins, for production planning and control tasks. “Process mining has emerged as a valuable tool in the realm of digital twinning as it proved its efficacy in tasks such as model generation, trace profiling, and performance evaluation. However, several challenges persist in different methodological and application areas, presenting valuable opportunities for both academia and industry.
Watch Tony Benedict, president of the Association of BPM Professionals International, discuss breaking down silos in process management
PEX Network publishes intelligent document processing report
PEX Network launched its new intelligent document processing (IDP) report, exploring the intricacies of the modern IDP landscape. The report paints a comprehensive picture of the key trends shaping IDP today, outlining use cases, success factors and the biggest challenges of adopting IDP in organizations. The report features case studies from Deutsche Post DHL Group, Manchester University NHS Foundation Trust and Ecclesia Group as well as guidance and insight from experts in the field.
“IDP solutions are becoming an integral part of enterprises looking to optimize their document-centric workflows because they provide an efficient combination of modern technologies such as computer vision, machine learning and natural language processing (NLP),” said Anoch Mane, research analyst at QKS Group.
READ MORE PEX RESEARCH & REPORTS NEWS
[inlinead-1]