Connect with us

Tools & Platforms

Agentic AI in Industry: The Technologies That Will Deliver Results

Published

on




Agentic AI promises to move industrial automation from rigid workflows to adaptive, intelligent systems. But the path forward isn’t purely technical. It demands investment in infrastructure, culture, and trust.

The AI landscape is evolving rapidly, and one of the most transformative developments on the horizon is agentic AI. These are AI systems capable of autonomous, goal-driven behavior. Unlike traditional machine learning models that produce a prediction or classification on demand, agentic AI systems can sense, plan, act, and learn over time, continuously improving and adapting to dynamic environments.

For industrial organizations, whether in manufacturing, energy, utilities, or logistics, the potential benefits are enormous: automated root-cause analysis, self-optimizing production lines, predictive maintenance with minimal human oversight, and even coordination across multi-agent systems for complex tasks like supply chain orchestration or grid balancing.

However, realizing these promises will require more than just smarter algorithms. It demands an entire ecosystem of technologies that can support autonomy, context-awareness, and continuous learning in real time, often at the edge. And even with these technologies, there often remain challenges that can impede success.

Enabling Technologies for Agentic AI in Industry

AI agents are autonomous systems designed to perceive their environment, make decisions, and take actions to achieve specific goals. Often, this work is done without constant human oversight.

Unlike traditional AI models that deliver a single output when prompted, agents operate continuously, adapting over time and responding dynamically to changing conditions. To function effectively, they need seamless access to diverse and up-to-date data sources, such as sensor streams, operational logs, and digital twins, to maintain situational awareness. They also must be capable of working in coordination with other agents, human operators, and enterprise systems. That requires interoperable protocols, shared context, and robust communication frameworks.

There are core technologies that can be used to make all of this happen. Some of the most important technologies to consider include:

Streaming Data Architectures: Agentic AI depends on real-time situational awareness. That requires continuous access to fresh, granular data from sensors, machines, and control systems. Traditional batch-oriented pipelines can’t meet this need.

Streaming data platforms, such as Apache Kafka and Apache Flink, provide the foundation for real-time data ingestion and processing. In an industrial context, these platforms can integrate telemetry from PLCs, SCADA systems, or IoT gateways and feed it directly to agentic systems for reasoning and action.

More importantly, streaming architectures enable feedback loops that allow agents to learn from the consequences of their actions in near real-time, which is crucial for adaptive behavior.

Vector Databases and Memory Systems: Agentic AI systems need memory for caching information and remembering past states, decisions, and outcomes. This memory enables reasoning over time, allowing agents to learn and plan more effectively.

Vector databases allow for semantic search and retrieval across time series, documents, or events. Used as episodic memory, they help agents recall relevant information to inform new decisions.

Memory is particularly critical in industrial settings where anomalies may develop gradually, or the context of an event only makes sense about past data.

Model Context Protocols (MCP): Agentic systems must interact with a variety of other systems, such as PLCs, MES, ERP systems, and digital twins, depending on their tasks. Emerging standards, such as MCP, aim to define how agents can manage context windows (i.e., the information they “see”) and utilize external tools (APIs, simulations, UIs) autonomously.

This is critical in industrial environments where the agent must fetch updated process parameters, launch diagnostics, or take control actions in real time. Without a structured, secure way to manage tool usage and context injection, agents will remain brittle and task-specific.

Simulation Environments and Digital Twins: Agents learn best in environments where they can explore and experiment. For high-risk industrial domains, real-world experimentation isn’t always feasible. Digital twins and simulators provide the safe space needed.

Platforms are needed that enable training and fine-tuning of agents in virtual environments. Such platforms can help accurately reflect plant behavior, including physical constraints, failure modes, and control logic.

Edge AI and On-Prem Inference: Latency, bandwidth, and data privacy concerns often prevent sending all industrial data to the cloud. Edge AI platforms enable local inference and decision-making close to the source.

That is essential for real-time autonomy in scenarios like robotic control, line inspection, or substation monitoring. Agentic systems can leverage edge compute to act quickly and only escalate issues to central systems when necessary.

See also: MCP: Enabling the Next Phase of Enterprise AI

Obstacles and Concerns That May Impede Success

Implementing AI agents presents several challenges for organizations, many of which stem from the complexity of integrating autonomous systems into existing industrial environments.

There are several important areas to consider that can impede the deployment of AI agents and limit their benefits. Any organization aiming to use AI agents on an enterprise scale must deal with the following:

Data Silos and Inaccessibility: Agentic AI requires broad access to operational data, including machine logs, sensor feeds, maintenance histories, and more. In most industrial organizations, that data is siloed across proprietary systems and legacy infrastructure. Without integration and normalization, agents can’t learn or act effectively.

Solving this will require both technical and organizational changes, including the use of modern data platforms, open APIs, and cross-functional governance.

Safety, Reliability, and Explainability: In mission-critical environments such as power generation or chemical manufacturing, agents cannot simply “try things and see what happens.” They must be safe, reliable, and explainable.

That creates a paradox. True autonomy implies some level of experimentation, but safety demands predictability. Techniques such as constrained reinforcement learning, human-in-the-loop oversight, and policy-based safety layers are emerging to manage this trade-off. However, organizations will likely find that these techniques are often in their early stage of development.

Skills Gap and Organizational Readiness: Deploying agentic AI systems requires a hybrid set of skills, including machine learning, systems integration, domain knowledge, and control theory. Most industrial organizations don’t yet have this mix in-house.

Training, upskilling, and hiring will be necessary; so will changes to how organizations think about automation. Rather than scripting every step, teams will need to focus on setting goals, establishing guardrails, and monitoring outcomes.

Final Thoughts

Agentic AI promises to move industrial automation from rigid workflows to adaptive, intelligent systems. But the path forward isn’t purely technical. It demands investment in infrastructure, culture, and trust.

Forward-thinking organizations must start laying the groundwork now by modernizing data pipelines, investing in simulation environments, and piloting constrained autonomy. Such moves will favorably position the organization to make use of intelligent agents in the future.





Source link

Tools & Platforms

Relativity Scales Generative AI Availability Across Asia

Published

on


RelativityOne users in five more countries will be empowered with enhanced document review and privilege identification capabilities

CHICAGO, July 7, 2025 /PRNewswire/ — Relativity, a global legal technology company, today announced that two of its generative AI solutions, Relativity aiR for Review and Relativity aiR for Privilege, will now be made available to all RelativityOne instances located in Hong Kong, India, Japan, Singapore and South Korea. Expanding on its previous availability, legal, investigation, and compliance teams in Asia will be equipped with the generative-AI powered document review solution and privilege review solution to help navigate the full spectrum of legal data challenges while reaping the benefits of better infrastructure and privacy.

Asia’s diverse legal landscape presents unique and evolving challenges, and legal teams across the region need technology that can keep pace,” said Chris Brown, Chief Product Officer at Relativity. “Whether it be for litigation, regulatory responses, or internal investigations, Relativity aiR products provide the necessary features to manage large volumes of data more effectively. As adoption grows across the globe, and real-world use cases continue to demonstrate impact, Relativity’s customers and partners can feel confident in the power and practicality of AI in their workflows.”

Enhancing the capabilities of legal teams across Asia with intelligent tools

Customers and partners in five additional countries will now be able to leverage aiR for Review and aiR for Privilege to deliver exceptional efficiency and accuracy in document and privilege review. This regional expansion underscores Relativity’s commitment to providing innovative solutions that align with the evolving needs of legal professionals in Asia and across the globe.

“Customers in Asia are facing a perfect storm — small teams, complex and diverse data sources, multilingual review, and constant pressure from clients to cut costs,” said Stuart Hall, Principal at Control Risks. “The launch of Relativity aiR in Asia couldn’t be more timely, offering Control Risks’ customers a real opportunity to simplify and streamline cross-border investigations and disputes with smarter tools and workflows.”

The introduction of Relativity aiR products in Asia is bolstered by the region’s growing demand for secure, scalable legal technology. Built within RelativityOne, these AI tools allow firms to harness the power of automation without compromising security or performance. By operating in a cloud-native environment, legal and compliance teams can eliminate the burden of managing physical infrastructure, standardize workflows across jurisdictions and redirect resources toward strategic analysis.

In response to the growing volume of investigative matters, organizations will be able to utilize aiR for Review to support a wide range of use cases beyond litigation — including internal investigations into fraud, bribery, corruption and whistleblower complaints. Legal and compliance teams can also rely on the tool for Know Your Customer (KYC) reviews, cross-border data transfer assessments and anti-money laundering efforts. Its versatility extends even further, supporting M&A due diligence, risk assessments, trade secret theft inquiries, white-collar investigations and HR-related matters.

For organizations concerned with data protection, Relativity’s cloud-native products, including aiR, offer peace of mind with enterprise-grade security and privacy controls. Backed by the company’s in-house security team, Relativity embeds protection into every stage of its product lifecycle. This security-first approach ensures that as firms adopt cutting-edge AI tools, their information is properly safeguarded.

Looking ahead, Relativity remains focused on empowering users through innovation, delivering rich insights and addressing their most pressing needs. In the coming months, new capabilities will be introduced within aiR for Review and aiR for Privilege. One upcoming enhancement is aiR for Review’s prompt kickstarter capability, which will greatly reduce manual work related to prompt criteria development. Soon, users will be able to upload case background documents — such as review protocols or disclosure requests—and an expert prompt that drives aiR for Review will automatically be produced, allowing users to accelerate analyses. This feature produces a comprehensive matter overview, including key people, organizations, term descriptions and relevance criteria. From there, teams can refine prompts as needed, accelerating the review process and enabling practitioners to take immediate action.

Additionally, aiR for Privilege users will soon be able to find privileged content faster by automating context building that the AI uses to make decisions. Furthermore, a brand-new entity classifier will more accurately identify and classify the entities within each case. This enhancement will help better identify and define the roles of individuals and organizations in a matter, improving precision and efficiency in privilege review.

Unlocking new possibilities for innovation

To achieve their goals with greater precision and reduced overhead, more than 200 customers have embraced aiR for Review, while over 140 have chosen aiR for Privilege to support their workflows. The scalability and transparent natural language reasoning of this industry-leading technology help customers secure faster results while uncovering deeper insights from data.

KordaMentha, an independent and trusted advisory and investment firm working across industries throughout Australia and Asia Pacific, has transformed its legal discovery approach since adopting aiR for Review. The solution has surfaced insights that conventional methods would have overlooked entirely. A recent case study highlights how aiR for Review enabled a defensible and comprehensive review under a tight disclosure deadline, in total saving 25+ days and reducing costs by 85%. With subject matter experts leading the process, KordaMentha was able to uncover several unanticipated findings that drove organizational change.

“Whether as a renowned center for international arbitration, a market with extensive regulatory and investigative demands, or a source of exponential data growth, Asia is a dynamic region uniquely suited to Relativity’s aiR suite,” said Roman Barbera, Partner at KordaMentha. “Building on RelativityOne’s proven ability to navigate diverse languages and data types, aiR delivers exceptional scalability and insight. We’re excited to deploy this trusted and secure AI solution in a region where KordaMentha is already deeply embedded, and where the need for fast, intelligent and defensible data analysis continues to grow.”

In addition to the current aiR product availability, Relativity aiR for Case Strategy, a cutting-edge solution that makes it faster and simpler for litigation attorneys to extract facts, craft case narratives and prepare for depositions and trial, is currently in limited general availability and is expected to become generally available to all regions with access to aiR products later this year.

For more information about the expansion of aiR availability in Asia, please register for the webinar “Transforming Legal Work in Asia: Introducing Relativity aiR for Review and aiR for Privilege,” taking place on July 22. The webinar will offer a first-hand look at aiR for Review and aiR for Privilege through live demonstrations and real stories from early adopters who’ve already transformed their practices. Request a demo from the Relativity team here.

About Relativity
Relativity makes software to help users organize data, discover the truth and act on it. Its SaaS product, RelativityOne, manages large volumes of data and quickly identifies key issues during litigation and internal investigations. Relativity has more than 300,000 users in approximately 40 countries serving thousands of organizations globally primarily in legal, financial services and government sectors, including the U.S. Department of Justice and 198 of the Am Law 200. Please contact Relativity at [email protected] or visit www.relativity.com for more information.

Media Contact: [email protected]

Logo – https://mma.prnewswire.com/media/445801/new_Relativity_logo_Logo_v2.jpg 



Source link

Continue Reading

Tools & Platforms

Why data center, tech firms are concerned about Chile’s AI regulation

Published

on


34,000+ projects in Latin America.

Contacts

43,000+ global companies doing business in the region.

Companies

102,000+ key contacts related to companies and projects

News

Analysis, reports, news and interviews about your industry in English, Spanish and Portuguese.



Source link

Continue Reading

Tools & Platforms

Player faith in technology shaken by storm around AI line-calling at Wimbledon | Wimbledon 2025

Published

on


When the Wimbledon organisers announced last year that electronic line-calling would replace line judges for the first time at the Championships this year, plenty of criticism could have been anticipated. Some people would take issue with the more sterile landscape on court and the lack of human touch, while the cull of around 300 linesmen and women would also surely be a sore point. It is difficult, however, to imagine they were prepared for the firestorm that has followed its long-awaited implementation at this tournament.

Electronic line-calling, or ELC, which uses automated ball-tracking technology has, after all, long been used in professional tennis tournaments, starting with the Next Gen ATP Finals in 2018. It has been four years since the Australian Open became the first grand slam to utilise the technology and this year, for the first time, the men’s tour, the ATP, is using ELC at all of its events. Although all other men’s clay-court events use ELC, the French Open is now the only grand slam that still employs human line judges.

Instead of this year offering Wimbledon to step into the future, however, the All England Lawn Tennis Club (AELTC) has spent the first eight days of the tournament defending its implementation of the technology.

For the first five days of the tournament the most significant blows were the parting shots from Jack Draper and Emma Raducanu, the men’s and women’s British No 1 players, who each criticised the ELC system following their defeats. Both players believed they had been subjected to incorrect calls. “It’s kind of disappointing, the tournament here, that the calls can be so wrong, but for the most part they’ve been OK. It’s just, like, I’ve had a few in my other matches, too, that have been very wrong,” Raducanu said.

The AELTC maintained that the system was working optimally and that ELC remains considerably more accurate than the line judges it replaced. Wimbledon employs Hawk-Eye, one of numerous ELC providerswhich uses a system that incorporates 10 cameras placed around the court, and which track the bounce of the ball. Hawk-Eye states that its margin of error is 2.2mm. Wimbledon had previously used ELC only as a safety net, allowing players to challenge calls conducted by line judges.

“It’s funny, because when we did have linesmen, we were constantly asked why we didn’t have electronic line-calling because it’s more accurate,” Debbie Jevans, the chair of the AELTC, told the BBC.

Emma Raducanu has not been impressed by ELC at Wimbledon. Photograph: Dave Shopland/Shutterstock

Then came a disastrous series of events on Centre Court. As Anastasia Pavlyuchenkova held game point on her serve at 4-4 in the first set against Sonay Kartal on Sunday, a backhand from Kartal clearly flew long but it was not called out. After a lengthy delay, it emerged that some of the ELC cameras had not been functional on Pavlyuchenkova’s side of the court for some time during the game. The umpire Nico Helwerth opted to replay the point. Around 10 minutes later, after losing that service game, Pavyluchenkova faced a set point on Kartal’s serve.

In the end, the AELTC was fortunate with the outcome. Pavlyuchenkova, who told Helwerth the tournament had “stolen” the game from her, recovered to win both the set and the match, limiting the significance of the error. The AELTC announced in a statement on Sunday night that the ELC had been accidentally deactivated on Pavlyuchenkova’s side of the court by one of the operators running the system.

skip past newsletter promotion

Bright on Monday morning, the Wimbledon chief executive, Sally Bolton, fielded a contentious scheduled meeting with the media, which was almost entirely centred around ELC. Bolton asserted repeatedly that the mistake was purely down to human error, that the protocols had been changed to prevent a similar issue and that ELC has otherwise been working accurately during the tournament. At the very least, the situation with Pavlyuchenkova also underlined the importance of having contingency plans for when technology fails, including the possibility of umpires using video replay.

Since the implementation of ELC, player reaction has largely been positive as it was rolled out on hard courts, with players recognising the greater accuracy provided by the system compared to human errors. However, after numerous dramatic moments during the clay-court season, as some players were frustrated with the differences between the ball marks and the ELC’s judgments, the first week of ELC at Wimbledon has been a difficult one. It is clear that faith in its implementation on the surface has diminished and both privately and publicly, players and coaches have expressed scepticism about its accuracy. As the tournament moves into the latter stages, it remains to be seen if that faith will be restored.



Source link

Continue Reading

Trending