HomeAI News
Why physical operations are uniquely positioned for AI transformation

Why physical operations are uniquely positioned for AI transformation

Hayo News
Hayo News
November 29th, 2023

Physical operations are at the heart of our economy, powering everyday life through the flow of goods and services — whether it’s transporting those goods, building infrastructure or keeping utilities running. These organizations exist across many industries including construction, transportation and logistics, food and beverage distribution, and more. In fact, they make up more than 40% of the U.S. GDP, but have been historically overlooked by technology that’s purpose-built to solve the complex challenges they face.

While technology has revolutionized how most people work and live, physical operations have continued to rely on outdated paper documents and manual processes. Now, that is finally changing. As advances in technology make it possible to capture more data from assets in the field and bring that information into the cloud, operations leaders are embracing digitization and transforming their physical operations into connected operations.

Companies that are embracing this transformation find themselves in a unique position to leverage an ever-expanding slate of AI and machine learning tools to drive better outcomes for their customers, employees, and ultimately, their bottom line.

So, what makes these industries so ripe for AI transformation? Ultimately, it comes down to data.

In the world of physical operations, companies deal with massive, petabyte-scale data in the cloud — and that volume can significantly multiply at the edge. What’s more, this data is multimodal, meaning it’s not just a single type of data like GPS location. It could be temperature data, inertial sensor data, text data, video files and more. This is an incredibly complex and rich set of information that must be analyzed together for real, tangible insights to be uncovered. AI is the natural fit for extracting value from this operational data, and with recent advancements in these tools, it is possible to extract far more value in significantly less time than with traditional analytics.

Foundation models, which are capable of a wide variety of tasks in a given domain, are a logical starting point. A main benefit to working with foundational models is that they can provide a quick start to any company, regardless of their size or budget. Companies can fine-tune foundation models with their own proprietary data, or use them to distill smaller, highly specialized models tailored to their specific operations. Because these subsequent models are orders of magnitude smaller, they can be run far more cost-effectively, or even on the edge.

Another consideration here is that these organizations don’t operate in what most would deem a “traditional” work or office environment. While some managerial roles are done in an office or at a desk, the majority of this workforce is on the frontlines — picking up trash, building highways, delivering groceries or driving thousands of miles in rural areas. Preventative insights and real-time alerts powered by AI are even more critical for these employees, who rely heavily on remote communication to do their jobs both safely and efficiently.

When you start to think about model development for this audience, engineering teams need to be sure that their specialized model is capable of working accurately and efficiently across a diverse array of tasks, equipment, geographic areas, temperatures, languages, measurement scales (i.e. metric vs imperial) and more. Moreover, models for physical operations are beyond text; they must be multimodal, working with a diversity of data from the field in order to provide helpful insights.

AI for impact: Accelerating time to insights and improve outcomes

True AI transformation for physical operations isn’t about implementing technology for technology’s sake. The adoption of AI should remain focused on tangible impact — from real-time alerts to predictive risk modeling. There are a few specific use cases where AI can be particularly impactful within this industry, for example:

Safety improvements: AI models can run on-dash cam devices in commercial vehicles to detect unsafe driving behavior — like tailgating or using a mobile device — and trigger audio alerts on the edge to correct drivers, in real-time. This can actually help prevent dangerous road incidents from occurring. For example, Samsara products helped prevent 120,000 crashes in 2022 alone, and customers like DHL Express have reported a 26% reduction in accidents, and subsequently a 49% reduction in related costs since using our AI Dash Cams.

From preventative to predictive maintenance: AI can be used to transform preventative maintenance to truly predictive maintenance. By leveraging existing trends within your organization’s data, machine learning models can reliably predict when a piece of equipment or a vehicle might need maintenance and proactively alert the user. This not only saves money on costly repairs, but allows mechanic teams to work smarter, not harder.

Automated workflows: Many operations workers rely on their mobile devices to perform day-to-day tasks, from proof-of-delivery receipts or Driver Vehicle Inspection Reports (DVIRs). AI can be used to streamline and simplify these processes by automating particular tasks and surfacing curated workflows for specific employees, cutting out the noise and only providing them with the information they need for a given task, at a given time.

What does the future hold?

The future of AI in physical operations is all about filling knowledge gaps and automating key workflows. With many organizations still using pen and paper logs or outdated spreadsheets, AI can provide visibility into trends within their operational data they didn’t even know existed and can present this information in a way that is easy to understand and actionable.

When it comes to the use of large language models (LLMs) like ChatGPT or Llama in physical operations, it’s likely that smaller, more specialized models will be derived from these larger models to serve specific use cases. The use of AI copilots within these industries will also be exciting to watch — leveraging LLMs can be particularly impactful when it comes to providing context-rich assistance or automating more mundane administrative tasks. LLMs are also an important enabler for allowing non-experts in the field to interface with complex technology using plain, everyday language.

The world of physical operations is exceedingly complex and ever-changing. It can be a challenge to wrap your head around the vast array of data formats and applications, especially as an AI team, but if you do, there’s so much to be done to simplify and optimize this space.

Reprinted from Evan Welbourne, SamsaraView Original


no dataCoffee time! Feel free to comment