How AI is improving operational efficiencies across oil and gas, healthcare, and more

Artificial intelligence (AI) is serving companies across a broad array of industries. It’s not about replacing human workers with machine counterparts but rather helping businesses operate more effectively in their fields and expand to new horizons and capabilities. I spoke with CEO and founder AJ Abdallat of Beyond Limits, an enterprise AI solution provider, to see how AI has been fulfilling its promise in the business realm.

SEE: TechRepublic Premium editorial calendar: IT policies, checklists, toolkits, and research for download (TechRepublic Premium)

Scott Matteson: How is AI playing a role in operational efficiency in the oil and gas, energy, manufacturing, and healthcare fields?

AJ Abdallat: AI is playing a critical role in operational efficiency across the energy value chain to optimize resource production, democratize domain expert knowledge, and increase value while reducing environmental risk. In the oil and gas sector, AI is enabling companies to optimize their production and improve asset maintenance in many ways, such as pinpointing drilling opportunities, inspecting pipes for problems with self-navigating robots, and predicting equipment wear and tear. 

For example, AI-powered sensors are often installed within permanent downhole gauges inside oil wells to monitor well pressure, temperature and integrity. AI is also helping companies in the sector better predict changing global demand to streamline supply chains and distribution. Companies can use AI to predict how an oversupply of resources or upcoming weather conditions in a certain region could impact the market there.

Energy production has become so difficult to predict and manage that California used to pay Arizona to take excess solar power they’d created. Challenges like these have spurred significant investment in AI to improve operational efficiencies by helping energy providers better balance loads, predict supply and demand, and otherwise optimize energy production, storage, and distribution. Additionally, AI is helping utilities providers better manage hazardous situations, such as with maintenance recommendations and scheduling, trip and fault prevention, outage prevention, and root cause analysis. This is particularly critical now as utilities providers are taking an increasingly conservative approach to hazardous situations, especially since PG&E’s landmark criminal conviction in March for its faulty power line igniting California’s deadly 2018 Camp Fire.

In healthcare, AI is improving efficiencies at both the public health level and individual level. At the public health level, AI-powered forecasting models are helping to predict the impact of COVID-19 on medical facilities and their patients while enabling facilities to determine logistical responses despite challenges with insufficient and continually-changing data. 

For example, a lack of homogenous data has made it difficult for hospitals to predict resource allocation needs, particularly for personal protective equipment. Beyond Limits’ AI forecasting model is helping them better determine the percentage of patients that may need ventilators and extracorporeal membrane oxygenation (ECMO) in the ICU and dialysis units. A few other similar examples include Penn Medicine’s COVID-19 capacity planning tool, CHIME; and Washington State’s predictive dashboard, a COVID-19 risk assessment tool. At the individual level, AI is being used to equip patient monitoring patches with AI technologies. Deploying these AI-assisted diagnostics algorithms in edge environments allows for real-time analysis of patients’ vital signs and recommended actions.

SEE: US parents now more accepting of robots in children’s healthcare, study shows (TechRepublic)

Scott Matteson: What are some unique aspects to each?

AJ Abdallat: In the oil and gas sector, companies have become increasingly reliant on AI since the shale bust in 2015-2016 as they required significant operational efficiency improvements to avoid bankruptcy. Similarly, recent challenges and low-price markets have revealed costly operational inefficiencies within the sector. At the start of the pandemic, companies quickly realized their immediate need to improve operational efficiencies and turned to AI to improve profitability, making up for the decreased market demand. British oil supermajor BP is a different story, however. BP began leveraging AI years before the pandemic, primarily to help decrease risk to personnel and assets following the Deepwater Horizon oil spill in 2010. Among other examples, BP is now leveraging an AI-powered well management system for detecting sand, which often accumulates around certain oil wells and over time can erode equipment and cause catastrophic events, increasing risk to personnel and operations.

In power, AI can be particularly valuable for natural gas power plants because their efficiency is extremely dependent on environmental conditions like temperature and humidity. AI can help predict these conditions and help human operators make more educated adjustments to bring a plant back in line with planning objectives. Additionally, rather than only relying on their own personal knowledge and experience, operators can leverage a cognitive AI system that combines encoded knowledge from a range of domain experts with machine learning techniques. This technology allows the system to “think” like an engineer and provide expert guidance to decision-makers. Cognitive AI systems can also provide an audit trail so human operators can understand how the AI reached its recommendations, as well as retrain the system with new business logic and knowledge to improve its decision-making in the future.

AI can be particularly beneficial for energy, healthcare, and other highly regulated sectors because AI systems can account for regulations and ensure its recommendations are aligned with those requirements. In healthcare, AI adoption has increased dramatically during the past year with more policymakers, doctors, nurses, and patients using AI technologies quickly and at scale in a short time frame out of necessity. Individuals are now using virtual and AI-powered healthcare services and wearing AI-enabled monitoring devices where they would have traditionally visited a doctor in person. With the rapid improvements in IoT and connectivity, the number of connected digital healthcare devices, as well as the volume of collected data, will continuously increase in the coming decades. Sophisticated AI technology and data-driven systems will be necessary for helping medical experts optimize their time and effort when observing and analyzing this meaningful and important data.

Scott Matteson: What are the challenges remaining to be solved?

AJ Abdallat: A prominent barrier to AI adoption across all industries is a lack of trust in the technology—whether that’s a lack of trust in its value, concerns about its biases, or fears about it replacing the workforce.

Many companies have only used industrial AI at the proof-of-concept (POC) and minimum-viable-product (MVP) scale, primarily because they haven’t been able to justify major investments in AI without a clear, proven demonstration of value. However, in the past year, many companies have started to see that value with their POCs and MVPs, and they’ve also gained a better understanding of how AI can be applied to various areas. That said, I look forward to seeing increasing adoption in 2021. 

SEE: IBM creates knowledgeable NLP system and adds AI governance capabilities to Watson (TechRepublic)

The widespread concern about AI bias is another major barrier to AI adoption and effectiveness. In fact, according to a recent report by DataRobot, 42% of AI professionals in the US and UK are “very” or “extremely” concerned about AI bias (which could be biased based on race, gender, age, location or specific data structures). This problem stems from how many AI systems are “black box” solutions, meaning they don’t provide visibility into their data analysis and how they reached certain decisions or recommendations. (The DataRobot report also found that 38% of AI professionals still use black box systems.) To eliminate AI bias—and increase AI trust and adoption—organizations need to leverage “glass box” AI solutions that provide clear audit trails explaining the reasoning behind recommendations and showing the evidence, risk, and confidence behind decisions. This allows users to see how the AI system reached a potentially flawed decision, and also makes it easier to correct the problem moving forward.

These audit trails are designed to be understood by people and interpretable by machines, serving as a partner rather than a replacement to human workers, and promoting continuous improvement throughout the organization.

Integration challenges have also been limiting to legacy industries such as energy. These industries have matured large-scale assets that were planned and constructed prior to this age of digital transformation. The foundations of these assets were not built to operate advanced digital technologies such as AI and ultramodern sensors. Companies transitioning to digital strategies have been hindered by aging computing networks, lack of sensor data, and incompatible systems. Looking ahead, companies will need to lean on their innovation teams to constantly consider digital technologies when planning for future platforms and large-scale assets so they can more easily integrate advanced AI and future-proof operations. 

Scott Matteson: How should IT professionals align themselves with these solutions and methodologies? 

AJ Abdallat: IT professionals need to leverage glass box AI solutions in order to develop trust and confidence in the solution without needing to do so blindly. Additionally, IT professionals should review the AI solution’s audit trails regularly to ensure there aren’t any biases or inaccuracies—and after confirming this consistently, IT professionals can follow the solution’s recommendations with confidence.

The shift to remote work has added another challenge in terms of user education, especially because many organizations have escalated their AI strategies since the start of the pandemic. The shift to remote has made it more challenging for senior-level workers to provide immediate guidance to junior-level workers. Therefore, it’s critical for junior-level users to diligently review the audit trails to better understand the outputs without having to call an expert for support.

Experienced IT professionals can play a major role in improving AI solutions and reducing the enterprise knowledge gap by sharing their expertise with the solutions. There is a concerning and growing knowledge gap in many industries including oil and gas, manufacturing and energy as many leading industry experts are retiring or set to retire soon. The incoming, less experienced IT professionals stepping into those roles, will need access to that lost senior-level expertise to keep operations running effectively. Fortunately, human knowledge can be embedded in many AI solutions so that knowledge is democratized across the entire workforce.

IT professionals should provide technical validation for the AI technologies to ensure these advanced applications can be integrated or layered on top of the company’s current systems. AI with deployment versatility is key, allowing the AI system to work in the cloud, on-premises, or at the edge in certain scenarios. IT professionals can also leverage data scientist expertise from the AI solution providers to ensure the AI systems are fully configurable by company operators with user-friendly interfaces. 

Scott Matteson: Where is this trend headed?

AJ Abdallat: AI technology can already leverage domain expert knowledge by embedding it into the system and then democratize that knowledge across an organization. The next layer is localizing and aggregating that knowledge to allow multiple users across multiple organizations to modify and add knowledge to the framework based on their unique situations. This is valuable because the AI system can learn from the knowledge inputs and combine the inputs with data to gain a broader understanding of different assets and use cases.

AI systems will become significantly more flexible and intelligent as they interact with more domain experts, problems, and data. Eventually, AI systems will be able to identify if decision-makers implemented or declined their recommended actions, whether the action taken did what it was supposed to do, and if the system was able to learn from that remediation action.

One industry that is well-suited for increasing AI adoption is the supply chain. The supply chain has an expansive network of people, businesses, and modes of transportation involved, and much of the system is currently manual (e.g. people still make phone calls to inform others about when trucks or ships are arriving with goods). This approach is just begging to be automated by AI, especially when you consider the amount of data used and the enormous scale of the supply chain industry.

How AI is being used at one hospital

I discussed the role AI can play in the medical field to learn some subjective examples, benefits, and challenges with Shana Bellus, director of admitting operations at Tufts Medical Center in Boston.

Scott Matteson: Can you provide a few examples of how you’re using AI at Tufts Medical Center?

Shana Bellus: We are utilizing Olive’s AI technology to facilitate patient flow through our community COVID testing site location at Tufts Medical Center. Our testing site operations initiated in March 2020 and have continued to evolve in response to the trends of the pandemic. Our volume has most recently peaked; we are now testing more than 500 patients a day. In order to serve the significant growth in the demand for testing that we anticipated throughout the latter months of this year, maximizing efficiencies in our workflow was critical. To achieve these goals, a few months ago we made a decision to pursue an innovative approach that would be a first for the organization: AI. 

SEE: 7 predictions for how technology will shape healthcare in 2021 (TechRepublic)

Current-state patients now have the autonomy to provide their information to the hospital directly and at their convenience, through submission of an online pre-registration form hosted on the hospital’s website. Following form submission, an Olive AI bot retrieves the information submitted and creates a pre-registration account for the COVID test that will be administered to the patient. The bot also documents the patients’ responses to the associated clinical questions required, within the hospital’s clinical EMR system. As a result, the point-of-service turnaround time is significantly quicker when the patient arrives for their test; patient wait times have materially decreased, and our testing capacity has increased.

Scott Matteson: What are some benefits AI can provide hospitals like Tufts Medical Center?

Shana Bellus: The use of automation allows the ability to refocus valuable human capital toward more meaningful initiatives versus the routine work that can be replicated by AI. In addition to the reliability and operational efficiencies gained, AI supports a positive patient experience and promotes employee satisfaction and engagement.

Scott Matteson: What are some challenges you’ve experienced with implementing AI?

Shana Bellus: Considering the novelty of using AI at our organization, we want to ensure everyone has an accurate understanding of what AI is, what it can help us achieve, and the purpose the organization aims for AI to serve. So far there has been a lot of expressed excitement regarding the possibilities that exist to further expand the reach of AI across the entire Wellforce system. Our greatest challenge will be how to prioritize the various opportunities identified.

Data, Analytics and AI Newsletter

Learn the latest news and best practices about data science, big data analytics, and artificial intelligence.
Delivered Mondays

Sign up today

Also see