How AI can impact healthcare, workflows & manufacturing after COVID-19

The crisis caused by COVID-19 has forced many people to adapt to significant lifestyle and work changes; however, given the current conditions it is difficult to accurately predict how and when the return to normalcy will take place, if it will at all.

The tech sector has been largely insulated from the more adverse effects of the crisis.

What’s more, the widespread disruption of workflows in other industries has given technologies, such as AI, opportunities that would be subject to more resistance under normal conditions.

AI gains a foothold in healthcare

For a long time AI has been heralded as the next big innovation in the healthcare industry. From diagnostics to treatment tracking and planning, AI has the potential to dramatically improve quality of care and lower costs.

However, due to the nature of the healthcare industry, development in this area has been hamstrung by a mix of regulatory pressure, entrenched processes, and disparate, non-uniform data. The crisis lifts these burdens.

While regulations won’t change overnight many governments are turning to tech to track the extent of the spread of the disease.

Recently, the NIH has begun to regularize and centralize the storage of data related to COVID-19 treatment, recovery and future health outcomes through the National COVID Collaborative (N3C), with many collaborating members in other countries as well.

This effort provides a single well maintained source of data to effectively deploy machine learning methods on, and consequently produce valid and meaningful results.

Biofourmis, a Boston-based health analytics startup, has been contracted by the University of Hong Kong to provide their AI enriched health monitoring and alerting platform to remotely monitor COVID-19 patients in quarantine.

Their platform eases strain on healthcare resources while still allowing for rapid response to deteriorations in a patient’s condition; in addition it also provides ample data to analyze to understand the progression of the disease over time.

In a more proactive direction, Healthmap, a project originally designed to predict influenza outbreaks was able to produce warning about a new outbreak and a potential pandemic on December 30, 2019.

BlueDot, a Canadian startup, was also able to predict where the likely spread targets would be based on aggregated travel data.

Results like these are promising, and due to the intensity of the current pandemic it is very likely that many health organizations and governments will be willing to pay into a program that could potentially allow them to avert crises entirely.

AI has also been deployed closer to the medical field as well, two teleradiology systems, Qure.ai and Lunit have been employed by doctors to analyse chest X-rays for the purposes of triaging patients. These platforms offer significant safety advantages, as they limit exposure to contagions while also providing a rapid and accurate assessment of the patient’s condition.

A lack of testing capacity in France forced some hospitals to switch to using chest X-rays as a fallback solution to triage patients, a process that normally takes up to a couple days to get results back.

Lunit’s platform is able to achieve 95% accuracy within 10 minutes, and due to that they have since been actively deployed by some of the largest hospitals in France, Italy, Portugal, and elsewhere.

Qure.ai has similarly been deployed in Italy, Mexico, and the US. And once the value of their services has been fully realized it will be difficult to stop using them.

While these two tools are effective for the triage of existing cases, researchers at NYU have been able to accurately predict how severe a case will become. In this study they assess various machine learning methods to estimate the severity of the virus, based off of a set of trackable biomarkers.

While methods like these are still in their infancy, the abundance of clinical data gathered by the platforms and projects that have been deployed in the field due to this crisis can easily be used to improve the reliability for COVID-19 cases and potentially other illnesses as well.

Coupled with many hospitals now willing to adopt AI based tools into their internal processes (and those that aren’t will be swayed by the trust generated by the ones that do), we could start to see the more data-driven approach to healthcare that has been perpetually just around the corner for the past decade.

Transforming business processes

For those who can, remote work has become the norm. And while many pundits have been intimating at this trend sticking around for long after the crisis, there are still significant hurdles faced by companies trying to move away from the office.

For one, there is a substantial lack of immediacy, especially as it relates to communication, scheduling, and management.

A significant barrier to the adoption of AI in more traditional sectors has been just how efficient their processes had become, making the switchover costs too great a hindrance. COVID-19 has changed that calculus.

Routine tasks and processes are starting to show cracks and productivity suffers as a result of the now increased lag times. Many companies are starting to see the value that automating some of their processes could have in the short term.

Creating numerous opportunities for new AI-based technologies to be developed and deployed to solve the problems that arise from managing the transition to greater automation while ensuring the quality of workflows in the short term.

Apromore is one such startup. Founded late last year, its AI-based process discovery system can help businesses find and track anomalies, but perhaps more importantly it use of AI can also explain them to some degree, helping to determine the root causes behind delays which can help businesses focus their automation efforts on their key pain points first.

Since all communication passes through the internet in the modern remote work ecosystem, it has recently become possible to implement methods to more appropriately track and optimize the productivity of workers.

Enaible is another platform that is creating ways to better judge how employees are performing by creating an AI system that can conform to the differing work styles of individual employees and determining how productive they are.

The benefits of employing a system like this are manifold, especially in remote work environments, as it becomes easier to determine which employees are more capable of maintaining work effectiveness; which need more attention from other team members; and where there are performance fluctuations due to external factors or internal team dynamics.

In fact, AI can aid in understanding underlying team dynamics, and even improve them. This market is drastically underserved, despite the substantial focus on building better teams and improving collaboration in recent years, the nature of face to face interaction has been a major blocker to the ability for digital services to inject themselves into the process.

It should be obvious that a transition away from offices and face-to-face meetings would pave the way for a disruptive transformation in how teams communicate and collaborate. The best part of this latter change would be its ability to apply to both rote, routine tasks and highly creative and technical ones.

For example, software development could even be greatly improved by an AI-based system that can be utilized as a focus for collaborative development and team enrichment by providing a distilled view of what needs to be done, shaving time off meetings or code reviews and allowing for more productive hours in the day.

And this is just the tip of the iceberg, unlike healthcare, AI’s penetration into processes control has mainly been due to a lack of data and a clear insertion point.

Now that every aspect is digitally defined, data collection and deployment are almost trivial, leaving ample opportunity for a disruptive startup. Assuming, of course, the trend towards delocalization holds.

Automation in manufacturing

It is well known that the manufacturing sector will benefit from more widespread automation. But it is unlikely that we will see much investment in the short term, as the volatility, supply chain disruptions and general decrease in demand caused by the crisis will limit investments in automation for most industries.

But some will see a boom caused by the changes in how customers will need to interact with their services. Particularly, retail and logistics could see a significant increase in automation coming out of the crisis.

Ecommerce is now more popular than ever, and an increase of demand coupled with an increased difficulty in getting employees, due to social distance and health requirements, would see them turn towards robots to help them bridge the gap.

Fetch Robotics, a startup producing autonomous mobile robots, had seen a 63% increase in inquiries due to the crisis in March. Some of which have paid off, as they have recently deployed a number of AMRs to airports to automate the disinfection of surfaces.

Regardless of the rate of recovery most companies will begin to see this shift, which is in line with long term trends, due to either a necessity to continue production despite health concerns (a problem solvable by augmenting the human workflow with AI and robotics) or due to the low switchover costs coming out of a contractionary period for most manufacturing companies.

The latter also benefits from a smaller public relations cost from increasing automation, which is the other primary factor inhibiting its adoption.

The current pandemic has negatively impacted many people, and will likely continue to do so for an extended period of time, and exacerbated many problems endemic to many aspects of public life. But with these problems brought to the fore so have the opportunities to try out new solutions.

Artificial Intelligence is uniquely poised to provide solutions for the near term problems as well as the long term systemic changes. Its immediate value is now readily apparent as many of the factors inhibiting adoption have been lessened considerably.

The future for AI is looking just as bright as before the crisis, if not more so.

Disclosure: This article is brought to you through a client of an Espacio portfolio company