Cybersecurity in healthcare is responsible for protecting the data that represents the life’s story of patients and infrastructure to enable proper care. Managing and securing the plethora of edge devices and the interoperability of all the technologies is an increasing challenge. There are four steps to take to enhance your healthcare cybersecurity: select a framework, leverage depth in defense, automate where possible, and test your environment.
Analysis is only available to clients at this time.
*When vendors’ names or quotes are shared as examples in this document, it is to provide a concrete example of what was on display at the conference or what we heard doing our research, not an evaluation or recommendation. Evaluation and recommendation of these vendors are beyond the scope of this specific research document.
To effectively integrate AI into healthcare, focus on three key areas: risk, impact, and value. Achieving a Patient 360 view requires orchestrating various tools. AI is embedded in many healthcare solutions including those for asset location, employee safety, and security. Always have a strategy to integrate AI into workflows. Successful integration depends on strong partnerships and clear communication about AI capabilities and limitations.
Target Audience Titles:
Chief Supply Chain, Logistics Officer, Procurement, Technology, and Data Officers
Supply Chain, Logistics, Procurement, Technology, BI and Data Science Directors
ERP Specialist, Supply Chain IT, Data Scientists, BI and related managers
Key Takeaways
Generic AI models don’t understand logistics-specific challenges.
Inconsistent, incomplete, and manually entered data hinder AI’s effectiveness.
Poorly structured processes and a reluctance to adopt AI-driven solutions slow innovation.
Onboarding new suppliers and standardizing systems remains difficult.
We took the most frequently asked and most urgent questions straight to the logistics and supply chain experts in the industry. This Whisper Report addresses the question regarding the biggest challenges using generative AI in supply chain and logistics. The first challenge, however, is not unique to that industry nor is it unique to generative AI. This challenge applies to a all analysis and analytics including all forms of AI – generative or not regardless the size of the models. Put simply, no matter how many ways you state it, when you put garbage data in you will get garbage results.
Given the dominance of a common answer, this raises the question, is the sector of logistics and supply chain in worse shape versus other industries? More specifically, is the data itself within logistics and supply chain the problem and if so, why? Put simply and as depicted in Figure 1, the challenges go far beyond the data. As Don Addington of Cloud 9 Perception put it, “in logistics space there is a level of complexity that is more complex than others.” These complexities come in for the following reasons.
Data doesn’t exist
There is an ideal digital world which is very different from the physical world. As Owen Nicholson from Slamcore pointed out, “If you are not seeing real world deployments with all the gnarly things that go wrong you are only creating idealized models that don’t work in the real world.” Distribution centers are full of human and robot workers as well as machines from multiple manufacturers. Unlike construction, many of these machines are in the same building they entered at the start of their usefulness as brand new machines long before generative AI term existed. Logistics is not the neat and tidy world of fintech transactions.
Data is inconsistent
As Ben Tracy of Vizion pointed out, “(many) skipped a few fundament steps, being useful and being reliable… They don’t monitor data quality, they don’t have consistency amongst data formats, and their systems are not exportable for the data that is inside of them.” Or what data professionals call it- ‘good old fashioned data quality’. To put it in the simplest terms possible, we all learned early in elementary school you need data in the same units to perform any math over the data. You do not add inches and feet together. You cannot add meters and feet together. You don’t speak globally about time without time zones. But perhaps most important, you cannot create data quality nor can you analyze data you haven’t or cannot export.
Data is manual and miss-keyed
If you are wondering how bad that data can be, Dawn Favier of Green Screens provided some hard facts, “its not uncommon to flag 35% of their (customers) data as dirty. Dirty meaning miss-keyed data, something tagged as full truck load when its partial.” Obviously, if one looked at data for a half truck and leveraged for a full truck, the resulting analytics are useless. With 35% of one’s data being dirty, there is work involved before you can even hope for insights.
Data lacks historical context For any AI to be successful, you need massive amounts of data over a very small problem so the mathematics behind the AI can provide useful information. As Atit Shah of Chetu explained, “
Even if you have the right collection of data, you can generate incorrect forecasting. A lot of people do not have a huge history or the history of the records so they go into the gen AI because everyone is doing it but it doesn’t meet their expectation.“ No matter how powerful the technology, all forms of AI need good data. Furthermore, the data must have context to be useful for any advance form of AI including generative AI.
Bad Processes One obvious reason for messy data is the messy, manual, and imprecise or undefined processes it represents. The biggest challenge as Bill Driegert of Flexport shared, is simply, “not slapping it (generative AI) on bad processes. There needs to be a lot of process engineering required to leverage AI.” If process re-engineering and establishing a clean data fabric is your organizations Mt. Everest, TBW Advisors LLC offers a lot of first-hand experience and expertise to teams and executive via inquiry. Any clients at this phase should schedule an inquiry to receive guidance. We will set up a plan of inquiries during your journey to give you any guidance we may have or can gather to assist you. The plan will cover milestones including but not limited to strategy reviews, presentation reviews, and architecture reviews. It is not an area to go through without a guide on your side even if the work is outsourced.
Resistant to change
It is always important to consider the culture of any organization when executing or desire to execute change management. As Erica Frank of Optimal Dynamics put it, “need to take a healthy assessment, how resistant are we to change, how are we going to challenge this from the top down.” As with any change management, executive buy-in with a business objective are critical to success. AI for the sake of AI is always a bad idea.
Perhaps the reason many in this space are resistant to change is the change is constant. As Jason Augustine of WNS put it, “Environment keeps changing every 3-6 months”. Thus discovering opportunities to align and integrate the transformational changes into these already occurring network constant changes is a less tumultuous approach.
Human Machine Interaction
Logistics, like manufacturing and construction, has a lot of machines in the loop. Those machines may or may not be intelligent machines. Thus as Dr. Mario Bjelonic of Rivr.ai shared, “the challenge will come up in terms of how the humans and robots will act as a team together.” Optimizing the total solution over this shared space is the true goal. But as one organization is optimized, what about working between each organization?
is the on boarding suppliers cannot be done by AI”. That’s correct. Bringing each and every machine into the system, or each and every supplier and the complex of array of data that that suppliers managed to coalesce together IS ITSELF NOT standardized thus cannot be automated.
Can’t use Generic Gen AI
As Balaji Guntur of Hoptek pointed out, “Most of the models are very generalized.” “AI is data hungry, and you need to train it on real data. The biggest challenge Generative AI in logistics is that the generative models don’t know what logistics is doing. This is the main challenge,” Aviv Castro, Sensos. In summary, as best put by Nykaj Nair of Sugere, “you need data highly accurate data that is relative to the companies supply chain.”
With all the challenges discussed, it may seem discouraging. It is important to realize the significant opportunity awaits thus easily providing business justification for the work to transform – carefully. As Justin Liu of Alibaba.com put it, “we are continuously adopting AI into our workflow into our latest and greatest features and functionalities to do their business more efficiently.” Rye Akervik of Shipsi believes the value is, “in adding it as a first layer to understand the (customer) issue.” Mick Oliver of Dexory shared, “We don’t see it as a challenge we see it as an opportunity and provide insights based on that data.” Rich Krul of Hoplite observed that the intelligent systems are, “way more efficient, people get their answers a little faster and thinks that is a good thing for the industry.” Most importantly as Georgy Melkonyan of Arnata pointed out, “Shouldn’t fear it (AI) is going to take your job, ai will not replace your job. The people that use ai are going to replace your job.”
*When vendors’ names or quotes are shared as examples in this document, it is to provide a concrete example of what was on display at the conference or what we heard doing our research, not an evaluation or recommendation. Evaluation and recommendation of these vendors are beyond the scope of this specific research document.
