Mitigating the Challenges of Integrating AI into Existing IT Infrastructures

Integrating new systems and applications into your organization’s IT infrastructure can sometimes feel like you’re setting up a Rube Goldberg machine. Everything needs to be pre-planned and meticulously set up in order to achieve the result you are after. One small misstep and suddenly you can be facing a chain reaction of your other systems being set off or it could cause something to bump out of alignment and cause a complete shutdown. For the important and personal work done in healthcare, either reaction can be catastrophic to your patients and your staff. The only way around this is more meticulous planning, so you can foresee the challenges before they happen and plan out how to mitigate the effects.

For some help on how to do this, we reached out to our incredible Healthcare IT Today Community to ask — What challenges do healthcare organizations face when integrating AI into their existing IT infrastructure, and how can these be mitigated? The following are their answers.

George Dealy, VP of Healthcare Applications at Dimensional Insight
Extending large language models and Gen AI with data already available in healthcare organizations’ systems will be among the biggest opportunities for AI in healthcare. General-purpose LLMs can be supplemented with both healthcare-specific content, such as medical literature, as well as data from operational and analytical applications. But getting this right comes with challenges, including protecting patient privacy and ensuring the validity of data produced by Gen AI. Putting in place disciplined, yet practical, data and AI governance structures coupled with ongoing efforts to monitor and improve data quality will help to mitigate these risks.

André Castro, Manager, Secure AI Products at Protegrity
AI does not change the nature of the adoption challenges from traditional data analytics, which can be summarized by data governance maturity issues. What it does instead is to massively incentivize organizations to mitigate this issue, given the cost of inaction. The fundamental investment organizations must make is in data governance (technology plus organizational changes).

This is the foundation upon which organizations might be able to leverage this technology. For many, data-driven decisions are still a mirage, let alone AI. For the top performers, aiming for agentic workflows, these initiatives will only be possible not only with high data governance maturity but with flexible and agile human-in-the-loop workflows, given the supervision necessity of these systems.

Jen Goldsmith, President at Tendo
Healthcare organizations face significant challenges when integrating AI into their existing IT landscape, particularly when dealing with legacy systems and accessing critical data for consumption by AI tools and platforms. Legacy systems, often built on outdated frameworks, may lack the compatibility and scalability needed to support modern AI tools. Additionally, siloed data in systems without direct data access or an ability to readily share data can hinder AI’s ability to generate meaningful insights.

To mitigate these challenges, organizations can adopt strategies like investing in interoperability solutions, such as APIs or middleware, to bridge the gap between legacy systems and AI platforms. Standardizing data formats and centralizing data storage can improve accessibility and ensure clean, usable data for AI analysis. Gradual modernization of IT infrastructure, alongside collaboration with vendors who specialize in healthcare AI, can further ease the transition while minimizing disruptions. By addressing these barriers proactively, healthcare organizations can successfully integrate AI and unlock its full potential.

Saji Rajasekharan, Chief Technology Officer at Experity
Healthcare organizations face several challenges when integrating AI into their current IT infrastructure, including data compatibility, system scalability, and ensuring patient privacy. Existing systems often struggle to integrate with modern AI solutions, creating challenges in data integration and real-time processing.

To mitigate these challenges, organizations should focus on selecting AI platforms that integrate seamlessly, ensuring strong data governance, and fostering collaboration between IT and clinical teams. By adopting a phased implementation strategy, healthcare organizations can leverage AI to enhance operational efficiency while maintaining compliance and patient trust throughout care delivery.

Suvajit Gupta, Chief Technology Officer at Cotiviti, Inc.
While AI is becoming an increasingly commonplace technology for healthcare organizations, it can be challenging to implement and requires specific resources that payers may not have access to. Healthcare organizations often run into issues such as difficulty creating appropriate datasets; challenges implementing AI-driven insights into legacy systems; not having the foundational investments into machine learning operations and workflow management; and lacking the ongoing resources to continuously establish, monitor, and update appropriate governance mechanisms.

To overcome these challenges, health plans should make the appropriate long-term investments or bring in a trusted partner that can effectively help them deploy AI within their organization—without replacing human expertise and decision-making.

Keavy Murphy, Vice President of Security at Net Health
Some leaders are hesitant to integrate AI programs into existing enterprise tech stacks, due to concerns about risk to data and shadow IT. To address this, leadership should provide clear guidelines or best practices to help staff, especially those with limited AI experience, on the most secure and compliant ways to utilize these tools. A strategic, incremental approach to adoption can ease internal hesitancy while allowing teams to gradually adapt to new programs. This method also gives advanced engineering teams the opportunity to explore new technologies to drive innovation and enhance the organization’s external services.

Hugh Cassidy, Chief Data Scientist at LeanTaaS
As healthcare organizations continue to adopt new AI tools and improve upon those already within their tech stack, striking the right balance between innovation and trust is essential. Leadership must clearly define the purpose, benefits, and limitations of AI applications for both staff and patients – explaining how AI is used to support decision-making, not replace it, to proactively alleviate concerns about automation overshadowing human expertise and demystify AI’s role in healthcare.

Additionally, implementing human-centric AI design processes is a critical piece of successful AI integration. Many healthcare organizations are already leveraging AI to streamline workflows, reduce administrative burdens on staff, and optimize manual tasks like scheduling. By engaging end-users in the design and feedback process, organizations can expand on those benefits, creating more intuitive interfaces and workflows that integrate seamlessly into existing processes, ultimately enhancing user comfort and minimizing disruption.

Lastly, robust monitoring systems are a must. Regular audits, feedback loops, and updates to address issues or improve accuracy are crucial for maintaining trust and demonstrating a commitment to ongoing improvement, as well as patient and staff well-being.

Anup Panthaloor, Executive Vice President of Health Plans and Healthcare Services at Firstsource
Ensuring AI applications have access to high-quality data is critical, and data in healthcare organizations often is in silos, sometimes in older, closed, proprietary IT systems. Application programming interfaces (APIs) can extract data that AI tools need, including cloud-based AI applications. Delivered as a service, these can enable healthcare organizations to access sophisticated AI tools even when IT landscapes have many legacy systems.

The best approach is to start small, tackling a key pain point with autonomous AI agents and co-pilots, such as automating prior authorization requests. Then build on these capabilities to extend AI deeper into a process, such as by automating supporting documentation retrieval and submission.

Matt Flath, Vice President of Asset Management at Onyx Equities
The adoption of AI in healthcare is picking up speed, and the industry needs to ensure that ample computing power is available to sustain this growth. Between this increased AI utilization and recent U.S. regulations aimed at promoting domestic AI development, all AI-using industries, especially healthcare, need to be mindful of the current capacity of data centers within the U.S. Growth in this sector. Healthcare and biopharmaceutical organizations will not only be competing against themselves for this valuable AI-data processing but also with other tech-driven industries.

Shaji Nair, Founder and CEO at Friska AI
Data silos are one of the most pressing challenges confronting healthcare organizations as they integrate AI more deeply into their existing infrastructure. For AI to function effectively, information must be centralized—a difficult task given that many healthcare organizations still operate in environments where patient, research, and treatment data are stored in separate systems, departments, or servers. Without centralized data, critical information cannot synchronize, which ultimately limits the effectiveness of AI by limiting its ability to learn.

Another challenge is migrating large datasets from older systems to newer AI platforms, which can create issues such as heightened cybersecurity risks. Integrating AI platforms with large systems like EHRs can also expose vulnerabilities and lead to compliance issues with regulations such as HIPAA. Data standardization is another factor that’s critical for successful AI implementation. AI relies on large, standardized data sets to deliver accurate predictions and diagnoses, but legacy systems often lack proper data governance and standardization.

Addressing this challenge requires platforms capable of managing a variety of data sources, as well as standardized information that can be easily accessed by the right individuals when needed. Standardization also offers numerous benefits, including improved data accuracy, which can enhance AI’s reliability and its ability to analyze large datasets more effectively.

Additional challenges include healthcare employees’ ability to adapt to new AI systems. Furthermore, implementing AI demands significant computing power, which may lead to additional investments or expenses for the organization to evaluate.

What great answers! Huge thank you to George Dealy, VP of Healthcare Applications at Dimensional Insight, André Castro, Manager, Secure AI Products at Protegrity, Jen Goldsmith, President at Tendo, Saji Rajasekharan, Chief Technology Officer at Experity, Suvajit Gupta, Chief Technology Officer at Cotiviti, Inc., Keavy Murphy, Vice President of Security at Net Health, Hugh Cassidy, Chief Data Scientist at LeanTaaS, Anup Panthaloor, Executive Vice President of Health Plans and Healthcare Services at Firstsource, Matt Flath, Vice President of Asset Management at Onyx Equities, and Shaji Nair, Founder and CEO at Friska AI for taking time out of your day to submit a quote to us! And thank you to all of you for taking the time out of your day to read this article! We could not do this without all of your support.

What challenges do you think healthcare organizations face when integrating AI into their existing IT infrastructures? How do you think these challenges can be mitigated? Let us know over on social media, we’d love to hear from all of you!