How healthcare AI will shape the future
How healthcare AI will shape the future
Get ready for the future of healthcare. AI promises personalized medicine, proactive care, and less-tired health professionals.
Key Takeaways
AI is revolutionizing healthcare. From personalized medicine to streamlining paperwork, AI is no longer just a buzzword, but a powerful tool with real-world impact.
Patient-centric benefits are key. AI improves outcomes, enhances the patient experience, and could even bridge healthcare access gaps in underserved regions.
Ethical considerations are paramount. Proactive steps around data privacy, bias mitigation, and transparent AI are essential for building trust and ensuring long-term success.
ROI is complex but achievable. The returns on AI investment in healthcare lie in cost reductions, improved efficiency, and a better reputation across the board.
The skills gap is a hurdle. Upskilling clinicians, administrators, and data teams is vital for hospitals to successfully harness AI's full potential.
Last year, I started Multimodal, a Generative AI company that helps organizations automate complex, knowledge-based workflows using AI Agents. Check it out here.
Recently, I’ve been down a rabbit hole about the use of generative AI in healthcare, and the companies excelling in this space. I’ve spoken to doctors, pharmacists, healthcare entrepreneurs, and other industry leaders.
Mark Michalski, CEO of Ascertain put it beautifully when I interviewed him on my podcast:
“Healthcare cannot continue to operate the way it has over the last few decades…healthcare needs to incorporate [AI] to continue to serve the population that it was meant to serve.”
And it’s true. Healthcare has to catch up with AI and integrate it fully to provide better health outcomes and streamline clinical practice. But because it's a highly regulated, data-heavy, and patient-facing industry, it can’t hurry either.
Here, we’ll explore what the future of AI in healthcare looks like, what current AI technologies can do for hospitals, pharma companies, and patients, and what industry professionals should be on the lookout for.
The early days of rule-based systems
Think of AI's beginnings in healthcare like the first clunky computers – pioneers, yes, but a far cry from the sleek machines we know now. In the 1970s and 80s, rule-based systems were the first wave of AI in medicine. These systems followed rigid, pre-programmed rules to aid in tasks like diagnosis. Here’s an example:
MYCIN: Developed in the 1970s, MYCIN helped identify bacterial infections and recommended antibiotic treatments. It was a knowledge-based system, encoding medical expertise in a set of rules. While revolutionary at the time, its reliance on hand-coded rules limited its flexibility and applicability to new situations.
Why rule-based systems mattered
Despite their limitations, rule-based systems laid the groundwork for what was to come. They proved the potential of computers to assist in complex medical decision-making, sparking the imagination of researchers and paving the way for more sophisticated AI techniques. They also highlighted the challenges of capturing human expertise in a rigid set of rules.
The age of machine learning
If rule-based systems were about following instructions, machine learning was about computers learning from experience. This shift from programming to pattern recognition unlocked a whole new level of AI potential for healthcare.
The late 1980s and 1990s saw a surge in machine learning applications in medicine. Machine learning algorithms could sift through vast datasets of patient records, lab results, and medical images, uncovering hidden patterns that human experts might miss. This newfound ability to analyze vast amounts of healthcare data led to breakthroughs in several areas:
Disease diagnosis: Machine learning models were developed to analyze medical images like X-rays and mammograms, identifying signs of disease with higher accuracy and earlier than human radiologists.
Drug discovery: AI algorithms could analyze molecular structures and patient data to accelerate drug discovery and development, leading to the creation of more targeted therapies.
Predictive analytics: I discussed predictive AI and its applications in my last post. It was super relevant for healthcare too, even in its early days. Machine learning models were used to forecast patient outcomes, risk of disease, and even optimal drug dosages. This proactive approach promised to revolutionize preventative care and treatment personalization.
How early models led to modern healthcare AI
The early era of AI in healthcare laid a crucial foundation for the field's future growth. Rule-based systems proved the concept of AI-assisted medical decision-making, while machine learning opened the door to data-driven discovery and personalized medicine.
These advancements, alongside the growing computational power of generative AI systems and the availability of healthcare data, paved the way for the explosion of deep learning and other advanced AI techniques that are transforming healthcare today.
How have modern AI systems transformed healthcare
The days of clunky, rule-based systems in healthcare are fading fast. AI is now a revolutionizing force, changing the way we deliver care. Here are the areas where AI is leaving the biggest mark:
Personalized medicine
Remember when getting a prescription felt like a gamble? A drug that worked for your neighbor might leave you feeling the same, or worse. AI is changing the game, and companies like Paige.AI and Insilico Medicine are charging ahead.
Your body's secret code: Our bodies contain a wealth of data. It's complex, but AI thrives on this complexity. Powerful algorithms can analyze these massive datasets, finding patterns that reveal the most effective treatment plan tailored just for you.
Oncology: precision strikes: Paige.ai are masters at analyzing cancer pathology images. Their AI can pick up minute details that might mean the difference between an early, curable diagnosis and a missed chance. This empowers doctors with critical information that informs targeted treatment plans unique to each patient's cancer.
Tailor-made medicine: No two patients are alike, and AI can finally honor that. By crunching vast datasets of genetics, medical history, and even lifestyle, it can pinpoint which medications (and dosages) are the best fit. Think of it as precision targeting for pharmaceuticals. In oncology, studies suggest personalized cancer treatments can significantly boost survival rates. Across the board, this means fewer harsh side effects and better outcomes. Insilico Medicine exemplifies this future. They use AI to analyze patient data and simulate how different drugs might work with their individual genetic makeup.
Jayodita Sanghvi, senior director of data science @ Included Health also commented on personalized patient care when I spoke to her on my podcast a few weeks ago:
"AI has the power to enable us to deeply understand each individual, what their clinical needs are, what their demographic needs are if they have any care gaps."
I see her point: patient groups who belong to overlooked sections of society can benefit from such personalized care even more.
Drug development is faster
Traditional laboratory-based drug discovery is like looking for a needle in a haystack, a slow slog with no guarantees. With AI pre-screening, scientists get a shortlist of promising leads. This focused approach accelerates the development of life-saving cures.
AI can analyze mountains of molecular data to identify promising compounds in record time. AI-powered simulations can even predict how a drug might interact within the body – essentially conducting 'virtual clinical trials'. This shaves years (and hefty funds) off that long, risky development process.
Atomwise: AI as a master chemist: Atomwise is a digital chemist with a near-magical understanding of molecular structures. Using machine learning, they analyze vast libraries of molecular structures, predicting which ones might interact with specific disease targets. This dramatically speeds up lead candidate identification – the crucial first step in the quest for new drugs.
I see such companies speeding up drug trials and production dramatically and taking a burden off researchers. As these systems get better, they’ll even be able to aid drug discovery for traditionally overlooked patient groups as well.
Real hospitals, real results
Here are some real-world examples of how hospitals today are using AI to change patient care:
Oncology: Memorial Sloan Kettering Cancer Center employs AI to design personalized radiation therapy plans. This targets the tumor more precisely, minimizing damage to healthy tissue – a huge win for patient health outcomes.
Cardiology: I also love what Mayo clinic did with AI in cardiology. They partnered with a generative AI company to develop an AI tool that can identify subtle signs of heart failure via ECG scans. This aims for proactive intervention, keeping patients healthier and reducing costly ER visits.
Neurology: AI models are showing promise in analyzing brain scans, helping diagnose Alzheimer's at the earliest stages. While there's no cure yet, early detection allows for better disease management and support.
General Medicine: University of Pittsburgh Medical Center uses AI for sepsis prediction. This gives doctors critical hours of warning that a patient's condition could deteriorate, saving lives through early treatment.
Fighting the paperwork monster
What stands out to me when it comes to healthcare, especially in the US, is the sky-high cost vs the moderate patient outcomes. Much of the healthcare cost for patients in the US is factored into expensive medications and machines. But a huge portion goes to document processing, claims, insurance management, pharmacy management, etc.
AI-powered automation can help you reduce the outright price of providing healthcare to your patients. This gives them better care without burning a hole in their pockets. Again, you’ll see this reflected in your reputation and bottom-line improvement.
Doctors didn't go to med school to wrestle with paperwork, but that's the reality for many. Most healthcare companies that we work with at Multimodal have mountains of backlogged paperwork that their staff just doesn’t have the time to get to. And they shouldn’t have to choose between improving health outcomes and sifting through paperwork.
AI is changing this. From automating prior authorizations for insurance to transcribing doctor's notes and electronic health records into structured data, AI handles tasks that drain clinicians' time and energy.
And if you’re not convinced yet, consider this. A McKinsey study suggests that nearly 30% of healthcare tasks are automatable. If you automated them, how many more hours could you dedicate to hands-on patient care?
Elevating patient experience
Frustrated patients stuck on hold, playing phone tag for basic questions? AI chatbots offer a lifeline. They provide instant answers to common queries, schedule appointments, and even send medication reminders.
It's not just about convenience – studies link proactive patient engagement through AI tools with improved health outcomes and greater patient satisfaction scores. This has a direct impact on hospitals' performance under value-based care models.
Provider relationship management
I see provider relationship management as a low-hanging fruit for most healthcare providers because it doesn’t take much to automate this. Building strong provider networks is key to delivering quality care and controlling costs. AI offers tools to optimize this complex process:
Smart plan comparisons: AI can analyze mountains of contract data across multiple plans, making it easier to identify the most cost-effective and comprehensive network options. This simplifies decision-making and negotiation.
Effortless communication: AI-powered platforms can generate personalized welcome letters, routine reports, and even initial claim denial explanations. This frees up staff for addressing complex provider inquiries.
Pinpointing network gaps: By analyzing patient data, demographics, and provider directories, AI can pinpoint areas lacking specialists or where wait times are long. This informs targeted recruitment efforts and proactive network expansion.
Performance insights: AI dashboards can track provider performance metrics (quality scores, claim processing efficiency, etc.). This data is valuable for contract negotiations and identifying providers needing additional support.
Pharma AI is the new norm
Hospitals get all the attention when we talk about healthcare. But pharmacies are actually some of the biggest beneficiaries of what AI has to offer:
Counseling, not just dispensing: As AI streamlines medication dispensing, pharmacists have more time for in-depth patient consults, improving adherence and outcomes.
Personalized regimens: With AI-driven analysis of patient data, pharmacists can flag potential drug interactions or tailor dosages more precisely.
Proactive outreach: AI tools can identify patients at risk of non-adherence to their medication and trigger pharmacist interventions to prevent complications.
Proactive refill reminders: AI analyzes patient data and prescription patterns, sending automated reminders (text, email, app) well before the medication runs out. This is especially helpful for chronic conditions.
Smarter authorization handling: AI can flag potential insurance snags with refills early, giving the time to sort it out instead of a last-minute scramble.
One-click refills: AI-powered chatbots or apps can handle simple refill requests instantly, freeing up pharmacists for more complex questions.
The world of pharmacy benefits managers (PBMs), insurance formularies, and drug pricing agreements is notoriously complex and paperwork-heavy. AI offers a way to cut through this red tape of the healthcare industry:
Decoding complex contracts: AI-powered natural language processing tools can analyze dense PBM contracts, extracting key terms, identifying potential cost discrepancies, and flagging ambiguous clauses. This saves time and prevents costly oversights.
Formulary optimization: AI can match patient data with insurance formularies and available drug options, suggesting the most cost-effective and clinically appropriate medication with minimal manual searching.
Automated prior authorizations: Instead of time-consuming form filling, AI can pre-populate prior authorization requests based on patient data and insurer rules, dramatically speeding up approvals.
For example, we recently helped a digital healthcare company automate their PBM tasks. Pre-AI, their analysts spent loads of time extracting essential pricing, discount, and rebate information, leading to decreased productivity and increased operational costs. Our solution helped them increase accuracy, reduce extraction time (from 10 to just 2 minutes per contract), and implement a high-tech system across the board.
If this isn’t a bottom-line improvement, I don’t know what is.
AI's 'Superpower Sight' into medical imaging
Radiologists are heroes of healthcare, but they're only human. Staring at scans for hours is exhausting for these healthcare professionals. AI isn't a replacement, but it's a tireless ally, and its vision is getting sharper by the day.
Beyond human limits: AI-powered radiology tools can detect subtle abnormalities in X-rays, CT scans, and MRIs that might escape even the most experienced radiologists. This is particularly revolutionary for early detection of diseases like cancer, where minutes saved mean lives saved.
The radiologist's assistant: AI isn't taking over radiologists' jobs, but it's becoming an indispensable tool. For instance, Lunit's AI can analyze chest X-rays for multiple diseases simultaneously, highlighting potential issues that warrant a doctor's closer look. This frees up radiologists' time for those complex cases and crucial conversations with patients.
These AI-powered imaging tools aren't just a showcase of cool tech. They're delivering tangible improvements for patients. We're seeing earlier disease detection, fewer false positives (those scary situations where additional tests are needlessly ordered), and the ability for doctors to prioritize urgent cases more effectively.
What does the future of healthcare AI hold?
Think back to the healthcare of today. Reactive, focused on fixing problems when they arise, often generalized – that could feel like healthcare from a bygone era by 2050. AI's potential is to turn the system on its head: proactive, preventative, and hyper-personalized.
Predicting problems before they start
Imagine a world where your health is monitored constantly, not through invasive devices, but through seamless tech built into your environment. Smartwatches that track far more than steps, and sensors in your home... all feed data to powerful AI algorithms.
Your body's digital twin: This AI isn't just tracking trends, it's building a digital twin of your health. Using predictive modeling informed by massive medical datasets, it can spot subtle deviations from your baseline that could signal a developing issue – heart trouble, cancer risk, even early signs of cognitive decline – often before any symptoms arise.
Proactive, not reactive: Instead of waiting until you're sick, the AI alerts your doctor or even an AI-powered virtual care system. Intervention happens earlier – a simple diet change, a medication adjustment, or preemptive treatment to stop an illness in its tracks. This is healthcare aimed at preserving your quality of life, not just frantic damage control.
The rise of AI-driven virtual care
Healthcare providers will always be invaluable, but AI could lighten their load and expand their reach globally. We're already seeing the early stages of this:
Your AI medical assistant: By 2050, these could become sophisticated virtual companions, collecting your daily health data, flagging concerns, and seamlessly connecting you with a human doctor when needed. This streamlines routine appointments, freeing doctors for the most complex cases.
Mental health at scale: The global shortage of mental health professionals is a crisis. AI chatbots are already offering compassionate, 24/7 support and therapy, guided by psychology experts. These tools could become vital lifelines for millions who otherwise would lack access to care.
The AI "Check-Up": Imagine an annual health checkup conducted largely by AI – analysis of bloodwork, full-body scans, and risk assessment based on your unique data. The doctor then steps in for a consultation, and their time is used for the most crucial part – explanations, empathy, and shared decision-making.
Making AI a bridge to global healthcare access
I have read countless cases of people in rural areas receiving inadequate treatment, often to the point of losing lives. This is especially true for the developing world. But one of the most powerful promises of AI in the healthcare sector is its potential to close gaps in care worldwide. Many regions lack specialists or even basic medical facilities. AI can change this.
Remote expertise, delivered locally: Doctors in rural areas using AI-powered imaging tools that deliver specialist-level analysis of scans in real-time could be the answer to healthcare access issues. Decision support systems could guide less-experienced clinicians through complex diagnoses and treatment choices.
Bridging the knowledge gap: AI can put the world's medical knowledge into the hands of healthcare workers anywhere. Language translation tools could ensure research is accessible in any language, while training platforms might use AI to personalize learning for clinicians at vastly different levels of expertise.
Telemedicine, supercharged: Today's telemedicine often feels limited compared to in-person care. AI could transform this, with AI-assisted tools analyzing a patient's voice patterns, facial expressions, and vital signs remotely – adding a layer of nuanced data to ensure remote consultations are thorough and accurate.
What’s the ROI of healthcare AI?
Let’s be real: it’s not easy to flesh out money for AI both because it's expensive and because you often can’t calculate its real ROI in an industry like healthcare. I did talk about this in detail in the AI ROI post earlier too. AI ROI comes from reducing costs and increasing efficiency. Some combination of these will give you a dry number to work with.
Sure, reducing costs is a major driver, but true healthcare ROI goes deeper. Let’s revisit the digital healthcare company I talked about earlier. Their complex pharmacy benefit contracts were a hotbed of inefficiencies and overspending. Our AI solution analyzed these contracts, identifying discrepancies and potential savings. The results?
80% reduction in contract review time
Millions of dollars in annual cost savings uncovered
Frees up in-house experts for strategic work, not tedious contract analysis
This illustrates how AI-powered gains lead to:
Lower long-term costs: Proactive, accurate contract management minimizes costly reimbursement errors.
The value of clinician time: Think of how this impacts pharmacists, who can focus on patient counseling instead of being buried in paperwork.
Reputation is priceless: Becoming known as a tech-forward, efficient healthcare provider has long-term competitive advantages.
But the costs matter
Healthcare AI ROI can be super tempting. But your budget and implementation matter a lot.
Upfront investment: Significant AI implementations require thought and investment. Budgeting for change management and staff training is crucial for a smooth rollout.
It's not magic: AI amplifies existing strengths. It won't solve fundamental problems with poor processes or low staff engagement.
Implementing a complex technology is always going to be a tradeoff. When you’re trying to decide whether or not to go ahead with this investment, think about your goals in regard to your patients, and how far AI will take your organization if you implement it. If you’re still confused, I’d suggest taking baby steps by rolling it out for smaller, affordable applications and iterating from there.
There are ethical concerns too
One of the key questions I encounter as I work with healthcare providers is about data. Most hospitals and health insurance providers have mountains of medical records and customer data. How do you get it in a secure, trustworthy, and explainable system and make sure it integrates seamlessly with the AI? How do you tune out the data noise and make it worthy of automation? Can you still comply with the regulations as you feed your data to AI?
Patient privacy: the bedrock of trust
Health data is undeniably sensitive. A breach doesn't just hurt the patient – it erodes the public's trust in the entire system. While regulations like HIPAA exist, AI demands extra vigilance:
De-identification isn't foolproof: Anonymized data can sometimes be re-identified with sophisticated techniques. Healthcare organizations need robust processes that go beyond simple name removal.
Consent matters: Patients should have a clear say in how their data is used, especially when it feeds AI development. This needs user-friendly, non-legalistic opt-in forms.
Cybersecurity is paramount: AI systems are only as secure as their infrastructure. Regular audits, staff training, and proactive threat detection are non-negotiable.
For these reasons, it’s critical to choose an AI partner that makes sure your data doesn’t leave your system. For instance, if you feed your patients’ data into ChatGPT or other excellent LLMs, you’ll get great results. But you’ll also make it a training tool that goes into a system that rids it of privacy.
Algorithmic bias: when 'Smart' systems discriminate
AI might seem objective, but it learns from the data we feed it. If that data reflects existing healthcare disparities (and it often does), the bias gets baked into the system. The consequences are serious:
Inaccurate diagnoses: An AI tool trained mostly on white patients might miss signs of diseases that present differently in people of color.
Inequitable treatment: If an AI system flags high-risk patients for intervention, but factors like zip code or insurance skew its results, vulnerable populations could fall through the cracks.
Fighting bias isn't an afterthought: Mitigation strategies must be in place from day one. Diverse datasets, rigorous testing, and having 'explainability' built into models (so we know why they make decisions) are key.
The 'Black Box' problem: can we trust AI we don't understand?
Some AI systems, especially deep neural networks, are so complex that even their creators can't fully grasp how they reach conclusions. This is a major hurdle for healthcare:
Explainability vs. accuracy: Highly accurate models might be opaque, making doctors hesitant to rely on them, especially for high-stakes decisions.
Regulatory hurdles: The FDA will want to see how an AI system works before wide approval. A lack of transparency makes this much harder. Researchers are actively developing 'explainable AI' techniques to bridge the gap.
Gaining patient buy-in: If a patient is told "The AI recommends surgery, but we don't fully know why," that's a tough sell. Communication strategies will be essential for acceptance.
The evolving regulatory landscape: playing catch-up
AI technology moves faster than the law. Here's what healthcare leaders need to watch:
FDA's evolving stance: The FDA is actively developing an AI-specific regulatory framework. This will likely bring stricter approval processes and ongoing monitoring requirements.
Liability questions: If an AI system makes a harmful error, who's responsible? The doctor, the hospital, the software vendor? Legal precedents are still being set.
Cross-border confusion: With patient data often flowing internationally, complying with multiple, sometimes conflicting, privacy laws will become a major headache.
Don't wait for scandals to force action. Hospitals that thoughtfully address these issues now gain a long-term advantage:
Ethics committees: Include AI experts alongside clinicians and legal reps to analyze potential use cases and proactively address risks.
Bias audits: Invest in regular testing of AI systems to uncover any hidden biases and mitigate them early.
Be an active voice: Industry leaders should advocate for sensible regulations that foster innovation while protecting patients.
The healthcare AI skills gap
The promise of AI in healthcare stumbles into a harsh reality: most hospitals simply aren't ready to harness it. This skills gap spans multiple levels and demands urgent action.
Doctors, nurses, and allied healthcare professionals are on the front lines, yet few have the training to understand what AI can (and can't) do.
Dr. Harvey Castro, another one of my guests on Pioneers, explained how in healthcare, AI implementation isn’t a top-down approach. Doctors and nurses are the ones implementing and using these systems. They’re the ones who need to test if the system works, how effective it is, and how convenient would it be for them. At 3 AM when they’re already overworked, the last thing they want is to jump through the loops of complicated AI models just to get to a diagnosis or treatment plan. The truth is, we don't need doctors who can write code, but we do need:
Understanding capabilities: What kinds of tasks is AI good at? Where does it fall short? This knowledge base guides smart deployment.
Critical thinking about AI outputs: Interpreting AI results requires clinical judgment. "The AI says X, but given the patient's history, does that make sense?"
AI as a communication tool: Can clinicians explain AI-driven recommendations to patients in plain language? This is key to building trust.
It's not just about tech teams
Massive training initiatives are crucial, but too often, these focus narrowly on IT staff. To drive true AI transformation, we need upskilling of:
Administrators: They make budget and staffing decisions that impact AI success. Understanding its ROI, limitations, and workforce implications is crucial.
Clinicians: From targeted workshops on AI for specific specialties to integrating AI concepts into medical school curricula, this is a long-term project.
Data wranglers: Clean, usable data is AI's fuel. Investing in data governance roles, not just data scientists is often overlooked.
The cost of inaction
Failing to bridge this skills gap has very real consequences:
Wasted investment: Expensive AI systems sit unused, failing to deliver ROI.
Poor patient outcomes: Misused or misunderstood AI can lead to missed diagnoses or inappropriate treatments.
Clinician burnout exacerbated: Poorly designed AI workflows add frustration, not efficiency, pushing staff away.
I also host an AI podcast and content series called “Pioneers.” This series takes you on an enthralling journey into the minds of AI visionaries, founders, and CEOs who are at the forefront of innovation through AI in their organizations.
To learn more, please visit Pioneers on Beehiiv.
Building responsible healthcare AI with the right partner
Great, we’ve established what AI can do across the board for healthcare professionals and patients. It’s also evident how using AI can reduce costs for organizations, and prices for patients. But I want to emphasize one thing here: choose the right AI partner if you’re a health professional looking to adopt AI.
The last thing you want is to end up with a system that makes you run into legal trouble, or worse, bad patient outcomes. The most famous AI companies here won’t be the ones making the cut if you’re trying to be thoughtful and better.
Look for a partner that serves your industry or similar ones. Get your data in order, but make sure it isn’t compromised. Moreover, demand a system that you can understand and explain if it ever comes to that.
In the next few weeks, we’ll dive further into the potential of artificial intelligence in healthcare, how doctors and other medical professionals can lean into AI to ease their lives, and how it can boost the ROI of healthcare organizations.
Stay tuned,
Ankur