In the early 2020s, artificial intelligence was like a shiny new toy. LLMs started writing poetry, generating surreal images and transforming electronic health record (EHR) systems. Back then, the conversation was dominated by future-tense verbs: AI will change medicine; AI could accelerate research.
Fast forward to 2026, and AI has spread to other facets of daily life. In the biopharma sector, the “hype cycle” has settled into a rhythm of pragmatic application. Still, conversations on Sermo indicate that physicians are unsure how far the evolution will go. “I think that nobody can predict how fast [AI] is going to evolve, it can be exponential or reach a [plateau],” writes one general practitioner in the community.
A pediatric neurologist has a similar take: “Everything remains to be decided, and it will clarify what the role of humans will be in the face of the advances that AI will bring.”
While the end result remains unpredictable, the direction is clear. AI is fundamentally shifting how therapies are conceived, tested and delivered. Here is how AI in drug development is reshaping the clinical landscape in 2026.
Accelerating the path to target identification
For decades, target identification (finding the specific molecules that a drug targets during drug development) was a game of educated guessing, often plagued by high failure rates.
The emergence of AI for drug discovery may speed up the process. By shifting to computational target selection before wet-lab validation, researchers are identifying viable targets with greater efficiency. This allows them to fail candidates in silico (on the computer) rather than in vivo (in animal models or human testing). That said, AI has its limitations. Models are often trained on incomplete or poor quality data, reducing their accuracy, according to a 2025 study. There’s also the possibility of training with biased data, resulting in biased AI systems, the study authors point out.
Physicians on Sermo are watching this shift. When polled on whether AI will meaningfully accelerate drug development in the next five to 10 years, 29% predicted significant acceleration, 37% said moderate improvements and only 3% said timelines won’t change meaningfully. When the community shared where they believe AI has the strongest potential impact, the top answers were early-stage target identification (28%), predicting drug-target interactions (28%) and designing clinical trials (15%).
As of February 2026, there are no current FDA-approved AI-designed drugs, but here’s a breakdown of how these advancements are transforming future drug development:
Early-stage target identification
Algorithms can now scan genomic datasets, proteomic information and scientific literature simultaneously to pinpoint causal links between biological targets and disease states.
An otolaryngologist on Sermo is encouraged by this application. “I think AI is extremely promising for drug development,” they share. “It can analyze huge datasets, identify potential targets faster, and reduce the time and cost of early-stage discovery. While it won’t replace human expertise, it can significantly accelerate research, improve accuracy, and open the door to treatments that would’ve been impossible to find with traditional methods.”
An ophthalmologist and Sermo member believes AI “will make it possible to rapidly accelerate the testing of promising molecules while eliminating others much sooner in the process.”
Predicting drug–target interactions
Once a target is identified, the next hurdle is binding affinity. Will the molecule actually stick? Historically, this required exhaustive physical testing. Deep learning models could help predict these interactions.
Designing clinical trials
Poor endpoints or unrealistic exclusion criteria can doom a study during the protocol design phase, before any subjects are enrolled. AI tools are now simulating trial protocols against historical data to predict potential bottlenecks. “AI is highly promising: it accelerates target discovery, predicts molecule behavior, and reduces early-stage failure rates,” writes a dermatologist on Sermo.
Some Sermo members remain cautiously optimistic. “AI is shortening timelines and sharpening the search for new molecules and targets, and it’s beginning to help design better trials and select patients,” notes a radiation oncologist. “But we still need much more independent validation and transparency before fully trusting its results in clinical practice.”
Patient recruitment and stratification
Finding the right patient for the right trial can be a logistical nightmare. AI is refining patient stratification by analyzing electronic health records to find eligible candidates who might otherwise be overlooked.
A general practitioner on Sermo thinks this could be one of AI’s strengths. “AI holds immense potential for drug discovery, accelerating the process and refining patient stratification, provided it’s robustly validated and integrated thoughtfully into practice,” they write.
Predictive safety: reducing clinical attrition while using AI in drug development
Sometimes during drug development, a candidate shows efficacy but fails on safety—often late in the game. Predictive safety models may help to mitigate this risk.
Computational analysis can help flag potential cardiotoxicity and metabolic issues before a drug enters a human body. By generating activity profiles that map how a molecule might interact with off-target receptors (the ones you don’t want it to touch), researchers can ensure that only high-quality candidates move forward.
Sermo members are split as to how much they trust AI-generated predictions during early research. 13% of participants in a poll said they’re “very confident,” 36% are “somewhat confident,” 31% are neutral and 17% are not confident.
Optimizing clinical trials with digital twins
A digital twin is essentially a virtual representation of a patient—or a control arm—built from vast amounts of health data. It allows researchers to simulate how a patient with specific characteristics might respond to a treatment. This has implications for the long-term goal of reducing control group sizes or supplementing traditional placebo arms in certain contexts. It has the potential to help reduce wasted capital and time by predicting outcomes before the deployment of real-world resources.
An OBGYN on Sermo considers this a critical tool for the future. “AI is going to be able to accelerate drug development,” they write. “Digital twins are going to be extremely useful.”
For a deeper dive into this specific technology, read Sermo’s full article on digital twins in healthcare.
Building trust through ‘Glass Box’ transparency
Despite the excitement, AI and drug development have a lingering relationship issue: the “black box” problem.
Initially, deep learning algorithms would spit out a correct answer (e.g., “This molecule will bind to this protein”), but they couldn’t explain how they arrived at that conclusion. In medicine, this isn’t good enough.
The goal is to transition to “glass box” AI—systems that offer traceable reasoning. Interpretability is especially important for high-risk AI medical applications. If an AI suggests a treatment pathway, a human physician needs to be able to audit the logic.
Physicians on Sermo are concerned with the possibility that providers could rely on AI data without thinking critically. In response to a poll asking what concerns members most about AI integration into drug development, the most popular answer was “over-reliance on technology by research teams,” with 26% of the vote. The factor that would increase their trust the most is independent validation of algorithms (25%). Evidently, many members are wary of the possibility of placing blind trust in the responses of black box models.
The transition from black box to glass box is still ongoing, from one general practitioner Sermo member’s perspective. “…many AI models still function as ‘black boxes’, and without transparent methodologies or robust prospective data, it’s difficult to gauge how well these tools generalize beyond curated datasets,” they reason.
In the U.S., AI tools used in drug development are not regulated as standalone products. Instead, the Food and Drug Administration (FDA) evaluates how AI-generated insights are used within the drug development process—particularly when they influence clinical trial design, safety assessments or regulatory submissions. While regulators have signaled openness to AI-assisted approaches, standards for validating adaptive or continuously learning models are still evolving.
AI’s role moving forward
In drug development, AI has the potential to replace brute-force screening with rational, predictive design. It could help usher in true precision medicine—delivering the right drug to the right patient, at the right time, with less waste. Physicians on Sermo are engaged in discussions around AI’s potential advantages and downsides in drug development and beyond. Join the community to access (and contribute to) the world’s largest drug ratings database, debate the clinical evidence of the first wave of AI-designed therapies and remain at the forefront of new advances.
