Discover how Simreka copilots improve formulation safety and performance prediction.
New product development (NPD) in materials science and chemical formulation has always been a high-stakes endeavor. Every experimental iteration consumes valuable resources, time, and capital—and carries the risk of failure, regulatory non-compliance, or safety issues that can derail entire projects. According to McKinsey’s State of AI 2025 report, 88 percent of organizations now report regular AI use in at least one business function, up from 78 percent just a year ago, driven largely by the technology’s ability to reduce risk and accelerate innovation.
The emergence of AI-powered experiment design represents a fundamental shift in how organizations approach NPD. Rather than relying solely on intuition, historical precedent, and trial-and-error testing, R&D teams can now leverage intelligent systems that predict formulation performance, anticipate safety concerns, and identify optimal experimental pathways before committing resources to physical testing. This paradigm shift is not just about efficiency—it’s about transforming risk from an unavoidable cost of innovation into a manageable, predictable factor that can be systematically minimized.
The High Cost of NPD Risk
Traditional new product development workflows expose organizations to multiple risk categories, each with significant consequences:
- Technical risk: Will the formulation deliver the desired performance characteristics?
- Safety risk: Are there toxicity, flammability, or environmental hazards?
- Regulatory risk: Will the product meet compliance requirements across target markets?
- Financial risk: What is the probability of ROI given development costs and market uncertainty?
- Timeline risk: Will delays enable competitors to capture market share first?
The stakes are substantial. Research shows that despite promising benefits like reduced development times and increased innovation, AI adoption in NPD is slow, with only 18% of U.S. firms having adopted AI by early 2024—largely due to concerns about implementation complexity and risk management itself.
In the chemical industry specifically, the global market for artificial intelligence was valued at US$1.3 billion in 2024 and is projected to reach US$5.2 billion by 2030, reflecting growing recognition that AI capabilities can simultaneously reduce risks while increasing R&D success rates.
How AI Transforms Experiment Design
AI-powered experiment design fundamentally changes the NPD risk profile by introducing predictive intelligence at every stage of development. Rather than discovering problems after expensive failures, organizations can now anticipate and mitigate risks before they materialize:
| NPD Stage | Traditional Risk Exposure | AI-Powered Risk Mitigation |
|---|---|---|
| Initial Formulation | Uncertain performance; multiple iterations needed | Predictive modeling identifies promising candidates upfront |
| Safety Assessment | Discover hazards only after synthesis | Toxicity and environmental impact predicted in silico |
| Regulatory Compliance | Late-stage discovery of non-compliant ingredients | Automated compliance checking against global databases |
| Performance Validation | Physical testing of many suboptimal candidates | Virtual screening narrows to high-probability successes |
| Scale-up | Unexpected manufacturing challenges | Process simulation identifies issues before production |
Simreka’s Virtual Experiment Platform exemplifies this risk-reduction approach through its forward and reverse simulation capabilities. Forward simulation predicts outcomes based on proposed formulations, while reverse simulation works backward from desired performance specifications to identify optimal ingredient combinations—dramatically reducing the experimental search space and associated risks.
Predictive Safety Assessment: From Reactive to Proactive
Safety risks represent some of the most consequential failures in NPD. A formulation that reaches late-stage development only to reveal toxicity issues can cost millions in wasted investment and cause serious reputational damage. AI transforms safety assessment from a reactive checkpoint into a proactive design constraint.
Modern AI-powered formulation tools can now anticipate environmental and health risks before synthesis. According to research on formulation machine learning tools, toxicity and CO₂ footprint prediction capabilities can anticipate environmental and health risks, with platforms providing explainable suggestions and reasoning behind each recommendation.
These systems analyze historical safety data, molecular structures, and regulatory databases to identify potential hazards across multiple dimensions:
- Human toxicity (acute and chronic exposure)
- Environmental impact (aquatic toxicity, biodegradability)
- Physical hazards (flammability, reactivity)
- Regulatory restrictions (ingredients banned in target markets)
- Carbon footprint and sustainability metrics
Simreka’s MatIQ – the AI Co-Pilot for Material Innovation integrates safety assessment directly into the formulation workflow through its MatQuest module, which accesses a massive corpus of patents, scientific literature, technical datasheets, and safety databases. This enables researchers to query potential safety concerns in natural language and receive evidence-based assessments before committing to physical experimentation.
Performance Prediction: Reducing Technical Risk
Technical risk—the uncertainty about whether a formulation will deliver desired performance characteristics—historically required extensive physical testing to resolve. AI-powered performance prediction shifts this paradigm by enabling virtual validation before laboratory work begins.
Recent advances in formulation AI have achieved remarkable accuracy. According to research on AI-directed formulation strategy design, systems have achieved an average area under the receiver operating characteristic curve above 90% for multiple decision tasks—demonstrating that AI predictions now match or exceed human expert judgment in many scenarios.
McKinsey research on AI in the chemical industry reports potential for two- to threefold acceleration in materials or molecule discovery with AI, enabling net new patentable chemistries optimized for end-state product properties. The report estimates that generative AI across commercial, R&D, operations, and support functions in energy and materials can create $80 billion to $140 billion in value.
MatIQ’s DataDive module exemplifies practical performance prediction by enabling natural language queries against historical experimental data. Researchers can ask questions like “Which formulations achieved tensile strength above 50 MPa with less than 2% volatile content?” and receive instant analytical insights with supporting visualizations—dramatically accelerating identification of successful patterns.
Regulatory Compliance: Automated Risk Screening
Regulatory compliance represents a critical risk dimension in NPD, particularly for companies operating in multiple geographic markets with differing requirements. Traditional compliance checking relied on manual keyword searches through regulatory databases—a time-consuming, error-prone process that often discovered issues only late in development.
AI-powered compliance tools transform this process by automatically comparing proposed formulations against comprehensive regulatory databases. These systems go beyond simple keyword matching to understand ingredient relationships, concentration thresholds, and context-specific restrictions. Data-driven AI tools can rapidly compare ingredients against safety databases to ensure compliance with regulatory bodies, as noted in industry analyses of formulation machine learning.
The risk implications are significant: discovering a non-compliant ingredient during late-stage testing can require complete reformulation, causing delays measured in months or years. AI-powered early screening eliminates this risk by flagging compliance issues before significant resources are invested.
Simreka’s Databank – the World’s Largest Material Informatics Platform provides the comprehensive ingredient and regulatory data foundation that enables this automated screening. By integrating compliance checking directly into the formulation design workflow, organizations can ensure regulatory viability becomes a design constraint rather than a late-stage discovery.
Intelligent Design of Experiments: Maximizing Information Gain
Even with AI-powered prediction, physical validation remains essential. The question becomes: which experiments should we run to maximize learning while minimizing risk and cost? This is where intelligent design of experiments (DOE) powered by AI delivers substantial value.
Traditional DOE methods relied on statistical techniques developed in the pre-digital era. Modern AI-enhanced DOE uses machine learning to dynamically optimize experimental plans based on accumulating results. According to recent research on AI-driven experimental design, the annual number of publications at the intersection of AI and engineering began to rise sharply around 2017, increasing sevenfold by 2024, reflecting rapid advancement in this field.
AI-enhanced DOE reduces risk through several mechanisms:
- Adaptive planning: Algorithms adjust experimental plans in real-time based on early results
- Exploration-exploitation balance: Systems optimize the trade-off between testing novel formulations and refining promising candidates
- Resource optimization: Experiments are prioritized by information gain per unit cost
- Constraint satisfaction: Safety, budget, and timeline constraints are built into experimental selection
Simreka’s Virtual Experiment Platform incorporates intelligent DOE principles through its data exploration capabilities, which query historical datasets to identify information gaps and suggest experiments that will yield maximum learning. This approach ensures that every physical experiment contributes meaningfully to knowledge accumulation rather than redundantly confirming what models already predict with confidence.
Generative AI for Formulation Innovation
The latest frontier in risk-reduced NPD is generative AI—systems that don’t just evaluate proposed formulations but actively generate novel candidates optimized for multiple objectives. This represents a profound shift from AI as an evaluator to AI as a creative partner in innovation.
Recent breakthroughs demonstrate the potential. According to research published in Nature Communications, generative AI methods can create digital versions of drug products from images of exemplar products, employing image generators guided by critical quality attributes such as particle size and drug loading. The pharmaceutical AI market has grown from US$200 million in 2015 to US$700 million in 2018, and was expected to reach $5 billion by 2024.
Simreka’s AI-Powered Formulation Generator brings this capability to materials and chemical formulation development. The system accepts application requirements, performance targets, and constraints as inputs—including verbal descriptions—and generates AI-suggested formulations that satisfy multiple objectives simultaneously. This approach dramatically reduces the risk of overlooking promising formulation spaces while ensuring that generated candidates meet predefined safety, performance, and regulatory requirements.
Real-Time Risk Monitoring and Mitigation
AI’s role in risk reduction extends beyond initial design into ongoing monitoring throughout the NPD lifecycle. Machine learning algorithms can analyze historical safety data to identify potential risks and predict unsafe conditions, while AI-powered monitoring systems provide real-time alerts for abnormal operating conditions, as documented in research on artificial intelligence for chemical risk assessment.
This continuous risk assessment capability enables organizations to:
- Detect emerging safety patterns across multiple concurrent projects
- Identify unexpected ingredient interactions before they cause failures
- Monitor regulatory landscape changes and flag affected formulations
- Track project risk scores and trigger interventions when thresholds are exceeded
The result is a shift from periodic risk reviews to continuous risk intelligence that enables proactive intervention rather than reactive damage control.
Building an AI-Powered Risk-Reduced NPD Workflow
Successfully implementing AI-powered experiment design requires thoughtful integration of tools, processes, and culture. Organizations seeing the greatest risk reduction benefits follow a structured approach:
- Establish data foundations: Ensure historical experimental data, safety records, and formulation specifications are digitized and accessible
- Define risk metrics: Create quantitative definitions of technical, safety, regulatory, and financial risk
- Integrate AI at design stage: Make predictive tools available during initial formulation ideation, not as afterthought validation
- Create feedback loops: Systematically feed physical experimental results back to refine predictive models
- Train hybrid teams: Develop R&D professionals who understand both domain science and AI tool capabilities
- Establish governance: Define approval workflows for AI-generated formulations and predicted risk assessments
According to McKinsey’s 2025 AI report, companies seeing the most value from AI often set growth or innovation as additional objectives, and half of AI high performers intend to use AI to transform their businesses. However, only 64 percent of respondents say that AI is enabling their innovation, and just 39 percent report EBIT impact at the enterprise level—highlighting that successful implementation requires strategic commitment beyond tool deployment.
Managing AI-Related Risks in NPD
While AI dramatically reduces traditional NPD risks, it introduces new risk categories that organizations must address. The same McKinsey research found that respondents reported acting to manage an average of two AI-related risks back in 2022, compared with four risks today.
Key AI-specific risks to manage include:
- Prediction accuracy: Understanding confidence intervals and model limitations
- Data security: Protecting proprietary formulation data and trade secrets
- Bias and fairness: Ensuring models don’t perpetuate historical biases in material selection
- Intellectual property: Clarifying ownership of AI-generated formulations
- Regulatory acceptance: Ensuring AI-designed products meet regulatory scrutiny
Organizations must balance AI’s risk-reduction benefits against these new considerations, implementing appropriate governance frameworks and validation protocols.
Conclusion
AI-powered experiment design represents one of the most significant advances in new product development risk management in decades. By shifting safety assessment, performance prediction, and regulatory compliance from reactive validation steps to proactive design constraints, AI enables organizations to reduce NPD risk while simultaneously accelerating innovation and reducing costs.
The evidence is compelling: two- to threefold acceleration in materials discovery, prediction accuracy above 90%, and potential value creation of $80-140 billion in energy and materials sectors alone. Yet despite these benefits, adoption remains at just 18% among U.S. firms, indicating substantial competitive opportunity for early movers.
The organizations that will lead the next decade of materials innovation are those that view AI not as a tool for incremental improvement but as a fundamental enabler of risk-intelligent NPD—where every formulation decision is informed by predictive intelligence, every experiment is optimized for maximum learning, and risk evolves from an unavoidable cost into a manageable, quantifiable parameter that can be systematically minimized. The question is not whether to adopt AI-powered experiment design, but how quickly your organization can implement it to capture competitive advantage before your competitors do.
Frequently Asked Questions
Q1. How does AI reduce safety risks in new product development?
AI reduces safety risks by predicting potential hazards before physical synthesis. Machine learning models analyze molecular structures, historical safety data, and regulatory databases to anticipate toxicity, environmental impact, and physical hazards. This enables researchers to identify and eliminate problematic ingredients during the design phase rather than discovering safety issues after expensive late-stage testing. Platforms like Simreka’s MatIQ provide toxicity and CO₂ footprint prediction with explainable reasoning, transforming safety from a reactive checkpoint into a proactive design constraint.
Q2. What level of prediction accuracy can we expect from AI-powered formulation tools?
State-of-the-art AI formulation systems have achieved above 90% accuracy (measured by area under the receiver operating characteristic curve) for multiple decision tasks, according to recent research. However, accuracy varies by application domain, data quality, and specific performance metrics. Organizations should expect high confidence predictions for properties similar to their historical data, with lower confidence for novel formulation spaces. The key is using AI to narrow the experimental search space rather than eliminate physical validation entirely—a workflow built into Simreka’s Virtual Experiment Platform.
Q3. How long does it take to implement AI-powered experiment design?
Implementation timelines vary based on organizational readiness. Companies with well-organized historical data and clear NPD risk metrics can see initial benefits within 3-6 months through pilot projects. Full enterprise integration typically requires 12-18 months and includes data infrastructure development, team training, and process redesign. The fastest path to value is starting with focused use cases (e.g., regulatory compliance screening or safety prediction) before expanding to comprehensive AI-powered workflows. Cloud-based platforms like Simreka’s Virtual Experiment Platform can accelerate deployment by eliminating local infrastructure requirements.
Q4. Does AI-powered experiment design work for small companies with limited historical data?
Yes, though the approach differs. Small companies can leverage pre-trained models and extensive external databases like Simreka’s Databank to compensate for limited internal data. Generative AI tools can suggest formulations based on verbal descriptions alone, without requiring extensive historical datasets. Additionally, intelligent design of experiments helps small companies maximize learning from every physical experiment, building predictive capability rapidly. The key is choosing platforms that combine proprietary enterprise data with comprehensive external material informatics databases.
Q5. What are the biggest barriers to adopting AI in NPD?
Research shows that despite promising benefits, only 18% of U.S. firms had adopted AI in NPD by early 2024. The primary barriers include concerns about implementation complexity, lack of in-house AI expertise, data quality and accessibility issues, unclear ROI expectations, and organizational resistance to changing established R&D workflows. Organizations overcome these barriers by starting with well-scoped pilot projects, partnering with experienced AI platform providers like the Simreka AI-Powered Formulation Generator team, investing in hybrid team training, and establishing clear metrics for risk reduction and efficiency gains.
Q6. How do we manage intellectual property concerns with AI-generated formulations?
IP management for AI-generated formulations requires clear governance frameworks established upfront. Organizations should define ownership protocols for AI-suggested formulations, ensure AI training doesn’t expose proprietary data to competitors, implement appropriate data security measures, and work with IP counsel to understand patentability of AI-assisted inventions. Leading AI platforms offer secure, private deployment options where proprietary data never leaves organizational boundaries—a configuration the team can walk through during a Simreka demo. The key is treating AI formulation tools as internal R&D resources with appropriate confidentiality and ownership protocols, similar to how organizations manage other computer-aided design tools.
Bibliographical Sources
- McKinsey & Company (2025). “The state of AI in 2025: Agents, innovation, and transformation.” Available at: https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai
- ChemCopilot (2024). “Formulation Machine Learning Tools: How AI Is Optimizing Chemical Synthesis and Product Performance.” Available at: https://www.chemcopilot.com/blog/formulation-machine-learning-tools-how-ai-is-optimizing-chemical-synthesis-and-product-performance
- Globe Newswire (2025). “Artificial Intelligence in Chemicals Research Report 2024-2030.” Available at: https://www.globenewswire.com/news-release/2025/02/25/3032214/0/en/Artificial-Intelligence-in-Chemicals-Research-Report-2024-2030-AI-and-IoT-Revolutionize-Chemical-Production-with-Efficiency-Sustainability-and-Smart-Manufacturing.html
- McKinsey & Company (2024). “How AI enables new possibilities in chemicals.” Available at: https://www.mckinsey.com/industries/chemicals/our-insights/how-ai-enables-new-possibilities-in-chemicals
- ScienceDirect (2024). “AI-directed formulation strategy design initiates rational drug development.” Available at: https://www.sciencedirect.com/science/article/abs/pii/S0168365924008988
- MDPI Applied Sciences (2025). “A Helping Hand: A Survey About AI-Driven Experimental Design for Accelerating Scientific Research.” Available at: https://www.mdpi.com/2076-3417/15/9/5208
- Nature Communications (2024). “In silico formulation optimization and particle engineering of pharmaceutical products using a generative artificial intelligence structure synthesis method.” Available at: https://www.nature.com/articles/s41467-024-54011-9
- ResearchGate (2024). “AI and New Product Development.” Available at: https://www.researchgate.net/publication/377460646_AI_and_New_Product_Development
- ScienceDirect (2019). “Artificial Intelligence for chemical risk assessment.” Available at: https://www.sciencedirect.com/science/article/pii/S2468111319300349
