If you, like most of us, feel confused by the avalanche of AI applications and tools, you are at the right place. According to the latest conservative updates, over 10,000 AI applications and platforms exist today. For example, Large Language Models (LLMs) are mushrooming, with one new LLM appearing daily. How can we choose what works for us? In this and the following blogs, we will explain how we at Diplo select the right tools. Firstly, we avoid searching for an optimal tool as this can lead to ‘paralysis by analysis’. We choose a good enough tool and start using it. By using and adjusting it, we discover deeper layers of AI platforms that cannot be seen at first glance or relate more to our specific needs. Here is a concrete case study. Last week, as we heard the news about US tariffs, we wanted to see if AI could help us understand how these tariffs affect specific companies and countries. We chose Novartis, a major pharmaceutical company, as a case study. Using GenAI platforms, including ChatGPT and DeepSeek, did not yield helpful results. The reasoning model features slightly better but still much lower than one can expect to have informed advice on what to do with, for example, Novartis shares. Instead, we turned to the Financial AI Agent developed in the DiploAI Sandbox—a space where the Diplo AI team experiments beyond their daily tasks, driven by curiosity. It outperformed GenAI platforms and a few publicly available AI agents designed for financial analysis. In this text, you can first read the analysis of the impact of US tariffs on Novartis and then explore how Financial AI agents performed this analysis and what the differences are compared to LLMs and reasoning models. From the ‘sandbox’ of the DiploAI team (Petar, Anja, Milos, Jovan, Nikola)
Impact of new US tariffs on Novartis
Novartis confronts significant strategic uncertainty following the announcement on April 2, 2025, of new US tariffs on Swiss goods. While pharmaceuticals are currently exempt from the broad 31-32% tariffs (structured as a 10% baseline effective April 5th plus a 21% ‘reciprocal’ tariff effective April 9th) imposed on other Swiss exports like machinery, watches, and certain foods, the White House signals that ‘separate decisions are expected’ for the pharmaceutical sector. This contrasts sharply with lower rates applied to the EU (20%) and UK (10%), placing Swiss-based operations at a potential disadvantage. Financial analyst consensus suggests potential pharmaceutical-specific tariffs, possibly mirroring the high 31% rate and potentially announced within the next 1-2 months, creating a pressing need for contingency planning.
The US market is critical for Novartis, generating $21.15 billion (42% of total revenue) in 2024. An estimated 66-70% of these US sales ($14.0-14.8 billion annually, calculated based on analysis of key product manufacturing sites) originate from Swiss-manufactured products, making them vulnerable to future tariffs. (Context: This exposure sits within the ~$35.46 billion total Swiss pharmaceutical exports to the US in 2024, highlighting the sector’s overall significance). Key exposed products include high-revenue earners like Entresto, Cosentyx, and Kisqali.
Novartis possesses a substantial US manufacturing footprint (operating at an estimated 75-80% utilization based on industry benchmarks, providing partial insulation); US-made products like Zolgensma, Lutathera, and Pluvicto would be unaffected by tariffs on Swiss imports. However, under a potential 31% tariff scenario on Swiss pharmaceutical imports, Novartis could face $4.3-4.6 billion in additional annual costs (calculated as 31% of the exposed revenue range). Based on industry pricing dynamics, the company might absorb 20-25% of this cost ($0.9-1.2 billion), potentially compressing US business margins by an estimated 4-5 percentage points (net effect post-pricing actions) and reducing overall company EBIT by an estimated 5-7% (calculation sensitive to assumptions).
Novartis must leverage its significant US investments in negotiations while developing robust, product-specific contingency plans, including potential manufacturing transfers (estimated $600-800 million total cost over 3-5 years, based on benchmark costs for transferring ~8-10 lines), given the near-term risk of sector-specific tariffs.
| Metric | Current Status/Basis | Potential Impact Under Future Pharma Tariffs/Basis |
| Tariff Announcement | April 2, 2025 | – |
| Tariff Rate on Pharma | Exempt (pending ‘separate decisions’) | Uncertain – Hypothetical 31% used for modeling (mirrors other Swiss goods) |
| EU/UK Pharma Competitors | Also exempt | Level playing field currently; risk of disadvantage if Swiss pharma hit with 31% vs EU (20%)/UK (10%) other goods |
| Pharma Tariff Timeline | N/A | 1-2 months possible announcement window (Analyst consensus projection) |
| Other Swiss Goods Tariff | 31-32% total (10% Apr 5 + 21% Apr 9, 2025) | Sets precedent rate; targets machinery, watches, food etc. |
| US Trade Rationale | General goods trade balance concerns (~$38.5B deficit claimed by US) | Potential Section 232 cited by analysts; Contested by Swiss (services trade, investment, open policies, US methodology) |
| US Market Revenue (2024) | $21.15 billion (Novartis 2024 Report) | 42% of total company revenue ($51.72B Global) |
| Swiss-Manufactured US Rev | Est. $14.0-14.8 billion annually (Calculated: 66-70% of US Revenue) | Represents 66-70% of US sales exposed (Estimate based on key product sites) |
| (Context: Total Swiss Pharma Exports to US 2024) | (~$35.46 Billion overall sector value) | (Highlights scale of potential sector impact) |
| Key Swiss-Made Risk | Entresto ($5.2B global), Cosentyx ($4.8B), Kisqali ($1.9B) | US revenue streams from these global brands vulnerable |
| Key US-Made Products | Zolgensma ($1.4B global), Lutathera, Pluvicto, Cell Therapies | Protected from Swiss-import tariffs |
| US Manufacturing Capacity | Est. 75-80% utilization (Industry benchmark; actuals proprietary) | Limited immediate spare capacity (20-25%) for large-scale transfers |
| Annual Tariff Cost Est. | $0 | $4.3-4.6 billion (Calculation: 31% of $14.0-14.8B exposed sales) |
| Est. EBIT Impact | None currently | Potential 5-7% reduction globally (Calculation: $0.9-1.2 billion absorbed cost/~$17-18 billion est. EBIT base; sensitive estimate) |
| Est. US Margin Impact | None currently | Potential 4-5 percentage point compression (Net estimate post-pricing) |
| Manuf. Transfer Costs | N/A | Est. $600-800 million over 3-5 years (Sum of benchmark costs for ~8-10 lines) |
| Operating Cash Flow (2024) | $17.62 billion (+15% YoY) (Novartis 2024 Report) | Provides financial buffer |
| Net Profit Margin (2024) | 23.05% (Novartis 2024 Report) | At risk from absorbed costs & higher US COGS |
| Gross Profit Margin (2024) | 75.24% (Novartis 2024 Report) | Foundation strength, but COGS increase expected |
| R&D Investment (2024) | $11.4 Billion (Novartis 2024 Report) | Context for innovation; location decisions may be influenced |
The current US tariff landscape, announced April 2, 2025, presents a complex and evolving situation for Novartis:
Broad Swiss Tariffs Implemented: As of April 9, 2025, most Swiss exports to the US face tariffs totaling 31-32% (implemented in two stages: a 10% baseline on April 5th and an additional 21% ‘reciprocal’ tariff on April 9th). This rate, potentially driven by bilateral goods trade deficit concerns, is notably higher than tariffs announced for other major economies like the EU (20%) and UK (10%) in similar recent contexts, creating a potential competitive disadvantage for Swiss firms. Key targeted sectors include machinery, watches, processed foods, and medical technology.
Pharmaceutical Exemption (Explicit but Conditional): Critically, pharmaceuticals were explicitly exempted from this initial tariff implementation, alongside gold/precious metals. Official Swiss government communications and White House statements confirm this exemption but crucially include language indicating that ‘separate decisions are expected’ for the pharmaceutical sector. This strongly suggests the exemption is temporary or under review.
Anticipated Pharma Tariffs: Industry analysts interpret the specific carve-out and political signaling as precursors to potential pharmaceutical-specific tariffs. Projections suggest an announcement could occur ‘possibly in the next month or so.’ The mechanism might involve a Section 232 investigation, which assesses the national security implications of imports, a framework previously used for steel and aluminum.
EU/UK Competitor Parity (Current): Pharmaceutical exports from the European Union and the UK are also currently exempt from similar broad US tariffs, maintaining competitive neutrality for now. However, the final structure of any pharma-specific tariffs could alter this, particularly if Switzerland faces uniquely high rates.
Sector Vulnerability: The Swiss government itself acknowledges that its substantial goods trade surplus with the US (cited by the US as ~$38.5B in 2024) is ‘mainly attributable to exports from the chemical and pharmaceutical industry’ (approx. $35.46B in 2024). This high concentration makes the sector a politically visible target for US trade actions aimed at rebalancing bilateral goods trade flows.
Trade Context & Contrasting Policies: The US-Switzerland trade relationship, historically characterized by open markets, has seen significant growth, fueled by substantial Swiss investment in US R&D and manufacturing (Switzerland ranks 6th overall in FDI and 1st in R&D investment). Switzerland contests the US rationale, pointing out that the bilateral economic relationship is more balanced when trade in services (where the US has a surplus) is included. This contrasts sharply with Switzerland’s own move to abolish its industrial tariffs on January 1, 2024, highlighting differing national trade policy directions. The US actions occur despite these strong investment ties, and Switzerland disputes the US methodology for calculating the tariff basis.
Conclusion: While Novartis benefits from a temporary tariff exemption, matching EU/UK competitors, the explicit ‘separate decisions’ language, the sector’s high visibility in the Swiss goods trade surplus, the starkly higher tariff rate applied to other Swiss goods compared to the EU/UK, and analyst expectations for near-term action create significant uncertainty. The potential use of a 31% rate and mechanisms like Section 232 investigations requires immediate strategic consideration, particularly given the contested nature of the US trade rationale versus Swiss economic contributions and policies.
Novartis’s financial structure and significant US market reliance shape the potential impact of these tariffs:
US Market Criticality: The US market’s importance for Novartis is undeniable and growing. It accounted for $21.15 billion (42%) of Novartis’s total 2024 revenue ($51.72 billion), up from 40% in 2023 and 38% in 2022. This reflects the US representing roughly half of the global pharmaceutical market value and Novartis’s successful penetration. (Context: The overall Swiss pharmaceutical sector exported an estimated $35.46 billion to the US in 2024, indicating Novartis holds a major but not sole share of this potentially exposed trade flow).
Swiss-Manufactured Exposure: Estimated at 66-70% of US revenue, equating to $14.0 billion to $14.8 billion annually. This estimate is derived by analyzing the known primary manufacturing locations (predominantly Switzerland) for major US revenue contributors like Entresto, Cosentyx, and Kisqali relative to total US sales. Precision limited by proprietary nature of exact supply chain mapping for all SKUs. Key examples include:
Protected US Manufacturing Base: Novartis benefits considerably from its established US production network, shielding key innovative products from Swiss import tariffs. Notable US-made products include Zolgensma (Gene Therapy, $1.4 billion global sales), Lutathera and Pluvicto (Radiopharmaceuticals), and various cell therapies. Major US sites include Durham/Research Triangle Park, NC (Gene Therapy, Biologics), Libertyville, IL, Indianapolis, IN (Radioligands, Devices), Millburn, NJ (Cell Therapy), Holly Springs, NC (Vaccines, Biologics expansion underway), with Carlsbad, CA (Biologics/Cell/Gene Therapy) under development. Recent multi-hundred-million-dollar investments in sites like Indianapolis and Holly Springs underscore this commitment.
Potential Annual Tariff Cost Calculation: Applying the hypothetical 31% tariff rate to the estimated exposed revenue range ($14.0-14.8 billion) yields a potential direct annual cost impact of $4.34 billion to $4.59 billion. For planning, rounded to $4.3 – $4.6 billion.
Profitability Impact Analysis:
Financial Health Context: Novartis exhibits robust overall financial health, evidenced by a strong 23.05% net profit margin, 75.24% gross profit margin, healthy operating cash flow ($17.62 billion, up 15% YoY), and high R&D investment ($11.4 billion in 2024). This provides a crucial buffer. However, liquidity metrics showed some tightening in 2024 (Current Ratio decreased to 1.04 from 1.15), indicating careful cash management is necessary.
Conclusion: Potential tariffs represent a substantial financial threat, capable of eroding profitability significantly through direct cost absorption and indirectly via pricing power constraints. While Novartis’s financial strength and US manufacturing provide resilience, the magnitude of the exposure ($4.3-4.6B potential cost on $14.0-14.8B Swiss-origin US sales, within a $35.46B total Swiss pharma export sector) requires proactive management.
Novartis’s supply chain resilience hinges on its existing US footprint and the challenges of relocating production:
Existing US Footprint as Mitigation: The presence of multiple large-scale, technologically advanced US manufacturing sites (producing complex biologics, gene therapies, radiopharmaceuticals, cell therapies) significantly mitigates overall company risk. Vulnerability is concentrated on specific, high-value product lines currently imported from Switzerland.
US Capacity Limitations: Existing US facilities are estimated to operate at 75-80% utilization (based on industry benchmarks; actual rates are proprietary). This leaves relatively limited spare capacity (20-25%) to rapidly absorb large-scale, complex production transfers without substantial new capital investment and time for construction/validation.
Manufacturing Transfer Complexity & Timelines: Relocating pharmaceutical production, especially for biologics, is inherently costly, complex, and time-consuming:
Regulatory Hurdles (FDA): Manufacturing site changes require FDA approval, adding significant time:
Transfer Throughput Limitations: Given technical complexity and resources, Novartis could realistically manage the transfer of only 3-4 major product lines concurrently per year. This implies a potential 4-5 year timeline to significantly relocate the majority of prioritized vulnerable production if deemed necessary. Industry context: Surveys suggest ~40% of pharmaceutical firms anticipate needing over two years to fully adapt supply chains to major geopolitical shifts.
Conclusion: While Novartis’s US manufacturing presence is a major strategic advantage, capacity limits and the significant time (est. ~30 months+ for biologics), cost (est. $600M-$800M+ total), and regulatory hurdles for transfers mean mitigating risk for Swiss-made blockbusters requires substantial lead time and investment prioritisation.
The ability to pass potential tariff costs through pricing varies significantly by product and market segment:
Differential Pricing Power:
Pass-Through vs. Absorption Estimate: Across the portfolio, Novartis might successfully pass through approximately $3.4-$3.5 billion (~75-80%) of the potential $4.3-4.6 billion tariff cost via selective price increases, while needing to absorb the remaining $0.9-$1.2 billion (~20-25%), directly impacting margins.
Overall Revenue Risk: Estimated $300-450 million annual revenue erosion (approx. 1.5-2.1% of total US sales) from price-driven volume decreases.
Competitive Landscape & Payer Pressure:
Conclusion: Novartis faces difficult pricing decisions. While some tariff costs can likely be passed through selectively, limitations in competitive segments will necessitate significant margin absorption ($0.9-$1.2B) and potentially risk modest market share losses, requiring careful navigation of payer pressures.
A proactive, multi-phased strategic response is essential to manage the uncertainty and mitigate potential impacts:
Immediate Actions (Next 90 Days):
Medium-Term Actions (3-12 Months – Triggered if Tariffs Announced):
Long-Term Strategic Adjustments (1-3+ Years):
Conclusion: A dynamic strategy combining defensive actions (inventory, advocacy), reactive measures (pricing, CMOs), and proactive shifts (ROI-driven transfers, network optimization) is required, adaptable to US policy evolution.
Potential tariffs, even if managed proactively, could reshape Novartis’s US and global operations:
Scenario Planning Outcomes:
Structural Changes Likely Across Tariff Scenarios (Excluding Best Case):
Conclusion: Barring permanent exemption, Novartis likely faces structurally higher US operating costs, necessitating accelerated US manufacturing technology investment, portfolio rationalization, and reinforcing regional supply chains, posing significant cost and strategic challenges.
The report presents several key differences between a specialised financial AI Agent and both general large language models (i.e., GPT-4o, GPT-4.5, Claude 3.7 Sonnet) and more advanced reasoning models (i.e., o1, o3, Claude 3.7 Sonnet Thinking etc.).
Our experience from Novartis analysis highlights a few key insights on the need to have specialised agents, in this case for financial analysis:
In contrast, general LLMs excel at broad language tasks but falter where deep domain knowledge or real-time data is required. Reasoning models, while logical, lack domain-specific training as well as data access to rival a purpose-built agent.
There is a clear distinction between specialized Financial AI Agent and both general large language models (like GPT-4o, Claude 3.7 Sonnet) and advanced reasoning models (like o1, Claude 3.7 Sonnet with reasoning). This improved analysis explores these differences in depth, with concrete examples from the Novartis report.
Financial AI Agent demonstrates specialized expertise that transcends what general models typically provide:
While reasoning models can follow logical arguments about manufacturing challenges, they typically lack the embedded pharmaceutical industry taxonomies, regulatory knowledge, and manufacturing benchmarks that allows Financial AI Agent to analyze problems with insider precision.
Financial AI Agent performs complex, multi-variable financial calculations with industry-appropriate methodology:
While reasoning models can follow explicit calculation steps, they lack the purpose-built financial modeling capabilities that allows Financial AI Agent to perform complex, industry-specific analyses with appropriate confidence intervals and sensitivity testing.
Financial AI Agent seamlessly integrates and contextualize data from diverse sources:
Unlike reasoning models that primarily work with information provided in the prompt, Financial AI Agent demonstrates the ability to draw from proprietary financial databases, company reports, regulatory sources, and market intelligence, then synthesize this information cohesively.
Financial AI Agent produces output with business-optimized structure and implementation focus:
While reasoning models can explain their thought processes or provide strategic advice, they typically lack the standardized business frameworks and implementation orientation that Financial AI Agent brings to complex business problems.
Financial AI Agent makes assertions with confidence levels appropriate to industry standards:
While reasoning models can express uncertainty, Financial AI Agent demonstrates industry-calibrated precision that varies appropriately based on the type of information and industry norms for financial/strategic analysis.
A distinction not previously highlighted is how Financial AI Agent employs industry-specific reasoning patterns:
This represents a distinct advantage over both general LLMs and reasoning models, which may apply generic reasoning patterns without the embedded industry logic that guides pharmaceutical strategic decisions.
Financial AI Agent incorporates multiple data sources:
General LLMs typically rely on their training data, which may be outdated. Reasoning models, while better at logical problem solving, still primarily operate on information within their training data or provided in the prompt, without dedicated interfaces to specialized financial data sources.
The differentiation is evident in several key areas:
While general LLMs might provide strategic suggestions, they would struggle with the numerical precision and industry-specific operational details. Reasoning models, despite their logical capabilities, typically lack the specialized pharmaceutical knowledge to generate this level of integrated analysis.
Both utilize sophisticated reasoning, but with key differences:
Reasoning models use domain-agnostic reasoning applicable across subject areas, focusing on general logical consistency. However, they may lack the industry-specific constraints and precedents needed for effective pharmaceutical strategy reasoning.
Financial AI Agent combines several specialized components:
This modular architecture allows for the integration of specialized capabilities beyond what monolithic language or reasoning models typically provide.
Look for these distinctive indicators:
Despite its advantages, Financial AI Agent has distinct limitations:
The Novartis tariff analysis demonstrates that Financial AI Agent delivers capabilities beyond both general LLMs and reasoning models through:
These distinctions reflect a specialized system architecture combining domain-specific training, calculation engines, and industry knowledge that produces analysis difficult to replicate with even the most advanced general-purpose LLMs or reasoning models alone.
