SNI: WEEK 07

tl;dr

This week’s AI-based developments across biopharma, medtech, life sciences, complex manufacturing and insurance reveal an acceleration of digital intelligence that is structural rather than incremental. 

AI is moving from advisory to autonomous, from augmenting human decision-making to executing workflows end-to-end. The enterprises that grasp this distinction are pulling ahead. Those treating AI as a bolt-on efficiency tool are discovering – through their share prices, if nothing else – that the market has already moved past them.

The key headlines from this week:

• Foundation models are entering drug design pipelines.

• Enterprise AI platforms are replacing fragmented point solutions in healthcare.

• Manufacturing is discovering that bolting AI onto legacy processes delivers less impressive results than redesigning operations around AI.

• Insurance watched its broker stocks crash after ChatGPT began quoting policies directly to consumers.

Emerging trends

Across all complex sectors, a common structural shift is visible. AI is transitioning from an advisory capability – generating recommendations, surfacing insights, assisting decisions – to an execution capability that autonomously performs workflows previously requiring human intermediation.

Isomorphic Labs’ IsoDDE doesn’t just predict protein structures; it designs drug candidates. MosaicOS doesn’t just flag anomalies; it drafts radiology reports. ChatGPT doesn’t just explain insurance; it quotes policies. Oracle’s supply chain agents don’t just recommend actions; they take them.

This shift from advisory to autonomous creates three consequences that cut across every industry covered here.

The first is a repricing of intermediation. Any business whose value proposition rests on connecting buyers with sellers, interpreting information for decision-makers, or executing routine professional workflows faces structural disruption. Insurance brokers, software vendors, clinical research organisations, and manufacturing system integrators all occupy this territory. The market is pricing this in – unevenly and sometimes excessively – but the direction is clear.

The second is a bifurcation between leaders and laggards. Companies redesigning their operations around AI - PepsiCo’s 20% throughput gain, Travelers’ 20,000-user deployment, BMS’s portfolio-wide clinical trial transformation - are pulling ahead of those bolting AI onto legacy processes. The MIT Sloan ‘productivity paradox’ research confirms what practitioners already know: AI delivers transformative returns when organisations restructure around it.

The third is a regulatory complexity that simultaneously constrains and protects.

The 1,250+ FDA-cleared AI devices that can’t get reimbursed, the 38 US state AI laws creating a patchwork of compliance requirements, the Colorado AI Act’s $20,000-per-violation penalties, and the EU AI Act’s August 2026 deadline all create friction.

For incumbents with compliance infrastructure, this friction is a moat. For AI-native disruptors, it is a barrier. The competitive outcome in each sector will depend on which force proves stronger.

Biopharma

The industry is shifting from AI as an experimental tool to AI as core infrastructure. A study found 78% of biopharma C-suite executives expect AI to play a central role in driving major change, yet only 9% report significant returns so far – a gap that is narrowing rapidly as billion-dollar deals become normalised.

The business model is also evolving: GSK’s $50 million licence of Noetik’s virtual cell foundation models established a new paradigm of licensing biological AI models as enterprise assets rather than project-scoped services.

AI drug design leaps from prediction to action

Google DeepMind spinoff Isomorphic Labs unveiled IsoDDE – the Isomorphic Labs Drug Design Engine.

IsoDDE goes far beyond AlphaFold’s protein structure prediction, representing a full-cycle computational drug design system that more than doubles AlphaFold 3’s accuracy on difficult targets.

It is up to 20× better than Boltz-2 on antibody benchmarks, outperforms physics-based gold standards on binding affinity prediction at a fraction of the cost.

It also identifies binding pockets from sequence alone – including sites that took researchers 15+ years to discover experimentally.

Demis Hassabis expects the first AI-designed drugs from Isomorphic to enter clinical trials by the end of 2026.

Deal flow

Takeda signed a $1.7 billion deal with Iambic Therapeutics for access to its NeuralPLexer AI drug discovery platforms, one of the largest AI drug discovery agreements to date.

Bristol Myers Squibb partnered with Evinova, AstraZeneca’s health tech subsidiary, to deploy AI-native clinical trial design across its entire global portfolio, with BMS’s CMO calling the transformation ‘an urgent necessity.’

Miami-based Famous Labs launched Heisenberg, a quantum-informed AI system that optimises which molecules to physically synthesise, addressing the bottleneck between computational generation and experimental resources.

Path to market

Generate Biomedicines progressed its Nasdaq IPO filing after dosing the first patient in Phase 3 trials of GB-0895, an AI-designed antibody for severe asthma – making it the first major AI biotech IPO with a pivotal-stage asset.

Insilico Medicine was featured in a Harvard Business School case study for Rentosertib, the world’s first drug with both an AI-discovered target and AI-designed structure to reach Phase 2, with Nature Medicine data showing meaningful lung function improvements.

South Korean startup Galux raised $29 million for AI-driven de novo protein design.

MedTech and digital health

The MedTech sector faces a paradox – unprecedented AI regulatory approvals alongside a growing ‘approval-to-payment’ gap where cleared AI tools cannot get reimbursed.

Enterprise platforms are replacing fragmented point solutions, but the real bottleneck is economic integration, not technical capability.

Research and real-world performance

The University of Michigan created an AI system that interprets brain MRIs in seconds, accurately identifying neurological conditions and determining urgency – a potential transformation of emergency triage workflows.

A sobering counterpoint emerged, however: a Nature Medicine study found that while AI chatbots could identify correct conditions in 94.9% of cases independently, accuracy dropped to less than 34.5% when used by actual patients, with chatbots fabricating information in several instances.

This gap between controlled and real-world performance is a critical consideration for consumer-facing health AI.

Enterprise platforms replace point solutions

Radiology Partners, the largest US technology-enabled radiology provider, launched MosaicOS – a cloud-based, AI-native radiology operating system combining diagnostic AI and workflow tools. Its drafting module uses a multimodal foundation model to pre-draft X-ray reports, while the recently acquired Cognita Imaging demonstrated 52% increased detection rates and a fourfold reduction in diagnostic errors.

MosaicOS is pursuing FDA clearance and represents the sector’s shift from fragmented point solutions to integrated platforms.

Regulatory developments

The FDA’s 6 January guidance eased regulation of clinical decision support software – the most significant deregulatory shift for AI health products in years. But a counterbalancing force has emerged: a Medicare Administrative Contractor proposed denying coverage for AI brain MRI technologies. And the ACR launched an AI Economics Committee to address the widening gap between FDA clearance of 1,250+ authorised AI devices and payer coverage.

Meanwhile, MedCognetics received FDA 510(k) clearance for CogNet AI-MT+, an AI breast imaging triage tool trained on diverse global datasets to address health equity concerns. RevealDx secured clearance for RevealAI-Lung, which reduced delayed cancer diagnoses from 45% to 25% in studies.

Complex manufacturing 

Manufacturing is experiencing a bifurcation. Companies that redesign operations around AI achieve transformative results – PepsiCo experienced a 20% throughput gain. But those bolting AI onto legacy processes are seeing less positive returns.

The skills gap is the critical bottleneck – despite manufacturing job cuts, severe shortages persist for AI and digital twin expertise. SMIC’s overcapacity warning suggests the infrastructure buildout may be outrunning actual productive use cases. The AI Factory

SEMICON Korea 2026 was the sector’s biggest event, with NVIDIA’s Timothy Costa outlining the ‘AI Factory’ vision – purpose-built infrastructure where AI manages the entire semiconductor manufacturing lifecycle. Not unlike our own ‘Second Nature Manufacturing’ vision.

NVIDIA introduced an agentic AI architecture connecting cloud systems, secure on-

premise factory servers and embedded AI within test cells. The roughly 550 exhibitors across 2,400+ booths at COEX Seoul reflected the AI-driven semiconductor boom, with the global market projected to reach $975 billion in 2026 - 26% growth.

The overcapacity warning

A critical caution emerged from SMIC, China’s top chipmaker, which warned that breakaway AI chip spending is pulling forward years of future demand, raising the risk that some data centres ‘could sit idle’. Co-CEO Zhao Haijun stated: ‘Companies would love to build ten years’ worth of data centre capacity within one or two years. ‘As for what exactly these data centres will do, that hasn’t been fully thought through’. This warning gained credibility alongside MIT Sloan research documenting a ‘productivity paradox’ in manufacturing AI adoption – early deployments that layer AI onto existing workflows without redesigning operations produce limited or uneven results. Microsoft cited 457% projected ROI over three years only for properly integrated AI implementations.

Enterprise deployment

Oracle launched AI agents for supply chain and manufacturing at its AI World event, embedding agentic AI across planning, procurement, manufacturing, maintenance and logistics workflows.

SAP detailed AI agent transformation in automotive supply chains, with executives expecting AI to boost product value by 22% and digital service value by 37% within three years.

The Pentagon’s $13.4 billion autonomy budget included Kodiak AI winning a Marine Corps autonomous vehicle contract and Overland AI’s $100 million raise for military ground vehicles.

Notable partnerships

The Siemens-NVIDIA industrial AI operating system is launching its first fully AI-driven adaptive manufacturing site at Siemens’ Erlangen factory in Germany, with PepsiCo reporting 20% throughput increases and 10–15% capex reduction as an early adopter. Dassault Systèmes and NVIDIA announced ‘Industry World Models’ – physics-validated AI systems for manufacturing simulation. Siemens also acquired Canopus AI for semiconductor wafer and mask inspection.

Insurance

The sector is simultaneously being disrupted by AI and emerging as a governance mechanism for AI risk across other industries. The dramatic broker stock crash mirrors the software selloff: markets are pricing in a world where AI-native distribution could bypass traditional intermediaries. But the regulatory complexity creates significant barriers to rapid disruption, suggesting the market reaction may have overshot in the near term while being directionally correct over the medium term.

The ChatGPT distribution shock

The most dramatic story came from the launch of ChatGPT-native insurance applications. Insurify released the industry’s first ChatGPT comparison app for auto insurance, drawing on 196 million historical quotes, while Spanish digital insurer Tuio became the first to receive OpenAI approval for real-time home insurance quoting inside ChatGPT.

The combined news triggered the sector’s worst trading day in years: the S&P 500 Insurance Index fell 3.9%, Willis Towers Watson plunged 12%, Arthur J. Gallagher fell 9.9% and Aon dropped 9.3%.

With ChatGPT’s 800 million weekly users and 12+ additional insurance AI apps in OpenAI’s pipeline, analysts at Berenberg and Morningstar called the selloff ‘overdone’ but acknowledged the long-term disintermediation threat to commoditised insurance lines.

Investment data

Gallagher Re’s Global InsurTech Report revealed Q4 2025 funding surged 66.8% to $1.68 billion, with full-year 2025 investment hitting $5.08 billion. AI-focused companies captured two-thirds of total insurtech funding ($3.35 billion across 227 deals), with (re)insurers themselves making a record 162 investments.

Gallagher Re’s Global Head of Insurtech stated: ‘AI is squarely the focus of most of the contemporary InsurTech world’.

Enterprise deployment

Travelers disclosed that 20,000+ professionals already use AI tools regularly, with Claude AI assistants deployed to nearly 10,000 engineers and analysts – one of the largest generative AI integrations in financial services.

Allianz UK scaled its ‘BRIAN’ underwriting tool from pilot to production across 260 underwriters, processing queries sourced exclusively from verified documents to ensure compliance.

Microsoft and Cognizant co-authored guidance on embedding agentic AI across claims workflows, while Sedgwick’s Sidekick improved processing efficiency by 30%+.

An Accenture survey of 430 underwriting executives found AI adoption is expected to jump from 14% today to 70% within three years.

Regulatory developments

The Colorado AI Act became the first comprehensive US AI consumer protection law covering insurance, requiring reasonable care against ‘algorithmic discrimination’ with penalties up to $20,000 per violation.

The Colorado Division of Insurance expanded governance requirements to auto and health insurers. The Center for Democracy and Technology published research exploring insurance as a governance mechanism for AI risk – using premium structures and policy conditions to incentivise safer AI practices across industries.

ATA launched a $750 million insurance facility at Lloyd’s specifically to underwrite the $7 trillion AI infrastructure boom, backed by Arch Insurance, Munich Re, and SCOR. 

Thank you for reading this week’s report. Come back next week for all the AI news you need to know in your sector.