Five themes dominated the 2026 Sustainability World Summit in Frankfurt: reporting maturity under the revised CSRD, data quality, Scope 3 measurement, climate risk as financial risk, and execution capacity.
Reporting maturity is shifting
Mandatory CSRD scope has narrowed.
The revised thresholds (more than 1,000 employees AND over €450M EU turnover) cut the in-scope population substantially compared with earlier drafts. Stop the Clock delayed Wave 2 (large EU companies) and Wave 3 (listed SMEs) by two years, and the Omnibus package simplified the rest.
The audit trajectory is firming up. Limited assurance is required now, with reasonable assurance to follow, and sustainability data is heading toward the same scrutiny as financial reporting. Mandatory scope is shrinking, but voluntary demand keeps climbing on the back of customer RFPs, CDP and EcoVadis questionnaires, and private-equity portfolio scoring. 70% of companies in a recent Bertelsmann Stiftung survey now see CSRD as an advantage rather than a burden.
Data quality is the foundation
Bad data breaks every downstream use case. The numbers already exist somewhere in the enterprise, scattered across procurement, ERP, EHS, HR, travel, energy, and supplier portals. The problem is that they sit in different systems, with mismatched units, inconsistent supplier proxies treated as primary values, and large gaps at facility level.
Dun & Bradstreet finds nearly half of companies have critical data gaps that create regulatory risk in the supply chain. 85% of European companies report that inaccurate or incomplete master data costs them every year, and Gartner puts the operating profit hit at up to 20%. Audit-grade requirements (lineage, provenance, version control, change history) are now table stakes for any tool sitting on top of that data. Companies have to improve the data before automating around it.
Scope 3 is moving from spend-based to activity-based
Spend-based factors are losing credibility. The same chemical can carry a different carbon footprint depending on supplier, geography, technology, and energy mix, and Scope 3 routinely exceeds Scope 1 and 2 by a factor of four to nine. Activity-based and primary supplier data are becoming the working standard for material categories. Even imperfect primary data outperforms polished spend-based proxies, which lowers the bar for supplier engagement. Purchasing decisions on large-volume materials can cut carbon footprint by an average of 38%, the kind of swing only activity-based data can detect. Supplier engagement is the bottleneck. Only 30% of suppliers respond to initial surveys, and survey fatigue is widespread.
Climate risk is financial risk
Physical climate risk has crossed from CSR concern to balance-sheet exposure. The WEF Global Risks Report 2025 ranks extreme weather as the top short and long-term risk identified by global risk experts. Scenario analysis, asset-level risk assessment, and net risk after insurance and adaptation are now standard inputs into enterprise risk management. Insured losses are climbing 5.9% per year and on track to double in a decade.
142 insured-loss natural catastrophes were recorded in 2025, and delayed action could cost over $1 trillion annually by 2050. A 3°C warming scenario could push insurance into unaffordable territory and destabilise key financial systems. The prudent response runs on two tracks. Adaptation for resilience and mitigation for decarbonisation, not one or the other. Insurers themselves are now a transmission mechanism, repricing climate risk into mortgage books, commercial real estate, and infrastructure debt.
Execution capacity is the binding constraint
Most enterprises have a transition plan on paper. The majority show slow or no progress against it; EY's 2025 Global Climate Action Barometer puts the figure at 64% with a plan and most of those stalled. ESG has evolved from optional commitment into the license to operate, and the constraint has moved from technology to delivery. In practice that looks like cross-functional ownership across CFO, COO and CHRO offices rather than a CSO solo lift, business cases that connect sustainability to operational decisions, and board-level accountability against named milestones. Deloitte's 2025 staffing survey ranks sustainability learning and development as the lowest-scoring capability area across the enterprise.
The talent picture cuts both ways. 70% of workers say they would accept the same job at a more sustainable company, but two-thirds can't find sustainability information on the company in question.
Implications
Reporting maturity is rising. Data quality, Scope 3, climate risk, and execution capacity are not keeping pace. Closing that gap will take a different kind of system, not a bigger team and not another dashboard.
Sustainability data is uniquely scattered. It sits across dozens of source systems, follows different units and methodologies, crosses organisational boundaries, and changes meaning depending on who is asking and why. Conventional automation cannot reason over it, because the meaning is not in any single record but in the relationships between records. No human team can keep up with that reading load in the time the decision allows. The next decade of enterprise sustainability will be carried by AI agents, not by interfaces.
For agents to be useful at this depth they have to go vertically into the data layer, not skim across the surface. That means understanding context, not just retrieving values. An emission factor only means something in relation to the supplier, the facility, the methodology, the regulatory framework, and the prior-period decision that set the baseline. Agentic AI in sustainability has to operate on top of a connected representation of all of that, with provenance attached to every value and audit trails attached to every change. Without that grounding, AI is a guessing machine. With it, intelligence compounds year over year and the system gets better, not noisier, every reporting cycle. That is where the industry has to go.
If you're working through any of this for your 2026 or 2027 cycle, we'd welcome the conversation.
