The Sustainability Reports Look Great. The Supply Chain Is Still On Fire.

There is a version of corporate sustainability that exists entirely on paper. It has targets, frameworks, disclosure schedules, and dashboards. It has been assured by consultants, reviewed by lawyers, and signed off by the board. It references the right acronyms in the right order.

And it is largely disconnected from the supply chain it is supposed to describe.

Not because the people behind it are dishonest. But because the data underpinning it is thin, the engagement driving it is shallow, and the generative AI being used to produce it is optimised for polish rather than truth.

The result is an entire industry that has become very good at describing a problem it has not yet seriously attempted to solve. I am not criticising intent. It is an observation about where the effort has gone, and where it hasn’t.

What the Numbers Actually Show

Let’s start with the basics.

Scope 3 supply chain emissions are, on average, 21 times higher than Scope 1 and 2 combined. They represent the overwhelming majority of most companies’ climate impact. Yet only 1 in 4 corporates even measures them. Of those, only 4 in 100 have a science-based target. And of those, across all EcoVadis-assessed companies in Europe, just 4% are on track to achieve it. (EcoVadis, 2025)

Four percent.

In Europe, roughly 35% of companies rely on proxy data, i.e., industry averages and spend-based calculations rather than actual supplier data. And 30% of suppliers provide no primary emissions data to buyers at all. (MIT, 2025)

The financial consequences for all of this are not abstract. Annual financial impact from supply chain emissions could exceed $500 billion by 2030. Supply chain disruptions, many linked to unmanaged environmental and social risk, already cost the global economy an estimated $1.6 trillion (EcoVadis, 2026). The EU’s 2030 climate plan anticipates carbon prices exceeding $120 per metric ton. This means that companies measuring Scope 3 with spreadsheets and spend-based proxies are not really managing this risk. They are merely deferring it.

This is a wide credibility gap. We have built a significant reporting architecture on foundations that a single honest conversation with a Tier 2 supplier would seriously challenge.

Tier 2 Is Where Reality Begins, and Where Visibility Ends

The picture at Tier 1 has genuinely improved. Nearly half of companies now report visibility into 75% or more of their Tier 1 suppliers, up from 27% just two years ago. That is real progress, and it reflects years of investment in supplier portals, SAQs, and disclosure programmes. (Accenture, 2026)

Tier 2 tells a different story.

Only 12% of companies have visibility into more than half of their Tier 2 suppliers (Accenture, 2026). The rest are operating with a blind spot in the part of the value chain where commodity extraction happens, where labour conditions are most precarious, where environmental violations are most likely, and where the actual emissions footprint of a product is largely determined.

This has been documented in every major supply chain sustainability study for the better part of a decade. What makes it urgent is what it means in the context of CSDDD, and the Omnibus rollback that has made it easier to defer action.

The Corporate Sustainability Due Diligence Directive was designed to change this. It was supposed to make Tier 2 visibility a legal obligation rather than a voluntary aspiration. The Omnibus package has scaled that back: smaller scope, longer timelines, fewer obligations. Every rollback sends the same signal to procurement teams: this is not urgent yet.

The risk, however, compounds quietly in the tiers nobody is looking at.

AI is Polishing the Reports with Problematic Data

Here is where the story gets more complicated.

Among companies using AI operationally in sustainability reporting, 53% are applying it to report writing and narrative generation (Reuters, 2026). In simple terms, most companies are prompting AI to simply ‘make this draft better/make it sound more professional’. Hence, AI is being used, at scale, to produce better-looking documents from the same underlying data.

The valuable applications, including tedious data gathering, supply chain data tracking, and scenario modelling, are each used by only 15% of organisations. These are precisely the applications where AI could have genuine impact: processing large, heterogeneous supplier datasets at speed, identifying anomalies, flagging gaps, and building a more accurate picture of what is actually happening in the chain.

However, I must say, the concentration on report writing is rational given current AI maturity. Report writing is a low-risk task. The AI generates text from data that humans have verified, and the output is reviewed before publication. Moving AI into data-intensive applications requires higher trust, more sophisticated integration, and a greater tolerance for error in the pipeline (Reuters, 2026). It is understandable that organisations start with the safer use case.

The concern is where investment and attention are flowing. A polished and well-structured sustainability report does not mean the underlying data is trustworthy. It just looks like it is. When AI is primarily used to improve the presentation of data rather than the quality of the data itself, it creates a false sense of security. Boards sign off on commitments grounded in proxy data. Procurement teams make supplier decisions based on narratives that AI has made convincing. Risk does not disappear because the report looks rigorous. It compounds quietly behind it, invisible until it isn’t.

And 47% of suppliers already say their customers are committed to sustainability on paper only (MIT, 2025). They see the targets. They receive the questionnaires. They do not believe the buyers are serious. That perception, held by the people who actually sit inside the supply chain, is the most important data point in this entire conversation.

The Only Intervention That Actually Works

Corporations that actively engage their suppliers are nine times more likely to deliver on their Scope 3 targets. (BCG, 2025)

Nine times. Companies that engage: who build real relationships with suppliers, co-invest in capability, link sustainability to commercial outcomes, and treat the supply chain as a partner in a shared problem.

Only 1 in 3 companies engage with suppliers in any meaningful way. Only 4 in 100 partner with them. (BCG, 2025)

This is the central tension of corporate sustainability in 2026. The evidence on what works is clear and consistent across every major study. The practice has not followed.

Partly this is about cost. High implementation costs are the second biggest barrier to Scope 3 reduction in Europe, cited by 18.4% of companies (MIT, 2025). The ROI case, however, is increasingly hard to ignore. Investing in climate action for the supply chain today could deliver 3 to 6 times return through avoided carbon-price exposure alone. The cost of engagement is real. The cost of not engaging is higher, and deferred costs have a way of arriving at the worst possible moment.

Partly this is about influence. Companies worry they cannot move suppliers who are larger, more powerful, or operating in markets where sustainability is not yet a commercial requirement. Suppliers, for their part, are not asking to be controlled. They are asking to be treated as partners: 45% want buyers to define clear and consistent metrics and goals; 39% want training and capability programmes; 37% want recognition for performance; 51% want sustainability performance linked to commercial outcomes (EcoVadis, 2026).

These are not unreasonable demands.

What Trusted Intelligence Actually Requires

The question this data raises is not technical. It is a question of where we want the effort to go.

Sustainability reporting, done well, is genuinely valuable. It creates accountability, drives disclosure, and builds the baseline of information that investors, regulators, and civil society need. The frameworks being developed under CSRD and CSDDD represent serious attempts to standardise that information and make it comparable across markets. None of this should be dismissed.

The challenge is that reporting and impact are not the same thing. A company can file a complete, assured, fully ESRS-compliant sustainability report and still have a Tier 2 supply chain where forced labour, ecosystem destruction, and untracked emissions are routine. The report describes a commitment. The supply chain contains the reality.

Closing that gap requires data that comes from suppliers, not merely about them. It requires human-governed AI, deployed against the messy, incomplete, heterogeneous datasets that represent supply chain reality, not against the clean outputs that already exist in corporate systems. It requires procurement relationships where sustainability performance carries commercial weight. It requires visibility past Tier 1, into the parts of the value chain where the actual decisions that determine environmental and social outcomes are made.

None of this is simple. None of it is cheap. And none of it fits neatly into an annual reporting cycle.

But the 9x multiplier on supplier sustainability does not come from better frameworks. It comes from companies that decide, seriously and commercially, to treat their suppliers as partners in a shared problem, and then do the slow, unglamorous, relationship-intensive work of building the trust that makes real data possible.

The reports will keep getting better. The supply chains will only change when the engagement does.