Introduction

This report explores the practical adoption of artificial intelligence (AI) within sustainability solutions, moving beyond marketing language to capture signals from 23 industry leaders. The findings indicate that AI is becoming a foundational layer in the sector, though deployment remains uneven. Expertise is shifting away from data collection and cleaning towards validation, interpretation, client engagement, and decision support.

Top survey findings

The survey reveals that AI use is widespread, with 83% of respondents using the technology internally and 78% reporting at least some client-facing applications. Large language models (LLMs) are the dominant technology, utilised by 83% of providers. Agentic or multi-step workflow systems are emerging faster in this sector (48%) than in the broader enterprise market. While 82% of respondents highlight efficiency gains as the primary value case, 70% cite trust, governance, or security as significant adoption bottlenecks. Consequently, 96% of organisations maintain human oversight or controls.

Why AI deployment matters

Sustainability workflows are unusually data-intensive, drawing on emissions data, supplier questionnaires, invoices, and regulatory frameworks. Teams require audit-ready outputs from fragmented and inconsistent inputs. McKinsey estimates that procurement functions currently use less than 20% of available data for decision-making. AI’s practical value lies in extracting data from documents, matching information to reporting schemas, identifying anomalies, and generating evidence trails to support consistent analysis.

Methodology and data sources

KanataQ received 23 unique responses from providers across software, data, advisory, and consulting sectors. The participant profile skews towards smaller firms, with 57% having 1-10 employees and only 13% having more than 200. Forty-three per cent of respondents describe their main offering as a hybrid of software and advisory services, reflecting a market where domain expertise remains central even as AI is embedded.

Adoption vs. reality: widespread use, uneven deployment depth

While 78% of providers report client-facing AI use, deployment depth remains uneven. Fifty-two per cent state that AI is core to their product or service, yet only 39% have embedded it into key client-facing workflows. This aligns with McKinsey’s 2025 State of AI survey, which found that while nearly nine out of ten respondents use AI, most have not yet embedded it deeply enough to achieve material enterprise-level benefits.

Where AI is being applied

Current deployment is concentrated in high-frequency, data-heavy tasks. The most common functions are data extraction and processing (70%) and calculations and analysis (70%), followed by workflow automation (65%) and reporting (57%). Less mature areas include quality assurance (35%) and risk or compliance monitoring (30%), which require stronger data lineage, audit trails, and accountability.

Workflow interpretation

Data ingestion and normalisation represent the highest-volume opportunities, as sustainability teams must convert supplier inputs, invoices, and policies into structured data. AI value is increasingly moving below the reporting layer and into the data and workflow layers. Future growth in risk and compliance monitoring is expected as providers demonstrate better explainability and traceability.

Technology layer

The technology stack is dominated by LLMs (83%), which are favoured for their flexibility in handling unstructured information tasks such as summarisation and extraction. However, 48% of providers are moving towards agentic systems or multi-step workflows. This compares to only 23% of general enterprises scaling agentic AI systems, according to McKinsey. PitchBook reported that venture activity in agentic AI reached $24.2 billion across 1,311 deals in 2025.

Spending level: AI is being deployed frugally by many providers

AI deployment is not necessarily tied to large budgets; 57% of respondents report spending less than $50,000 annually on AI-related development. This suggests many providers are integrating AI through existing platforms and APIs. Long-term differentiation is expected to come from proprietary data, workflow ownership, and the generation of trusted client outcomes rather than model investment alone.

Key constraints

The leading barriers to integration are data security (39%) and internal team capacity or skills (39%). Grouped together, trust, governance, security, and reliability issues appear in 70% of responses. Sustainability providers handling sensitive supplier and emissions data must address these prerequisites to ensure controlled deployment and avoid risks such as unauthorised “shadow AI” use.

Human oversight

Human oversight is nearly universal, with 96% of respondents reporting controls for AI-enabled workflows. This is viewed as a feature of mature deployment rather than a sign of immaturity. Leading platforms like EcoVadis and ERM use AI to support experts rather than replace them, ensuring assessments remain reliable and defensible while increasing the speed and depth of insights.

Value creation: efficiency today, product transformation tomorrow

Efficiency remains the primary value driver, with 82% of respondents mentioning gains in time, cost, productivity, or speed. A second wave of value is emerging through product architecture, cited by 41% of respondents. This involves building AI-native workflows and capabilities, such as real-time tracking and granular data analysis, that were previously difficult to achieve without AI.

Case study: Earthchain

Earthchain is a UK-based platform that uses AI to convert fragmented business data into audit-ready carbon intelligence. The company applies computer vision to read documents and smaller specialised models to process data into activity-based inputs. This approach improves both speed and quality by keeping humans in the loop for final validation while providing confidence scores and audit trails for transparency.

Six AI deployment archetypes

The report identifies six archetypes for AI deployment: copilot overlays for search and summarisation; extraction and structuring engines for manual bottlenecks; estimation and enrichment layers for data gaps; evidence and assurance automation; continuous monitoring systems; and agentic orchestration to coordinate actions across modules and tools.

Strategic takeaways

Generic AI features, such as basic “chat with your report” tools, are becoming harder to defend. Differentiation for providers will likely strengthen around proprietary data assets, workflow ownership, and assurance readiness. Buyers increasingly expect AI-enabled productivity but will prioritise solutions that are secure, auditable, and provide clear answers on where AI sits in the workflow.

Future market outlook: from AI features to AI workflows

The most significant market shift is structural, as AI moves from being a feature attached to tools to a workflow layer connecting analysis, reporting, and decision support. The near-term market is moving towards bounded autonomy, where AI systems are supported by human experts responsible for accountability. Transitioning to AI-supported operating systems will be incremental and highly dependent on data quality.

Explore kanataq

KanataQ is a marketplace connecting businesses with trusted sustainability and ESG solutions. The organisation has redesigned its intelligence algorithms with AI to make processes faster and more scalable. It is also preparing to launch Matchy, an AI agent designed to help buyers identify best-fit solutions through email-based discovery and reasoning.