Alteryx research finds that 53% of organisations struggle to translate business context into AI systems, exposing a widening gap between rising technology investments and the ability to deliver measurable business outcomes.
Artificial intelligence investment is accelerating across enterprises, but the ability to convert that spending into dependable business value remains uneven. Organisations are deploying AI tools, experimenting with agents and expanding automation initiatives. Yet a fundamental question persists: do these systems understand how the business actually works?
New research from Alteryx suggests that this is becoming one of the most significant obstacles to enterprise AI adoption. According to its 2026 IT Leader Research: The State of AI Ownership, Agents, and ROI, 53% of organisations struggle to translate business context into AI systems and workflows. The finding comes at a time when 80% of technology leaders expect their organisations’ AI spending to increase over the next two years.
The contrast is important. Enterprises are prepared to commit more capital to AI, but many have yet to establish the organisational knowledge, data foundations and collaboration models required to make those investments consistently productive. The challenge is no longer simply gaining access to AI models or deploying applications. It is ensuring that AI understands the rules, exceptions, objectives and operating processes that shape business decisions.
The research, based on a survey of 1,400 technology leaders globally, also indicates that AI is already generating returns for many organisations. Some 69% report moderate or significant returns on their AI investments. However, the gap between this progress and the difficulties in translating business context suggests that current returns may not automatically translate into sustainable, enterprise-wide value.
The business-context problem
AI models can process large volumes of information, identify patterns and generate responses, but their usefulness in an enterprise depends on the quality and relevance of the context they receive. A model may be technically capable of analysing financial data, forecasting demand or recommending supply chain adjustments. However, without understanding the organisation’s policies, business rules, operational constraints and decision-making practices, its recommendations may not align with the way the enterprise functions.
Business context is often more complicated than the structured data stored in enterprise applications. It includes assumptions used in financial forecasts, approval hierarchies, regulatory obligations, pricing policies, customer exceptions and the practical knowledge employees acquire through years of experience. Much of this information remains distributed across spreadsheets, documents, email conversations, macros and departmental systems.
Consider a financial forecasting process. The underlying data may show revenue, expenses and cash flow, but the forecast is also shaped by assumptions about market conditions, business-unit targets, seasonal patterns and management adjustments. An AI system that sees the numbers but not the assumptions may produce an output that appears credible while missing the reasoning behind the decision.
The same problem applies to tax compliance, procurement, inventory planning and customer servicing. Supply chain recommendations, for example, need to account not only for stock levels and demand forecasts but also for supplier constraints, delivery schedules and operational priorities. Without access to these rules, AI-generated recommendations may be incomplete or impractical.
The Alteryx findings underline why business context is moving from a supporting consideration to a core requirement for enterprise AI. In the survey, 77% of technology leaders said business context is essential to generating accurate and relevant AI outputs. This suggests that enterprises increasingly recognise that model capability alone cannot guarantee business relevance.
“AI systems need more than access to data; they need the right business context to generate outputs that are relevant, reliable and aligned with organisational requirements.”
— Sabya Sen, Vice President, IMEA & APAC, Alteryx
India’s AI ambitions meet data-readiness constraints
The issue is particularly relevant to India, where organisations are expanding AI programmes across banking, manufacturing, retail, services and the public sector. Investment momentum is strong, but the foundations required to support AI at scale remain uneven.
A separate Dun & Bradstreet AI Momentum Survey for India, released in August 2026, found that all surveyed Indian businesses had AI-related projects underway, while 69% planned to increase their AI investments. However, only 4% said their enterprise data was fully ready to support AI at scale.
The contrast highlights a structural challenge for Indian enterprises. AI adoption is advancing, but the data architecture and governance practices needed to support widespread deployment are not progressing at the same pace. Although 73% of surveyed businesses reported some measurable returns from AI investments, limited data readiness could constrain the next phase of adoption.
For organisations, data readiness involves more than storing information in modern platforms. Data must be accessible, accurate, consistent, appropriately governed and connected to the business processes in which it is used. Enterprises may have invested in cloud data platforms, analytics tools and AI applications, but fragmented systems and inconsistent definitions can still prevent teams from establishing a reliable view of business operations.
The problem becomes more pronounced when AI agents are introduced. Unlike conventional analytical tools that primarily generate reports or recommendations, agents are increasingly expected to perform tasks, interact with applications and execute workflows. Their actions can affect customer records, financial processes, operational systems and business decisions. In such environments, incomplete context is not merely an analytical limitation; it can become an operational and governance risk.
Sabya Sen, Vice President, IMEA & APAC, Alteryx, believes the next phase of enterprise AI will depend on connecting technology with the knowledge and processes that drive business operations.
“India is moving rapidly towards wider enterprise AI adoption, with organisations increasingly investing in AI to improve productivity, efficiency and business outcomes. However, the next phase of this journey will depend on how effectively businesses connect AI with the knowledge and processes that drive their operations,” Sen said.
He added that making business logic accessible, governed and repeatable will be critical as enterprises move from experimentation to scaled deployment.
Data access remains a bottleneck
Even when organisations have invested in data platforms, employees who understand business operations may not have the access or technical capabilities needed to use the information effectively. The Alteryx research identifies this as another obstacle to enterprise AI adoption.
Only 18% of organisations have achieved fully self-service access to cloud data for business users. Another 38% report a mixed model of data access, while 15% say business users remain largely dependent on technical teams.
This has implications for the speed and quality of AI development. Business teams often understand the operational problem, the relevant exceptions and the decisions that need to be improved. IT teams, meanwhile, manage data infrastructure, security, integration and technology governance. When access to data and the ability to build workflows remain concentrated within technical departments, the people closest to the business may struggle to translate their knowledge into usable AI applications.
The resulting dependence can create bottlenecks. Business users must explain requirements to technical teams, wait for data preparation and then review outputs that may not fully reflect operational realities. As AI initiatives multiply across departments, this model can become difficult to scale.
Self-service access, however, should not be confused with unrestricted access. Enterprises need to balance usability with security, privacy, data quality and regulatory obligations. The objective is to give authorised business users the ability to work with relevant data within defined governance boundaries.
This requires common data definitions, clear ownership, reusable business rules and mechanisms to track how information is prepared and used. Without these controls, self-service analytics and AI can create new inconsistencies rather than resolve existing ones.
IT and business collaboration becomes essential
The question of who owns AI is also evolving. Traditionally, enterprise technology strategy, infrastructure and application delivery have been led by IT departments. But AI initiatives increasingly cut across business functions, making a purely technology-led approach insufficient.
The Alteryx survey found that 71% of technology leaders believe AI initiatives are most successful when IT and business teams work closely together. Around two-thirds also believe AI and agent-based systems are most productive when managed within business functions.
These findings point towards a more distributed model of AI ownership. IT must continue to provide the architecture, integration, security and governance framework. Business teams need to define the intended outcomes, supply operational context and validate whether AI-generated outputs are useful in real-world situations.
This shared model can help close the gap between technical capability and business relevance. For example, a finance team may understand the assumptions and approval rules behind budgeting, while IT ensures that the relevant data sources are integrated and access is controlled. Together, they can develop an AI-enabled workflow that reflects actual business requirements rather than simply automating an existing manual process.
The same principle applies to manufacturing, procurement, customer service and supply chain operations. AI projects are more likely to deliver measurable outcomes when the people who understand the processes are involved in designing, testing and governing the systems.
However, collaboration needs to be supported by clear accountability. Enterprises must define who owns business rules, who approves changes to workflows, who monitors agent behaviour and who is responsible when an automated decision produces an incorrect outcome. As agents gain greater autonomy, these questions will become more consequential.
From AI experiments to repeatable business outcomes
The transition from AI experimentation to enterprise-wide deployment requires organisations to rethink how they manage business knowledge. Rather than treating spreadsheets, macros, documents and employee expertise as disconnected resources, enterprises need to make the underlying logic visible and reusable.
This is where governed workflows and analytics automation can play a role. Business rules can be documented, translated into repeatable processes and maintained as operational requirements change. When these rules are accessible to AI systems, organisations can improve the consistency of outputs and make it easier to understand how a recommendation or action was generated.
Alteryx positions its Alteryx One platform around this requirement. The platform is designed to help business users prepare AI-ready datasets and define, maintain and update business logic through visual workflows. The company describes this logic as visible, understandable, repeatable and auditable. The broader proposition is to connect data platforms with the decisions and operations that run the business.
The underlying principle is relevant beyond any single technology platform: enterprises need a dependable layer of business logic between raw data, AI models and operational execution. Without it, AI initiatives risk becoming isolated experiments that are difficult to govern, reproduce or scale.
The challenge is also to measure AI value in business terms. Increased model usage, higher numbers of automated workflows or faster data processing do not necessarily demonstrate improved business performance. Organisations need to assess whether AI reduces operating costs, improves customer experience, shortens decision cycles, increases productivity or reduces risk.
This requires baselines, clear performance indicators and ongoing monitoring. It also means evaluating AI systems not only on the quality of their outputs but on whether those outputs lead to better decisions and outcomes.
The next phase of enterprise AI
The Alteryx research captures a central tension in the enterprise AI market. Organisations are increasing their spending and reporting returns, yet many still lack the data readiness, business context and operating models required to scale AI consistently.
For Indian enterprises, the opportunity is substantial, but the next phase will demand greater attention to the foundations beneath AI adoption. Data quality, business-process documentation, access controls, governance and collaboration between IT and business functions will determine whether AI can move beyond isolated use cases.
As agent-based systems become more capable of taking action, the cost of overlooking these foundations could rise. An AI system that misunderstands a business rule may produce an inaccurate recommendation; an autonomous agent operating with that misunderstanding could carry it into an enterprise workflow.
The strategic priority, therefore, is not simply to deploy more AI. It is to make organisational knowledge accessible to AI in a way that is controlled, explainable and continuously updated. Enterprises that achieve this will be better positioned to turn AI investments into repeatable business outcomes. Those that do not may find that rising expenditure produces more experimentation without a corresponding improvement in enterprise performance.
The next competitive advantage in AI may not come from using the most powerful model, but from ensuring that the model understands the business it is expected to serve.
