How Hex Thinks Shared Context Will Unlock Enterprise AI ROI

As enterprises move beyond AI pilots, shared context and open infrastructure will determine whether intelligent agents deliver measurable business value at scale.
The enterprise AI conversation has changed. Over the past two years, organisations have invested heavily in large language models, copilots and generative AI applications, driven by the promise of greater productivity and faster decision-making. Today, however, boardrooms are asking a different question. Rather than debating which model to adopt, leaders want evidence that AI is delivering measurable business outcomes.
That shift took centre stage at MachineCon USA on July 24, where Carlos Aguilar, Head of Product at Hex, joined leaders from Fivetran, dbt Labs and other enterprise data teams for a panel titled ‘The AI-Native Mandate: Building the Foundation for Measurable AI ROI’.
The discussion reflects a growing consensus across the industry: organisations can no longer think of AI as a standalone capability. They need a foundation that enables people and AI systems to reason, collaborate and make decisions from the same trusted understanding of the business.
Hex believes the biggest obstacle to scaling enterprise AI is no longer the model itself, but the context surrounding it.
As Aguilar puts it, “Models are commoditising fast. Context isn’t. Every agent, workflow, and application pulling from a different, ungoverned understanding of the business doesn’t produce intelligence; it produces inconsistency at scale. Shared context is becoming as fundamental to the AI stack as the database was to the application stack.”
The same shift is visible across the broader AI ecosystem. As Remy Thellier, Head of AI/ML Partners at Snowflake, notes, “C-suites aren’t asking for another AI pilot. They want to see it working in production, with numbers behind it. Our joint customers get there by keeping AI analysis grounded in governed data on Snowflake, and now Snowflake’s AI functions run right inside the Hex workflow. That’s the difference between a demo and something that can actually be leveraged every day to drive significant value for the organisation.”
AI Doesn’t Have a Model Problem
Many enterprises believe they are AI-ready because they have invested in cloud platforms, data warehouses and business intelligence tools. Yet, these architectures were built for analytics and reporting, not for autonomous AI agents capable of continuously reasoning across workflows.
Different departments often calculate the same business metric differently. AI assistants generate conflicting recommendations because they interpret business logic independently. Employees spend more time validating AI-generated outputs than acting on them. According to Hex, these are not failures of AI models but of fragmented context.
Hex argues that many organisations delay AI adoption while trying to perfect their data estate. Instead, those seeing measurable returns begin with a recurring business process, assemble enough trusted context to support that workflow, and allow their data foundation to mature as AI delivers value. Rather than treating data quality as a prerequisite, they improve it by solving real business problems.
Shared Context as AI’s Competitive Advantage
For Hex, the defining challenge of enterprise AI is not connecting more data. It is ensuring every analyst, application and AI agent interprets data in exactly the same way.
Business metrics such as revenue, customer lifetime value or churn often have different definitions across departments. Human analysts understand these nuances through experience, but AI agents cannot. Two systems can access the same dataset yet arrive at different conclusions if they rely on different business definitions.
Aguilar believes the semantic layer is becoming a critical component of the AI stack. Rather than embedding business logic into individual dashboards or applications, organisations should create reusable semantics that every analyst and AI agent can build upon. Shared business definitions, governed metrics and common analytical logic become organisational assets rather than project-specific configurations.
Their customer experiences reinforce this philosophy. At ClickUp, combining product usage, customer engagement and billing data enabled AI to identify high-value churn risks, helping customer success and marketing teams personalise interventions and save more than $1 million in revenue churn. At Huckberry, bringing together historical sales data, promotional calendars and finance expertise produced more accurate inventory forecasts and savings exceeding $1 million annually. In both cases, AI generated value because it had access to shared business context rather than isolated datasets.
For the company, AI should complement analysts rather than replace them. Analysts contribute business judgement and institutional knowledge, while AI contributes speed, exploration and automation. Better decisions emerge only when both operate from the same trusted context.
Open Infrastructure is Not Just About Moving Data
As organisations prepare for an increasingly agentic future, open data infrastructure is becoming a strategic necessity.
Open table formats such as Apache Iceberg, interoperable APIs and portable architectures allow enterprises to move data across technologies without repeatedly rebuilding pipelines. But Hex argues that openness cannot stop with storage.
As organisations adopt multiple foundation models and specialised AI agents, the challenge is no longer moving data between systems, it is preserving business meaning wherever that data goes. An Open Data Infrastructure therefore extends beyond open formats to include an open semantic layer, ensuring business definitions, governance policies and analytical context remain portable across tools. Without context portability, data portability alone simply recreates fragmentation.
Rather than waiting for every data quality issue to be resolved, Hex advocates starting with a single, measurable business process, building enough trusted context to make AI useful, and allowing the foundation to mature through production. Every successful workflow expands the organisation’s context layer, making future AI initiatives easier to deploy.
Building for Measurable AI ROI
Hex’s CIO playbook encourages organisations to begin with business outcomes rather than platforms, identify where trusted context already exists, design workflows that business teams can own, measure success through operational outcomes rather than AI metrics, and allow the data foundation to evolve alongside real work.
This philosophy is evident across Hex’s customer stories. At EliseAI, AI reduced quarterly business review preparation from days to minutes while keeping analysts in control of the underlying reasoning. At Supabase, AI consolidated information spread across eight systems into a single workflow, enabling support engineers to resolve customer requests in seconds while maintaining human oversight before actions were executed. In both cases, AI succeeded because it became part of an existing business process rather than another standalone application.
Boardrooms will increasingly judge AI investments by business outcomes rather than technical achievements. Organisations that start with real business processes, build just enough trusted context to solve them, and allow that foundation to mature through production will be best positioned to realise measurable AI ROI.
[This article is part of the Brand Content initiative at AIM.]
Key Takeaways
- Recognize the need for shared context to enhance enterprise AI effectiveness and ROI.
- Transition from AI pilot projects to scalable implementations demanding measurable business outcomes.
- Establish open infrastructure that enables collaboration between people and AI systems.
- Prioritize understanding organizational context over merely selecting advanced AI models.
- Engage in industry discussions to align AI strategies with evolving enterprise demands.