| Definition Chat with Data is the practice of querying business data through natural language, asking a question in plain text or by voice and receiving an immediate, data-grounded answer rather than navigating a dashboard or requesting an analyst report. In 2026, Chat with Data is evolving rapidly: LLMs are making it conversational across multiple turns, agentic AI is making it proactive, multimodal interfaces are extending it beyond text, and deeper integration with enterprise systems is making it available wherever decisions are made. |
Chat with Data has moved from a compelling concept to an enterprise reality in a short period of time. The question has shifted from ‘is this possible?’ to ‘how is it evolving, and what does the next generation look like?’
This blog covers the recent trends reshaping how Chat with Data works, the integration patterns that are making it accessible where decisions happen, and where the technology is heading in the near term. For the foundational guide covering what Conversational BI is, how it works technically, and how to implement it, see our Conversational BI complete guide. This blog picks up where that guide ends.
Table of Contents
Where Chat with Data Stands in 2026
The foundational technology for Chat with Data has matured significantly. Large language models have replaced the narrow, keyword-matching NLP systems that powered earlier natural language query tools, enabling genuinely flexible, context-aware interpretation of business questions. Users can now phrase a question in any number of ways and receive a consistent, accurate answer, something that earlier systems struggled with when even minor variations in phrasing broke the query.
The integration layer has deepened too. Chat with Data is no longer primarily a standalone tool with a dedicated query interface. It is increasingly embedded within the platforms, channels, and applications where business teams already spend their time: Microsoft Teams, Slack, Power BI, ERP interfaces, and CRM dashboards. The transition from ‘a tool you go to’ to ‘a capability that comes to you’ is the most significant operational change in how Chat with Data is used day to day.
According to a Gartner survey cited by the existing chat-with-data ecosystem, 85% of customer service leaders plan to explore or pilot conversational GenAI solutions, reflecting how broadly the natural-language interaction model is being adopted across enterprise functions, not only analytics.
Recent Trends Reshaping Chat with Data
From Single Queries to Multi-Turn Conversations
The first generation of natural language query tools answered one question at a time. Each query was independent: the system had no memory of what was asked before and could not build on a previous answer. Modern Chat with Data systems, powered by LLMs, maintain conversational context across multiple turns.
This changes how users interact with data fundamentally. A finance manager can ask ‘What are our top five cost categories this quarter?’, receive the answer, and follow up with ‘Which of those is growing fastest year over year?’ without restating the context. The conversation flows the same way it would with a knowledgeable colleague, with each answer building on the last.
Multi-turn dialogue also enables collaborative data exploration: a team in a meeting can ask a sequence of progressively more specific questions in real time, drilling from a high-level summary into the root cause of a trend, all within a single conversation thread.
Agentic AI: From Responsive to Proactive
The most significant trend in Chat with Data for 2026 is the shift from reactive to proactive analytics through agentic AI. In a reactive system, a user asks a question and the system answers it. In an agentic system, the AI monitors data, identifies patterns that are business-relevant, and surfaces those insights without being asked.
An agentic Chat with Data system might detect that a specific product’s margin has declined three weeks in a row, identify that the driver is a cost increase in a particular input, and surface this finding to the relevant category manager before the trend becomes a P&L problem. The insight arrives in the user’s Teams channel or email inbox as a narrative summary with the supporting data, rather than waiting for the user to ask the right question.
This capability requires more than LLM integration: it needs orchestration logic that defines which metrics to monitor, what constitutes an anomaly worth surfacing, and who should receive which insight. Getting the governance of proactive AI right is as important as the technology itself.
Multimodal Interfaces: Beyond Text
Chat with Data began as a text interface. The next evolution is multimodal: systems that accept voice, images, and combinations of input types alongside text, and that return answers in the most useful format for the context.
Voice-based Chat with Data is now production-ready for a range of enterprise use cases. A warehouse manager can verbally ask about stock levels during a physical inventory review without touching a device. An executive can ask a data question out loud during a review meeting and have the answer displayed on screen within seconds.
Image-based input is an emerging capability: users can share a screenshot of a chart or a photo of a physical document and ask the system to interpret it in the context of their data. While this is in early deployment for most enterprise users, it represents an extension of the ‘chat with your data’ model into the physical and visual world that will become more prominent through 2027 and beyond.
Narrative AI and Data Storytelling
Returning a chart or a table in response to a data question is useful. Returning a chart alongside a plain-language narrative that explains what the data means, what drove the pattern, and what it implies for a decision is significantly more valuable for business users who may not interpret the visualisation instinctively.
Narrative AI in Chat with Data generates the explanatory text automatically, drawing on the retrieved data and the context of the question to produce commentary that mirrors what an experienced analyst might say. A sales leader asking ‘Why did North America underperform target in Q2?’ receives not just a variance table but a coherent explanation: ‘North America missed target by 8%, primarily driven by a 23% decline in enterprise deals in the financial services vertical, which offset a 12% overperformance in the technology sector.’
This narrative layer significantly reduces the interpretation burden on business users, particularly those who interact with data less frequently and may not have an intuitive feel for how to read a chart or a pivot table.
Integration Patterns: Where Chat with Data Is Being Deployed
The value of Chat with Data is proportional to where it is available. A capability that requires a user to switch to a separate application is less valuable than one embedded in the tools where they make decisions. Here are the integration patterns driving the most adoption in 2026.
Embedded in Collaboration Platforms
Microsoft Teams and Slack are the primary collaboration environments for enterprise teams in 2026. Chat with Data integrations in these platforms allow users to ask data questions in the same channel where they discuss the decisions those questions inform. A sales team reviewing a deal in a Teams thread can ask directly within that thread for the account’s revenue history, deal stage progression, or market opportunity data, without opening a separate BI tool.
These integrations typically work through bot interfaces or app integrations that connect the collaboration platform to the underlying semantic model and data warehouse. The response arrives as a formatted message within the thread, visible to all participants, turning a data query into a shared team reference rather than a private analyst request.
Within Existing BI Platforms
Microsoft Power BI’s Copilot and Q&A features, Tableau’s Ask Data, and Snowflake’s Cortex all offer Chat with Data capabilities built directly into the BI platform. For organisations already invested in these platforms, these native capabilities are often the fastest route to Chat with Data deployment because they connect directly to existing semantic models and data sources.
The trade-off of native platform capabilities is scope: they are typically limited to the data connected within that platform. An organisation using Power BI connected to a data warehouse that covers finance and sales can answer finance and sales questions through Power BI Copilot, but not questions that require data from a separate HR system that is not part of that model.
Surfaced Within ERP and Operational Systems
An emerging integration pattern brings Chat with Data directly into ERP and operational platforms: SAP, Oracle, and Microsoft Dynamics 365 all have active AI assistant capabilities that allow users to ask data questions within the system they use for operational work. A procurement manager in SAP can ask about supplier performance without leaving the procurement module; a financial controller in Dynamics can query period-close status without switching to a separate reporting tool.
This embedded-in-operations model reduces the context switching that undermines adoption of separate analytics tools. When data access is available within the system of action rather than requiring a switch to a system of record, usage rates are consistently higher.
API-Driven Integration for Custom Applications
For organisations with custom internal applications, customer-facing portals, or industry-specific software, API-driven Chat with Data integration allows conversational data access to be embedded within those applications directly. A retail company might embed Chat with Data in its store operations portal so that store managers can ask inventory and performance questions without needing a separate BI platform. A financial services firm might embed it in a client reporting portal so that relationship managers can answer client questions on the spot.
API-driven integration requires more development effort than native platform integrations but offers the highest degree of control over the interface, the governance model, and the scope of questions the system will answer.
The Governance Shift: From Access Control to Answer Control
As Chat with Data matures, the governance conversation is evolving. The first generation of governance concern was access control: ensuring that users can only ask about data they are authorised to see. This is now relatively well-solved through role-based security enforced at the semantic or data layer.
The emerging governance challenge is answer control: ensuring that the AI-generated answer is accurate, appropriately caveated, and does not overstate certainty about results that are based on incomplete or ambiguous data. This is particularly important for organisations using Chat with Data in regulated contexts, where the accuracy of a data answer has compliance implications, or in customer-facing applications, where an incorrect answer reaches an external audience.
Best practice in 2026 includes: surfacing the underlying data source and query logic alongside every answer so users can verify the basis; flagging answers based on incomplete data rather than silently returning a partial result; maintaining audit logs of every question and answer for compliance review; and setting clear scope boundaries so the system does not attempt to answer questions outside its defined data domain.
What the Near Future Looks Like for Chat with Data
Personalised Data Experiences
The next generation of Chat with Data will learn user preferences and context over time. A CFO who consistently asks about margin and cash position will receive answers that prioritise those dimensions. A regional sales manager will receive answers contextualised for their geography without specifying it each time. This personalisation is not about storing personal data inappropriately but about the system learning the user’s role, priorities, and frequent questions to make every interaction more efficient.
Predictive and Prescriptive Answers
Current Chat with Data primarily answers descriptive and diagnostic questions: what happened, and why. The near-term development is predictive and prescriptive capability: not just ‘what happened to our churn rate last month?’ but ‘which customers are most likely to churn in the next 30 days, and what action is most likely to retain them?’
This requires the integration of predictive models with the Chat with Data layer, so that natural language questions can trigger not just data retrieval but model inference. Organisations that have built predictive models on their data will be able to expose those models through a conversational interface, making their predictive capability accessible to a much wider audience of business users than the analysts who built the models.
Cross-System Data Conversations
Today, most Chat with Data implementations are scoped to a single data platform or a well-defined subset of the organisation’s data. A question that requires joining data from the CRM, the ERP, and the HR system in a single answer is typically not yet possible through a single conversational interface.
Data fabric and data mesh architectures are moving this frontier by creating federated semantic layers that span multiple data domains and sources. As these architectures mature, Chat with Data will be able to answer questions that cut across the full breadth of an organisation’s data estate, making the conversational interface a genuinely unified entry point to enterprise intelligence.
AI Agents That Take Action, Not Just Answer Questions
The most forward-looking development in Chat with Data is the transition from systems that inform to systems that act. An AI agent that detects a supply risk, surfaces the insight to the procurement team, and, with appropriate authorisation, initiates a supplier contact or a purchase order adjustment is not merely a Chat with Data system: it is an agentic analytics system that closes the loop between insight and action.
This capability is in early production in a small number of enterprise environments in 2026. It represents a fundamental change in what business intelligence means: from a function that produces reports and answers questions to one that participates directly in operational decision-making. The governance requirements, particularly around authorisation, auditability, and human oversight of AI-initiated actions, are correspondingly more demanding and will be a defining focus of enterprise AI deployments through 2027 and 2028.
How Data Semantics Delivers Chat with Data
Data Semantics delivers Chat with Data through a conversational analytics capability built on the client’s existing data infrastructure, designed to integrate with the platforms and channels where their teams already work.
- Semantic model development: Every Chat with Data implementation is underpinned by a business-aligned semantic model that ensures natural language questions return answers consistent with how the organisation defines and measures its performance.
- Platform integration: Data Semantics layers Chat with Data capabilities onto Microsoft Power BI, Azure Synapse, Snowflake, and Databricks, leveraging existing data investments rather than requiring a parallel data infrastructure.
- Collaboration channel embedding: Chat with Data is deployed within Microsoft Teams, WhatsApp, or custom portals, so users can ask data questions within the flow of their work.
- Governance and compliance: Role-based access control, audit logging, and ISO 27001 and SOC 2 Type 2-compliant practices ensure that Chat with Data operates within the organisation’s security and compliance framework.
- Ongoing model refinement: As business definitions evolve and new data domains come into scope, Data Semantics provides continuous semantic model updates and model retraining to keep the Chat with Data capability current.
Explore how Data Semantics delivers Chat with Data for enterprise teams
Conclusion
Chat with Data is not a finished product in 2026. It is a rapidly evolving capability with a clear trajectory: toward more context-awareness through multi-turn dialogue, more proactivity through agentic AI, broader reach through deeper integration, and ultimately toward systems that not only answer questions but participate in the decisions those questions inform.
The organisations that are building their Chat with Data capabilities now, with the right semantic foundations, governance models, and integration patterns, are positioning themselves to benefit from each successive capability improvement as it becomes available. The data infrastructure decisions made today determine how quickly the next generation of Chat with Data capability can be deployed.
Connect with Data Semantics to discuss how Chat with Data can be deployed within your analytics environment.
Frequently Asked Questions
How is Chat with Data different from the Conversational BI concept?
Conversational BI is the broader capability: the technology, architecture, and practice of using natural language to interact with business intelligence systems. Chat with Data is a specific implementation of Conversational BI, typically referring to a conversational layer connected to an organisation’s data warehouse or analytics platform. The two terms are closely related; ‘Chat with Data’ tends to emphasise the interaction model (a dialogue with data), while ‘Conversational BI’ tends to emphasise the broader category of analytics technology. See our Conversational BI complete guide for the foundational coverage.
What data sources can Chat with Data connect to?
Chat with Data implementations can connect to any structured data source that has been modelled and documented in a semantic layer: cloud data warehouses (Snowflake, BigQuery, Azure Synapse, Redshift), on-premise databases, ERP systems via data connectors, and third-party SaaS platforms where data can be extracted into the data model. The scope of questions a Chat with Data system can answer is determined by the scope of data connected to its semantic model, not by any inherent limitation of the natural language interface.
What is agentic analytics and how is it different from Chat with Data?
Chat with Data is reactive: the user asks a question and the system answers it. Agentic analytics is proactive: the AI monitors data continuously, identifies patterns or anomalies that meet defined criteria, and surfaces insights to relevant users without waiting to be asked. Agentic analytics is an emerging capability that extends Chat with Data by adding autonomous monitoring and proactive notification. The most advanced implementations in 2026 are beginning to go further still: not just notifying users of an insight but recommending or initiating a response action, with appropriate authorisation controls.
How does narrative AI improve Chat with Data output?
Standard data query output is a table or chart. Narrative AI generates a plain-language explanation of what the data means alongside the visualisation: what drove the result, how it compares to expectations, and what it implies for a decision. This explanation makes data accessible to business users who may not interpret a visualisation instinctively and significantly reduces the analytical burden placed on the person receiving the answer. Narrative AI is powered by LLMs prompted with the retrieved data and the context of the question, producing commentary that reflects the actual data rather than generic observations.
