April, May 2026 | The Astute Update
- Astute Dimension
- 1 day ago
- 10 min read
News in Brief
The AI race has moved to business capability at scale
OpenAI Frontiers framed the current phase of AI adoption around real-world deployment, responsible enterprise rollout, secure operating models, and AI becoming part of how companies operate at scale. [14]
SAP’s planned acquisition of Dremio is an interesting move in the race to provision enterprise AI capabilities, giving SAP a stronger foundation for governed context across SAP and non-SAP data. [1]
Oracle is pushing the control layer closer to enterprise data, with identity-aware access and runtime context for users and agents. [3]
Jedox 26 strengthens the planning layer with explainable KPI insights, natural-language analysis, a next-generation Integrator, native Python support, improved JSON handling, and self-service SAML. [5]
Finance leaders are moving beyond AI experimentation. KPMG’s 2026 finance survey found that active AI use across finance has more than doubled in two years, with the strongest gains in decision quality, forecast accuracy, and responsiveness. [7]
The theme from the major providers in April and May is capability at scale. We see the big players attempt to answer the question how can AI that work inside business processes, respect access rights, and use trusted data.
Main Story
The race is now business capability at scale
The enterprise AI conversation is maturing.
For the past two years, much of the market has focused on model capability, pilot deployments, experimentation, and productivity gains. That phase is not over, but the centre of gravity is shifting. The real race is now about turning AI into consistent business capability at scale.
OpenAI Frontiers made this shift clear. Across events in San Francisco and London, OpenAI brought together leaders to discuss how organisations are putting AI to work across industries. The themes were practical: moving from experimentation to real-world deployment, rethinking products and workflows, improving customer experiences, and bringing AI into organisations responsibly, securely, and at scale. [14]
That is the key signal. AI adoption is no longer just a technology question. It is becoming an operating-model question.
Enterprises are now asking how AI fits into work, how it connects to company data, how it supports teams, how it scales securely, and how leaders manage the change. OpenAI’s business platform messaging reinforces this direction. It positions frontier models, ChatGPT for Business and Enterprise, specialised agents, API capabilities, company-data integrations, admin controls, SAML SSO, enterprise-grade security, compliance support, and privacy protections as part of the enterprise AI stack. [15]
This matters because frontier AI capability alone is not enough. To become operational capability, AI needs the platform layers that determine which data it can use, which business context it understands, which permissions it inherits, and which controls apply when it acts.
That brings the discussion back to SAP and Oracle.
SAP is putting business context at the centre of enterprise AI. Its planned acquisition of Dremio is one of the clearest signals that enterprise AI is moving toward open, governed, semantically rich data foundations. SAP said Business Data Cloud will become an Apache Iceberg-native enterprise lakehouse, with SAP and non-SAP data coexisting on an open foundation. SAP also said Dremio’s catalog will serve as a discovery and semantic layer, giving connected engines access to business context, access rights, and lineage. [1]
At SAP Sapphire, SAP also framed enterprise AI as a context and execution problem. SAP stated that its future AI agents need to understand business data, end-to-end processes, security, compliance, and governance frameworks, and verify identity and access authorisations before returning outputs. [2]
Oracle is approaching the same challenge from the control side.
Its Deep Data Security capability for Oracle AI Database 26ai is designed to enforce fine-grained access policies directly during SQL execution, so only authorised rows, columns, and cell values are returned. Oracle also describes identity and context-aware access, identity propagation, dynamic masking, controlled privilege elevation, and centralised auditing. [3]
Oracle’s AI Data Platform Workbench updates add another important layer. The April 2026 update describes session attributes that allow a user’s identity to be passed from a front-end application into the agent and LLM context. This helps enforce row-level and column-level security when agents query data. [4]
In simple terms, both SAP and Oracle are building capabilities that help AI agents understand the data permissions of the users they are acting on behalf of.
That is a critical foundation for enterprise AI.
Permission-aware context enables AI systems to retrieve relevant information without exposing sensitive data that the user should not see. It allows different people to ask the same question and receive answers based on their legitimate access. It supports sensitive-data masking, user-level audit trails, controlled agent behaviour, and clearer accountability for decisions and recommendations.
Without this discipline, AI can scale risk as quickly as it scales productivity. An agent may produce a useful answer, but from data the user should never have accessed. With permission-aware context, AI becomes more useful because it becomes more bounded.
This is where the enterprise AI market is heading.
OpenAI is helping define the frontier of what AI can become inside organisations. SAP is investing in the data and semantic layer needed to ground that capability in enterprise context. Oracle is strengthening the permission and enforcement layer needed to control how AI interacts with enterprise data.
The next phase of enterprise AI will not be won by model capability alone. It will be won by organisations that can connect AI to trusted data, business meaning, user permissions, governance controls, and real workflows.
The race is no longer just about who has the most powerful AI.
It is about who can turn AI into safe, useful, permission-aware business capability at scale.
Jedox Update
This quarter has seen the release of Jedox 26. This release adds to Jedox capabilities providing capabilities finance teams can use today. JedoxAI’s Data Insights feature supports guided KPI analysis, helping users compare numbers, identify trends, surface drivers, break down performance across dimensions, and turn insights into presentation-ready outputs. The release also expands natural-language interaction, making it easier for users to create and refine Views and explore data. [5]
The Integrator update is also important. Jedox describes a next-generation Integrator with a more modern interface, native Python support, improved JSON handling, and AI-assisted guidance for troubleshooting, templates, and contextual support. [5]
That is not just a technical upgrade. In planning and EPM, integration is where a lot of trust is won or lost. If data flows are fragile, delayed, or poorly understood, AI-assisted analysis will inherit that weakness. Better integration capability gives planning teams a stronger foundation for using AI in a way that is connected to business reality.
Jedox 26 also adds self-service SAML configuration, helping administrators manage authentication more independently and improve control across the platform. [6]
The bigger message is that AI in planning becomes more useful when it is grounded in structured performance data, business logic, and explainable workflows. That is a good position for Jedox. Its role is not to be a generic AI layer. Its role is to help finance and planning teams turn trusted business data into better forecasts, analysis, and decisions.
Other News
Regulators are moving from principle to operating discipline
APRA’s April letter to industry is a useful Australian reference point. APRA observed that AI adoption is moving quickly, but governance, assurance, and operational controls are not keeping pace. It highlighted risks around prompt injection, data leakage, insecure integrations, exploit injection, autonomous agents, weak preventative controls, and insufficient technical restrictions. [12]
APRA also called out privileged access management, controls over agentic and autonomous workflows, supplier visibility, third-party and fourth-party dependencies, and continuous monitoring across the AI lifecycle. [12]
The Bank of England, FCA, and HM Treasury also issued a May statement on frontier AI models and cyber resilience, reinforcing the link between AI adoption, cyber risk, third-party oversight, and operational resilience. [13]
For boards and executives, the direction is practical. AI governance is no longer a policy document sitting beside the platform. It needs to be designed into access, data, workflow, assurance, and operating controls.
Finance AI is shifting from efficiency to decision quality
KPMG’s 2026 Global AI in Finance report found that active AI use across the finance function has moved from 30% to 75% in two years, with 76% of organisations actively leveraging AI in financial planning. The report also found that 70% reported improved decision-making quality, with gains clustering in judgment-heavy areas such as planning, forecasting, and risk assessment. [7]
KPMG also found that organisations with stronger governance and controls report significantly better AI outcomes, in some cases three to six times the rate of significant improvement compared with those without. [7]
That aligns closely with the OpenAI, SAP, Oracle, Workday, OneStream, and Jedox stories. AI performance is increasingly tied to context, control, and trust.
Workday extends agents while keeping policies and controls in view
Workday’s May announcements show the same market pattern in a different domain. Workday made its Sana Self-Service Agent available in Microsoft 365 Copilot, allowing HR and finance tasks to be completed from within Microsoft 365 while Workday maintains policies, approvals, business rules, and role-based permissions. [8]
Workday also announced Sana for IT Service Management and a Travel Agent, stating that these agents follow the same security, controls, and governance models customers already rely on, and are connected to Workday’s data context. [9]
OneStream sharpens its Finance AI position
OneStream completed its acquisition by Hg in April, in an all-cash transaction valuing the company at approximately US$6.4 billion. The company linked the move to longer-term investment in Finance AI and enterprise finance management. [10]
At Splash 2026, OneStream announced a Finance Agentic Layer built around a governed entry point for agentic requests, identity and access resolution, rate limiting, semantic grounding, tools, skills, and native finance agents. [11]
OneStream also published survey findings from more than 350 senior Finance and IT executives, reporting that 47% had made a material business decision based on inaccurate, incomplete, or outdated financial data in the previous 12 months. As vendor-backed research, it should be read as directional. Still, it reinforces the same point: AI will amplify data quality and governance problems if the underlying foundation is weak. [11]
Closing Note
OpenAI Frontiers captured the shift from experimentation to real-world deployment. SAP and Oracle showed the kind of enterprise platform work needed underneath that shift: governed context, data access, identity, permissions, and control.
Jedox continues to strengthen the planning layer where trusted performance data becomes business decisions.
For Finance, EPM, and transformation teams, the next wave of AI value will come from platforms that can turn capability into controlled execution. Faster answers are helpful. Governed, explainable, decision-ready capability is far more valuable.
Thanks for Reading!
We trust this edition kept you informed with the latest developments, from technology updates and platform enhancements to independent analyst recognition and planning insights.
Warm regards,
Astute Dimension
Helping you translate planning into performance.
References
[1] SAP News Center — “SAP to Acquire Dremio to Unify SAP and Non-SAP Data to Power Agentic AI”, 4 May 2026.Used for SAP’s planned Dremio acquisition, SAP Business Data Cloud positioning, Apache Iceberg-native lakehouse direction, open catalog and semantic layer, access rights, data lineage, and knowledge graph context.https://news.sap.com/2026/05/sap-to-acquire-dremio-unify-sap-and-non-sap-data-power-agentic-ai/
[2] SAP News Center — “SAP Unveils Business AI Platform to Power the Autonomous Enterprise”, 13 May 2026.Used for SAP Sapphire context, SAP’s framing of enterprise AI as a context and execution problem, and SAP’s statement that agents need to verify identity and access authorisations before returning outputs.https://news.sap.com/2026/05/sap-sapphire-keynote-business-ai-platform-power-autonomous-enterprise/
[3] Oracle — “Safely unleash AI on enterprise data with Deep Data Security”, Oracle AI Database 26ai.Used for Oracle Deep Data Security, database-enforced authorisation, row-, column-, and cell-level security, identity and context-aware access, identity propagation, dynamic masking, controlled privilege elevation, and auditing.https://www.oracle.com/security/database-security/features/deep-data-security/
[4] Oracle AI Data Platform Blog — “What’s New in Oracle AI Data Platform April 2026”.Used for Oracle AI Data Platform Workbench updates, session attributes, identity passed into agent context, and row-level and column-level security enforcement when agents query data.https://blogs.oracle.com/ai-data-platform/whats-new-in-oracle-ai-data-platform-april-2026
[5] Jedox — “Jedox 26 Major Release | AI-Powered Business Planning”.Used for JedoxAI, Data Insights, natural-language analysis, KPI explanation, next-generation Integrator, native Python support, improved JSON handling, AI-assisted Integrator support, and self-service SAML.https://www.jedox.com/en/blog/jedox-26-major-release/
[6] Jedox — “Jedox 26: AI for FP&A, Planning, and Data Integration”.Used for Jedox 26 feature details including Data Insights, Reporting Agent, presentation-ready outputs, Integrator enhancements, Cube Delta Extract, native Python support, improved JSON handling, self-service SAML, and Microsoft 365 add-in improvements.https://www.jedox.com/en/services/whats-new/
[7] KPMG — “Global AI in Finance 2026: The Decision Advantage”.Used for KPMG’s survey of 1,013 senior finance leaders, AI adoption across finance, 76% active AI use in financial planning, 70% improved decision-making quality, decision-heavy use cases, forecast accuracy, governance and controls, and data quality findings.https://assets.kpmg.com/content/dam/kpmgsites/uk/pdf/2026/05/ai-in-finance-report.pdf
[8] Workday Newsroom — “Workday Brings Sana Self-Service Agent for HR and Finance Into Microsoft 365 Copilot”, 13 May 2026.Used for Workday Sana in Microsoft 365 Copilot, HR and finance task automation, existing approvals, policies, business rules, role-based permissions, and Workday data remaining in the trusted system.https://newsroom.workday.com/2026-05-13-Workday-Brings-Sana-Self-Service-Agent-for-HR-and-Finance-Into-Microsoft-365-Copilot
[9] Workday Newsroom — “Workday Announces Sana for IT Service Management and New Travel Agent”, 21 May 2026.Used for Workday Sana for ITSM, Travel Agent, governance positioning, access changes, policies, controls, and Workday data context across HR, finance, and IT workflows.https://newsroom.workday.com/2026-05-21-Workday-Announces-Sana-for-IT-Service-Management-and-New-Travel-Agent
[10] OneStream — “OneStream Announces Completion of Acquisition by Hg for $6.4 Billion”, 1 April 2026.Used for OneStream’s acquisition by Hg, transaction value, finance management positioning, and OneStream’s stated direction on Finance AI.https://www.onestream.com/news/onestream-announces-completion-of-acquisition-by-hg-for-usd6-4-billion/
[11] OneStream — “What We Announced at Splash 2026: A New Foundation for Finance” and “Companies Are Scaling AI on Data They Don’t Trust, New Study Finds”.Used for OneStream’s Finance Agentic Layer, identity and access resolution, semantic grounding, and the survey of 350+ senior Finance and IT executives on data quality and AI risk.https://www.onestream.com/blog/what-we-announced-at-splash-2026-a-new-foundation-for-finance/
[12] APRA — “APRA Letter to Industry on Artificial Intelligence”, 30 April 2026.Used for Australian regulatory context on AI adoption, governance maturity, cyber risk, prompt injection, data leakage, agentic workflow controls, privileged access management, supplier visibility, assurance, and continuous monitoring.https://www.apra.gov.au/apra-letter-to-industry-on-artificial-intelligence-ai
[13] Bank of England, FCA, and HM Treasury — “Joint statement on Frontier AI models and cyber resilience”, May 2026.Used for UK regulatory context on frontier AI, cyber resilience, access management, third-party oversight, and operational risk.https://www.bankofengland.co.uk/news/2026/may/boe-fca-and-hm-treasury-joint-statement-on-frontier-ai-models-and-cyber-resilience
[14] OpenAI LinkedIn post supplied by user — OpenAI Frontiers, San Francisco, London, Tokyo, and Seoul, May 2026.Used for the OpenAI Frontiers framing: leaders discussing AI deployment across industries, movement from experimentation to real-world deployment, AI-enabled product, workflow and customer-experience redesign, responsible and secure organisational rollout, Sarah Friar and Sam Altman discussion, and expansion of the Frontiers series to Asia.
[15] OpenAI — “AI Platforms to Accelerate your Business” / OpenAI Business.Used for OpenAI’s business-platform positioning around frontier models, ChatGPT for Business and Enterprise, specialised agents, workspace agents, API platform, company-data integrations, enterprise-grade security, admin controls, SAML SSO, compliance, and data privacy.https://openai.com/business/