TL;DR
Senior executives at Dell, Microsoft, Salesforce, ServiceNow and Snowflake say 2026 will be the year organizations demand measurable returns from AI and stronger governance. Their forecasts emphasize on-premises control, agent safeguards, data governance and more efficient distributed compute.
What happened
Executives from major enterprise technology vendors issued forward-looking assessments for AI in the workplace for 2026, converging on a few core priorities: demonstrating return on investment and tightening governance. Dell CTO John Roese argued that AI deployments have been rushed into production without adequate policies and predicted a shift toward private, on-premises or controlled environments to preserve security, governance and cost control. ServiceNow’s Heath Ramsey urged companies to start with small, high-impact use cases that save time and money, backed by a single entry point for policies and approvals. Snowflake CISO Brad Jones highlighted the risk of improperly permissioned documents feeding generative tools, stressing data governance. Microsoft leaders framed trust, clear agent identities, limits on system access and protocols for data created by agents as essential, and forecast denser, distributed compute architectures—so-called linked superfactories—to lower costs and boost efficiency.
Why it matters
- Enterprises will increasingly demand measurable financial returns from AI investments rather than experimental pilots.
- Stronger data governance will be central to preventing accidental exposure of sensitive information to generative or agentic systems.
- On-premises and controlled AI environments could shift deployment models and procurement decisions for infrastructure.
- Changes in infrastructure design and resilience planning will affect hardware suppliers, cloud providers and data-protection vendors.
Key facts
- Dell CTO John Roese predicts a move toward running AI locally in on-premises or controlled ‘AI factories’ to improve control and stability.
- ServiceNow’s Heath Ramsey advises starting with small, end-to-end AI projects that address tasks 'bleeding time and money' and using a single policy entry point.
- Snowflake CISO Brad Jones warns that improperly permissioned documents or data sets could be exposed if fed to generative or agentic AI.
- Microsoft’s Vasu Jakkal says agents need clear identity, restricted system access, data-management protocols and protections akin to human users.
- Microsoft CTO Mark Russinovich forecasts denser distributed compute across linked AI systems to reduce costs and increase efficiency.
- Microsoft opened a linked AI supercluster in Wisconsin that spans 1.2 million square feet, uses thousands of Nvidia GB200 GPUs, is rated at 337 MW, and processes 865,000 tokens per second.
- Dell predicts AI will change resilience and disaster-recovery priorities, including protecting vectorized data and other AI-specific artifacts.
- Salesforce EVP Adam Evans predicts brands will be defined by their AI agents, which act as personalized, evolving brand ambassadors.
What to watch next
- Adoption rates of on-premises and private AI deployments versus public-cloud-first strategies.
- Emergence and adoption of standardized governance frameworks, policy gateways and approval workflows for AI agents.
- Expansion of linked AI superclusters or 'superfactories' and their reported impact on costs and throughput.
Quick glossary
- AI agent: A software entity that performs tasks or makes decisions with varying degrees of autonomy, often interacting with systems or users.
- On-premises (on-prem): Computing infrastructure located physically within an organization’s facilities, managed directly by that organization rather than by external cloud providers.
- Data governance: Policies, procedures and controls that determine how data is managed, accessed, protected and used across an organization.
- Vectorized data: Numerical representations of information (often text or images) used by machine learning models to perform similarity search, classification or other tasks.
- AI supercluster / superfactory: Large, highly interconnected deployments of compute and storage designed to run extensive AI workloads at scale.
Reader FAQ
Will AI reliably deliver ROI across enterprises in 2026?
Industry leaders say ROI is the defining question for 2026 and that measurable value is expected, but widespread, uniform ROI outcomes are not confirmed in the source.
Are companies moving their AI workloads back on-premises?
Dell’s CTO predicts a shift toward on-premises or controlled environments for greater governance and control, according to the source.
What is the biggest governance risk with generative AI and agents?
Snowflake’s CISO warns that improperly permissioned documents or data sets may be exposed if fed to generative or agentic systems.
Will larger, linked AI compute facilities reduce costs?
Microsoft leaders predict denser, distributed AI systems and linked superfactories will drive down costs and improve efficiency, per the source.
AI + ML 2 Tis the season when tech leaders rub their crystal balls 2026 is the year where AI must meet ROI in the enterprise, and the key to…
Sources
- Tis the season when tech leaders rub their crystal balls
- 9 AI Predictions Every Business Leader Should Know For …
- CFOs and other leaders share AI predictions for 2026
Related posts
- AI Employees Don’t Pay Taxes — How Automation Could Erode Public Revenue
- Why LLMs Are Not Fun: A Developer’s Case Against AI-as-Teammate
- 2025: When AI’s Frenzy Met a Hard Reality Check on Growth