Enterprise Software Watch: AI Agents Begin Challenging the SaaS Model

Watch Tech Series · Enterprise Software Watch | September 24, 2026

The latest enterprise software developments as AI agents reshape SaaS, enterprise automation, data platforms, governance, and software economics.

EXECUTIVE BRIEF

Enterprise software is moving from software that employees use toward AI systems that perform work across software.

Six developments today — a $77M raise for AI “employees,” a new AI harness from Salesforce, confidential AI from VAST Data, two acquisitions, and a $100M vertical AI round — point to the same structural shift. This edition covers what happened and what it means for the enterprise software stack. It is a companion to today’s AI Watch and AI Infrastructure Watch, keeping the focus on the software business model and the enterprise application layer.

1. Ema’s $77M Round Puts the SaaS Model Under Pressure

Source: TechCrunch

Ema raised $77 million in Series B funding, bringing total funding to $140 million. Its platform uses teams of AI agents to automate processes across HR, IT, and finance. The company says it has more than 50 active enterprise deals and over 1 million active enterprise users.

The round lands in a category that has quietly become crowded: AI agents performing work traditionally handled by SaaS applications. Ema’s model is built around what it calls “AI employees” — agent teams that execute processes rather than provide tools for humans to execute them.

The pricing model is where the disruption becomes visible. Ema charges on tasks and outcomes rather than seats. That is a structural departure from the per-user pricing that has defined enterprise software economics for two decades. If agents perform work, the number of human seats becomes a less meaningful unit of value.

The convergence at play is between enterprise software and IT services. A traditional SaaS platform sells the tool. A traditional services firm sells the outcome. Ema is attempting to sell both through the same vehicle — and its pricing reflects the outcome rather than the tool.

What It Means: The important question is not whether Ema becomes another SaaS vendor. It is whether AI agents change what companies actually buy from enterprise software vendors. Seat-based pricing, per-module packaging, and the application-as-tool model all assume that humans are the ones performing work. When agents perform the work, the unit of value shifts. Incumbents that cannot reprice around outcomes will find that their revenue model is misaligned with the value their customers are buying.

The CODEW Lens: Ema is not a SaaS company with AI features. It is a test case for whether the SaaS business model survives agents.

2. Salesforce’s “Enterprise AI Harness” Points to a New Application Layer

Source: Salesforce

Salesforce introduced its Trusted Enterprise AI Harness, an architecture designed to give AI agents business context, reasoning and planning capabilities, the ability to take actions across systems, and enterprise controls.

The architecture addresses a problem that has become visible as agents move from demos into production: agents that operate inside a single application are limited. Agents that operate across multiple enterprise applications need a layer that provides three things the applications themselves do not expose cleanly.

Layer	What It Provides
Business context	The meaning of enterprise data — customers, deals, accounts, policies — beyond raw records.
Reasoning & planning	The ability to break a goal into steps and decide what to do next.
Action across systems	The capability to execute tasks in applications other than the one the agent lives in.
Enterprise controls	Governance, permissions, auditability, and escalation.

Business context is the interesting piece. In the SaaS era, the application itself encoded the context: the CRM knew what a customer was, the ERP knew what a supplier was. In the agent era, that context has to be extracted, structured, and made available to systems that are not the original application.

What It Means: Enterprise application vendors increasingly need to provide more than applications. They need to provide the context, permissions, and orchestration layer through which agents operate. Salesforce’s harness is an attempt to claim that layer before a competitor does — or before the model providers themselves move up the stack. If the future enterprise stack has an “AI harness” between models and applications, whoever owns that layer owns the interface between intent and execution.

The CODEW Lens: The application is no longer the end product. It is one layer in a stack where context, control, and orchestration determine who owns the customer.

3. Enterprise Data Becomes the Battleground

Source: GlobeNewswire

VAST Data announced DataEnclave, a confidential AI runtime that allows leading AI models to operate against sensitive enterprise data using confidential computing. The company says the system can run inside customer data centers or trusted cloud hardware while giving organizations greater control over data, model selection, and deployment.

The announcement addresses one of the more persistent limits on enterprise AI adoption: the data that would make AI most valuable is often the data that is hardest to move. Financial records, health records, legal documents, and customer data subject to regulatory constraints cannot simply be sent to a public model API. But running models inside the enterprise has historically meant giving up access to the frontier models that produce the best results.

DataEnclave is a software-layer answer to that problem. By running within a confidential computing environment — where neither the infrastructure operator nor the model provider can access the underlying data — it separates the question of where the model runs from the question of who controls the data. The enterprise keeps the data, and the model provider keeps the weights.

What It Means: Enterprise AI adoption increasingly depends on solving a basic software problem: how do you let AI work with valuable corporate data without surrendering control of that data? This is a data-platform problem before it is a model problem. Data platforms and AI application infrastructure are converging — the same software stack increasingly has to handle storage, governance, model access, and inference environment as one coherent system. Vendors that can solve the sovereignty problem will have a structural advantage in regulated industries.

The CODEW Lens: Confidential AI is not a feature. It is a prerequisite for the enterprise AI market that does not yet exist at scale.

4. Progress Completes Domo Acquisition

Source: The Manila Times

Progress Software completed its acquisition of Domo’s AI and data platform business, saying the deal advances its strategy around trusted AI and agentic outcomes.

The deal fits a pattern that has been building through 2026: enterprise software vendors buying data and analytics capability rather than building it. Progress has historically been a portfolio company — infrastructure, development tools, and application experience. Domo adds a data platform that can serve as the context layer for AI features across the rest of the portfolio.

Analytics vendors were once a distinct category. As AI agents began operating across enterprise applications, the value of being the place where enterprise data is modeled and governed increased. Context is what makes an agent useful; context lives in the data layer. That is why data platforms have become acquisition targets rather than adjacent businesses.

What It Means: The enterprise software M&A story is increasingly about acquiring the data and context required to make AI useful. Standalone data platforms are becoming strategically more important, not less — because the value of an AI agent depends on the quality of the data it can see. Expect more acquisitions of data platforms by application vendors, and more application vendors trying to argue that their own data is a differentiator.

The CODEW Lens: Data platforms are no longer infrastructure. They are the substrate on which AI features and agent workflows run.

5. Nutanix Acquires Ryax for Agentic AI Infrastructure

Source: HPCwire

Nutanix agreed to acquire Ryax Technologies, whose platform provides scheduling and performance optimization for AI workloads. Nutanix plans to integrate the technology into its Kubernetes platform and broader AI strategy.

Nutanix built its business on hyperconverged infrastructure — the idea that compute, storage, and networking should be managed as a single software-defined system. The company’s expansion into Kubernetes was a natural extension of that thesis. The Ryax acquisition extends it further into AI workload management.

Agentic AI creates a different kind of workload pattern than the batch-oriented or request-response workloads that enterprise infrastructure has historically optimized for. Agents spawn sub-tasks, call external tools, wait on responses, and resume — often in patterns that are neither predictable nor steady-state. Scheduling and performance optimization for that workload profile is not a solved problem.

What It Means: Enterprise software boundaries are becoming less clear. The application layer increasingly depends on AI infrastructure capabilities underneath it — orchestration, scheduling, cost governance, and performance management for workloads that behave differently than traditional enterprise software. Vendors that were once strictly infrastructure companies are moving up, and application companies are moving down. The teams that will win are those that can present a coherent stack across both.

The CODEW Lens: Agentic AI is not just a new application pattern. It is a new infrastructure workload, and the software that manages it is part of the enterprise application stack whether vendors call it that or not.

6. Numeral’s $100M Round Shows AI Moving Into Specialized Enterprise Workflows

Source: Numeral

Numeral raised $100 million in Series C funding for its AI-powered sales-tax compliance platform. The company plans to expand into software, manufacturing, distribution, and wholesale.

Sales-tax compliance is exactly the kind of workflow that AI agents are well-suited to attack. It is rules-based but jurisdictionally complex, high-volume but structured, high-stakes but not strategic, and constantly changing. It has historically been handled by a mix of internal staff, external advisors, and specialist software — all of which are expensive relative to the value the workflow itself produces.

Numeral is a vertical AI company, not a horizontal platform. The distinction matters for how the market is likely to evolve. General-purpose AI vendors are competing on breadth — the ability to handle many workflows across many industries. Vertical AI vendors compete on depth — the ability to handle one workflow better than anyone else, with domain knowledge encoded into the product and into the training data.

What It Means: AI doesn’t have to replace the entire ERP or CRM. It can first attack high-value, highly structured workflows inside them. That is a lower-risk entry point for enterprises than replacing a system of record, and a faster path to measurable ROI. Expect the pattern to repeat across tax, compliance, procurement, contract review, and other workflows where the rules are complex, but the outcome is well-defined.

The CODEW Lens: Vertical AI is not a niche. It is the most realistic path to enterprise AI ROI in the next 18–24 months — because it sells an outcome customers already understand how to buy.

The Enterprise Software Shift

Enterprise software is entering a transition from systems of record to systems of action.

Traditional SaaS applications primarily store information, provide workflows, and give employees tools to perform work. AI agents increasingly sit across those systems, interpret business context, make decisions, and execute tasks.

That creates a new competitive layer: the software that controls the agent, the data, the workflow, and the permissions may become as strategically important as the application itself.

Today’s six developments each touch a different part of this shift:

Development	Layer of the Stack
Ema ($77M)	Agent execution and the business model for it
Salesforce AI Harness	Context, orchestration, and controls
VAST DataEnclave	Data sovereignty and confidential AI runtime
Progress / Domo	The data and analytics substrate
Nutanix / Ryax	Agentic workload infrastructure
Numeral ($100M)	Vertical workflow automation

The CODEW Lens: The enterprise software market is not being replaced by AI. It is being reorganized around a new stack in which applications, context, orchestration, data, and infrastructure all have to be priced and packaged differently than they were before.

Sources

→ TechCrunch — Ema’s $77M funding and AI employees
→ Salesforce — Trusted Enterprise AI Harness
→ GlobeNewswire — VAST Data DataEnclave
→ The Manila Times — Progress Software / Domo acquisition
→ HPCwire — Nutanix / Ryax acquisition
→ Numeral — $100M Series C

Next in Enterprise Software Watch

→ Agentic Pricing: How AI Changes What Enterprises Buy
→ The AI Harness Layer: Who Owns Context in the Enterprise Stack?
→ Vertical AI vs. Horizontal Platforms: Where the ROI Actually Lands

The CODEW Stat

6 developments · $177M in disclosed rounds · 2 acquisitions · 5 layers of the enterprise stack Today’s enterprise software developments touched every layer of the emerging agent-era stack: the business model for agent execution (Ema, $77M), the context and orchestration layer (Salesforce AI Harness), the data sovereignty and confidential AI runtime (VAST DataEnclave), the data and analytics substrate (Progress / Domo), agentic workload infrastructure (Nutanix / Ryax), and vertical workflow automation (Numeral, $100M). Together they point to a single structural shift: enterprise software is moving from systems of record to systems of action, and the competitive layer is shifting from the application to the software that controls the agent, the data, the workflow, and the permissions.

THE CODEW · ENTERPRISE SOFTWARE WATCH

Editorial Note

The Enterprise Software Watch examines the developments reshaping enterprise software and AI agent platforms, including agentic architectures, SaaS pricing models, AI control planes, enterprise data and governance, non-human identity security, M&A across the agentic stack, and the competitive dynamics among software companies.

Educational content only. Not investment or business advice. Analysis is based on company announcements, official product disclosures, investor relations releases, and reporting from TechCrunch, Salesforce, GlobeNewswire, The Manila Times, HPCwire, and Numeral. Metrics referenced are labeled as reported, calculated, or CODEW-derived. Some products referenced may be affiliate partners — see our Affiliate Disclosure for full details. Platform coverage, data sources, and methodologies can change as the intelligence platform evolves.