Enterprise Software Watch | September 28, 2026
ERP turns agentic, CRM pricing fractures, workflow platforms drop the "sidecar" model, and Databricks goes shopping again — this week's enterprise software stack, top to bottom.
The editorial question behind this edition: is enterprise software moving from a system of record and a system of workflow toward a system of AI-driven execution?
Seven threads this week say yes, from different directions: SAP, Oracle and Microsoft embedding agents directly into ERP; Salesforce and HubSpot fracturing CRM pricing around outcomes instead of seats; ServiceNow declaring the "sidecar AI era" over; Salesforce building a context-and-controls layer for agents that cross applications; and Databricks going shopping again to fold spreadsheets into its AI coworker. This edition works down that stack — agents vs. SaaS, ERP, CRM, workflow, infrastructure, and the capital behind it — as a companion to today's AI Watch and The Term Sheet.
1. AI Agents vs. Traditional SaaS: The Pricing Model Is the Tell
Source: TechCrunch / SaaStr
The clearest version of the AI-agents-vs.-SaaS thesis is still Ema, the AI-employee startup whose $77 million raise this month was built explicitly around agents performing HR, IT, and finance work that enterprise software and IT services firms have historically billed for — priced on tasks and outcomes, not seats.
What's new this week is that the incumbents are visibly absorbing the same logic. HubSpot's Agent Hub now charges $0.50 per resolved conversation and $1.00 per qualified lead rather than a flat seat fee, while Salesforce is running at least three pricing models simultaneously — traditional per-seat licenses, Agentforce folded inside the seat as of a September 3 repricing, and its Fin-derived agent charging $0.99 per outcome. Neither company has committed fully to outcome pricing, because doing so risks cannibalizing seat revenue they still depend on.
What It Means: When AI-native challengers and seat-based incumbents both start pricing on outcomes in the same quarter, that's not a startup trend anymore — it's a market repricing itself around a different unit of value. The vendors best positioned aren't necessarily the cheapest; they're the ones that can run outcome and seat pricing side by side without the accounting getting confusing for buyers.
The CODEW Lens: Watch pricing pages, not press releases. Vendors will keep calling this "agentic AI." What they're actually renegotiating is what a customer pays for.
2. ERP: From System of Record to System of Action
Source: InfotechLead
SAP, Oracle, and Microsoft are all converging on the same architecture: agents that don't just answer questions about ERP data but monitor processes, flag exceptions, and execute transactions inside predefined permissions. Oracle's clearest move is 22 Fusion Agentic Applications spanning ERP, HCM, supply chain, and customer experience, operating natively inside Fusion Cloud rather than as a bolt-on chatbot. SAP's Joule is doing the equivalent inside S/4HANA — Bosch has reported a 20% developer productivity gain using it, while Microsoft is governing agentic workloads through Copilot Studio, with systems integrator Atos preparing to manage 19,000 AI agents across its Microsoft stack.
Gartner's numbers frame the pace: 62% of cloud ERP spending is projected to go toward AI-enabled solutions by 2027, up from just 14% in 2024, with finance organizations using AI-embedded cloud ERP expected to close their books 30% faster by 2028.
What It Means: ERP was the enterprise software category most insulated from agent disruption, because its data is too regulated and too load-bearing to hand to an unproven agent. That's exactly why the incumbents' agentic push matters more here than anywhere else — if SAP, Oracle, and Microsoft can make agentic ERP trustworthy at this scale, it closes the door on AI-native ERP challengers before they get a foothold.
The CODEW Lens: ERP is where "system of record becomes system of action" is easiest to prove and hardest to fake — which is why it's the best test of whether agentic enterprise software is real.
3. CRM: The Per-Seat Model Starts to Fracture
Source: SaaStr / Industry Pricing Trackers
CRM is the category where the pricing fracture from Section 1 is most visible, because sales and service software has run on per-seat economics for two decades. HubSpot's Agent Hub (renamed from Breeze) now bills $0.50 per resolved conversation and $1.00 per recommended lead, layered on top of its existing seat tiers rather than replacing them. Salesforce's own agent economics moved again on September 3, folding Agentforce inside the standard seat for some editions while still running Fin's per-outcome pricing elsewhere in the portfolio — three pricing models operating inside one company.
Microsoft Dynamics 365 has mostly held to tenant- and seat-based licensing for now, which makes it the control case worth watching: if Dynamics starts moving toward outcome pricing too, it confirms this isn't a Salesforce-versus-HubSpot fight but an industry-wide repricing.
What It Means: No CRM vendor has fully committed to outcome-based pricing, and that hesitation is the story. Seats remain the predictable, budgetable unit that CFOs want; outcomes are the unit that actually reflects what agents deliver. Vendors running both at once are hedging against being wrong about which model wins.
The CODEW Lens: A vendor running three pricing models isn't confused. It's buying time to see which one the market actually pays for.
4. Workflow & Automation: ServiceNow Ends the "Sidecar" Era
Source: ServiceNow Newsroom
ServiceNow has positioned its entire product portfolio as AI-native rather than treating AI as a separately purchased add-on — what the company calls moving past the "sidecar AI era," where intelligence gets bolted onto software after the fact instead of built into it. The centerpiece is its Context Engine, which grounds agent decisions in an organization's actual policies, relationships, and decision history, drawing on a claimed 85 billion workflows and 7 trillion transactions run through the platform, plus data folded in from ServiceNow's Traceloop, Veza, Pyramid Analytics, and data.world acquisitions.
The framing matters as much as the feature: ServiceNow's argument is that fragmentation — hundreds of applications per enterprise, each with its own data model and security perimeter — is the actual blocker to agentic workflows, not a lack of AI horsepower.
What It Means: Workflow and automation platforms are converging on a shared conclusion with ERP vendors: the constraint on agentic AI isn't model capability; it's enterprise context and governance. Whoever owns the layer that supplies that context to agents — not whoever has the flashiest agent — controls the workflow layer going forward.
The CODEW Lens: "No more sidecar AI" is a positioning statement dressed as a product launch. What it really signals is that AI stopped being a line item and became a precondition for renewal.
5. Enterprise AI Infrastructure: Who Owns the Context Layer
Source: Salesforce
Salesforce's Trusted Enterprise AI Harness, introduced this month, is the clearest articulation yet of what sits underneath agentic enterprise software: business context (what a customer, deal, or policy actually means), reasoning and planning (breaking a goal into steps), action across systems (executing tasks outside the agent's home application), and enterprise controls (governance, permissions, auditability). ServiceNow's Context Engine and Oracle's process-native Fusion agents are effectively building the same four layers under different names.
That convergence is the real signal this week: three major vendors independently arrived at nearly identical architecture for what an agent needs before it can be trusted with real enterprise work.
What It Means: A shared architecture across competitors is usually a sign a category is maturing past marketing and into infrastructure. The vendor that supplies the best version of the context-and-controls layer — not necessarily the best underlying model — is positioned to become the default substrate other companies' agents have to plug into.
The CODEW Lens: The application used to be the product. Increasingly, the context layer underneath it is the actual moat.
6. M&A & Funding: Databricks Keeps Shopping
Source: TechCrunch
Databricks acquired Row Zero, a Seattle spreadsheet startup that scales to roughly a billion rows and connects directly to governed enterprise data, on September 24 — terms undisclosed. The rationale, per CEO Ali Ghodsi, was that Databricks' own finance team had already paired Row Zero with Genie, the company's AI coworker, to avoid the ungoverned "spreadmart" files that create security gaps once agents start touching sensitive company data. Row Zero will give Genie a native, governed spreadsheet interface across web, desktop, and mobile.
The deal lands weeks after Databricks closed a $5 billion round in August at a $190 billion valuation, up from $134 billion just six months earlier, and follows a string of 2026 acquisitions including AI security firm Panther and agent-platform team Electric. Ema's $77 million round, covered in Section 1, is the venture-side mirror of the same trend: capital is flowing to both AI-native challengers attacking enterprise software categories and incumbents buying the pieces they need to defend them.
What It Means: Databricks isn't just buying a spreadsheet tool — it's buying a familiar interface for a governance problem AI agents are about to make much worse. Expect more of these small, interface-layer acquisitions as data platforms race to give business users a safe way to work with AI-touched data without exporting it into ungoverned files.
The CODEW Lens: When a company explicitly says it's "scouting for more startups to acquire" right after closing a deal, treat this week's acquisition as the first of several, not the last.
7. What to Watch
- Whether Microsoft Dynamics follows HubSpot and Salesforce into outcome-based pricing, confirming an industry-wide shift rather than a two-vendor fight.
- Whether ERP agentic adoption actually hits Gartner's 62%-by-2027 pace, or stalls the way broader enterprise agent pilots have (see this week's AI Watch on the 88% pilot-to-production gap).
- Which vendor's context-and-controls layer (Salesforce's Harness, ServiceNow's Context Engine, or an ERP-native equivalent) becomes the de facto standard other companies build agents against.
- More interface-layer acquisitions from Databricks and similar data platforms, aimed at governance rather than raw AI capability.
- Whether AI-native challengers like Ema convert early enterprise deals into renewals, the real test of whether outcome pricing sticks past a first contract.
- Enterprise software M&A pace into Q4, as capital-rich AI-native vendors and cash-flush incumbents both go shopping for the same category of governance and interface startups.
The Enterprise Software Signal
Enterprise software is moving from a system of record and a system of workflow toward a system of AI-driven execution — and every vendor covered this week is converging on the same four-part architecture to get there.
Whether the entry point is ERP (SAP, Oracle, Microsoft), CRM (Salesforce, HubSpot), workflow (ServiceNow), or the infrastructure underneath all three (Salesforce's Harness), the same components keep appearing: enterprise context, reasoning and planning, cross-system action, and governance controls. That convergence, more than any single product launch, is the strongest evidence that agentic enterprise software has moved past marketing.
| Development | Layer of the Stack |
|---|---|
| Ema / CRM outcome pricing | Pricing & business model |
| SAP / Oracle / Microsoft ERP | System of record → system of action |
| Salesforce / HubSpot CRM | Customer workflow & seat economics |
| ServiceNow Context Engine | Workflow orchestration & governance |
| Salesforce Enterprise AI Harness | Enterprise AI infrastructure |
| Databricks / Row Zero | M&A & capital allocation |
The CODEW Lens: The enterprise software market isn't choosing between incumbents and AI-native challengers. It's rebuilding pricing, architecture, and governance simultaneously — and the vendors moving on all three at once are the ones actually positioned to own the agentic transition rather than get disrupted by it.
Sources
→ TechCrunch — Ema raises $77M as AI starts eating into enterprise software and services
→ SaaStr — HubSpot's shift to per-resolution AI pricing, and Salesforce's multi-model repricing
→ InfotechLead — How AI is changing ERP software in 2026: SAP, Oracle and Microsoft push agentic ERP
→ ServiceNow Newsroom — ServiceNow moves beyond the sidecar AI era
→ Salesforce — Salesforce introduces the Trusted Enterprise AI Harness
→ TechCrunch — Databricks buys Row Zero and is scouting for more startups to acquire
The CODEW Stat
7 threads · 62% of cloud ERP spend AI-enabled by 2027 · 3 pricing models inside one CRM vendor · $190B Databricks valuation This week's enterprise software developments touched every layer of the agentic transition at once: pricing and business model (Ema, HubSpot, Salesforce), ERP's shift from record to action (SAP, Oracle, Microsoft, backed by Gartner's 62%-by-2027 forecast), CRM's seat-economics fracture, workflow orchestration and governance (ServiceNow), the enterprise AI infrastructure layer underneath it all (Salesforce's Harness), and the capital chasing the governance gap agents create (Databricks/Row Zero, fresh off a $190 billion valuation). Together they answer this edition's opening question: enterprise software is moving toward a system of AI-driven execution, and the vendors treating that as an architecture problem — not a feature to bolt on — are the ones pulling ahead.
Editorial Note
The Enterprise Software Watch examines the developments reshaping enterprise software and AI agent platforms, including agentic ERP and CRM, SaaS pricing models, AI control planes and context layers, enterprise data and governance, workflow automation, 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, SaaStr, InfotechLead, ServiceNow Newsroom, and Salesforce cited above. 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.
Reviewed by Erwin Castro
on
Monday, September 28, 2026
Rating:
No comments: