Workday Buys Agent Identity, ServiceNow Puts a Passport on Every Agent, Schneider Electric Pays $22.6B for PTC, and ERP Vendors Race to Govern What Their Agents Can Actually Do
The Battle Has Moved From Building Agents to Governing Them
For two years, enterprise software competition was measured by how many AI assistants a vendor could ship. That phase is ending. The announcements that matter this month are not about adding another copilot — they are about who controls what an agent is allowed to touch, where its permissions come from, and who can revoke them. Workday signed a definitive agreement to acquire C1.ai, an agentic identity platform for people and agents, at Workday Rising on October 13, with closing expected by January 31, 2027. ServiceNow followed on October 15 with AI Passport, a cryptographically signed credential that authorizes what an agent can do at runtime, shipping in Q4 2026.
The urgency is visible in the numbers. C1.ai's research found that 95% of large U.S. companies already run autonomous agents, yet only 21% report having a mature way to govern them, according to Deloitte's State of AI research cited in the acquisition announcement. ServiceNow's 2026 Banking AI Maturity Index found a similar gap: 100% of banks report using AI, but only 24% have governance to match. Agents are in production. The controls are not.
Two adjacent developments widen the frame. Schneider Electric agreed to acquire PTC for $22.6 billion in equity value, an all-cash deal at $205 per share and a 42.3% premium, aimed at building a unified "digital thread" of engineering and operational data for industrial AI. And the ERP vendors — Oracle, SAP and Infor — have spent the last month converting agentic promises into generally available products, with SAP's May agentic releases now shipping and Oracle's Fusion Claw offering 25 governed agentic applications. The through-line: enterprise software is consolidating around five control points — identity, runtime authorization, workflow ownership, the industrial data thread, and monetization.
Enterprise Software at a Glance
| Development | Vendor | Control Point | Deployment Reality |
|---|---|---|---|
| C1.ai acquisition | Workday | Agent & human identity | Signed — closes by Jan 31, 2027 |
| AI Passport | ServiceNow | Runtime authorization | Announced — Q4 2026 |
| PTC acquisition | Schneider Electric | Industrial data thread | Signed — pending regulatory |
| Agentic ERP releases | Oracle, SAP, Infor | Workflow ownership | Shipping — SAP GA this month |
| Agentforce work units | Salesforce | Monetization | Live — $1.5B ARR, mostly contracted |
| Employee service agents | Zendesk | Frontline workflow | Launched Oct 20 — early access |
Signal summary: Only one of these six developments is fully live and revenue-generating. The rest are signed agreements, Q4 availability targets, or pending regulatory review. That gap between announcement and deployment is the defining feature of the current agentic software cycle — and it is where the competitive outcome will actually be decided.
The Stories Shaping Enterprise Software
Workday Buys C1.ai to Become the Identity Layer for People and Agents
What changed: Workday signed a definitive agreement to acquire C1.ai, an agentic identity platform for people and agents, announced at Workday Rising on October 13, 2026. C1.ai connects to the systems a business runs on — email, CRM, code repositories, finance systems and cloud infrastructure — granting access at the right moment and removing it when it is no longer needed. Financial terms were not disclosed. Closing is expected in Workday's fiscal 2027 fourth quarter, ending January 31, 2027.
Strategic significance: Workday is making a specific argument: HR systems are the source of identity, because every employee enters and leaves the enterprise through them. "Every new hire, job change, and exit is a Workday event," the company states in its announcement. Co-founder and CEO Aneel Bhusri put it more bluntly to analysts: "HR is the source of identity", and ServiceNow doesn't have it. Salesforce has it for salespeople and the people that are in their system. But the one thing about every company is that everybody in the company is an employee". The acquisition turns that upstream position into a governance product that covers both human workers and the agents acting on their behalf.
Competitive implications: This is a direct move into territory occupied by Okta, Microsoft Entra, and SailPoint — but framed differently. Workday is not selling a standalone identity provider; it is positioning itself as an agentic governance layer that works with existing IDPs. "We can work with other IDPs whether it's Microsoft or Okta, but we now have that agentic layer that manages human and agent governance," Bhusri said. If that framing holds, Workday becomes a complement to identity providers rather than a replacement — which lowers adoption friction but also caps how much of the identity stack it can capture.
Business-model implications: Workday is explicitly moving toward consumption. Bhusri outlined five monetization routes — Flex Credit-based agents, APIs, Workday Data Cloud, Workday Extend AI for custom applications, and the traditional core subscription. CFO Zane Rowe disclosed that 300 customers are signed up for Flex Credits and more than 5,500 customers are using organic AI agents, while Sana, Extend AI, HiredScore, and Paradox are each at or approaching the $100 million revenue mark. Accenture has signed for Flex Credits covering 400,000 employees. Bhusri called consumption "a far better model than a people-based subscription-based model" — an unusual admission from a company built on per-employee subscriptions.
Deployment reality: The acquisition is signed but not closed. C1.ai's capabilities will not be integrated into Workday's stack until fiscal 2027. C1.ai's claim that 95% of large U.S. companies run autonomous agents while only 21% have mature governance comes from its own research and Deloitte — treat it as directionally useful rather than independently audited.
What's next: Watch whether Workday publishes agent-governance adoption metrics after closing, and whether the C1.ai integration extends beyond Workday's own systems into third-party applications at meaningful depth. Also watch the Anthropic partnership announced at the same event: Claude connected to Sana agents and Workday's HR and finance capabilities through the Model Context Protocol, with the Sana-in-Claude plugin in early access and general availability expected in late 2026.
The CODEW Take: Workday has identified a real control point and a credible claim to it. But "HR is the source of identity" is an argument, not yet a product advantage. The test is whether enterprises will accept their HR vendor as the governance authority for agent permissions across the whole application estate — including systems the HR vendor does not own.
ServiceNow's AI Passport Enforces What Agents Can Do While They Run
What changed: On October 15, ServiceNow announced AI Passport, a cryptographically signed credential that every AI agent — first-party or third-party — presents at runtime to verify its authorized scope. When an agent stays within scope, it operates freely. When it drifts, ServiceNow applies graduated responses: redirecting, restricting, reconnecting with proper context, or invoking a kill switch at machine speed. AI Passport, shadow AI detection, and AI Cost Management are all expected in Q4 2026. NVIDIA is a design and launch partner, implementing enforcement through its Open Agent Safety Platform, including OpenShell and the Sentry security framework.
Strategic significance: ServiceNow is making a sharper architectural bet than Workday. Where Workday anchors authority in the HR record, ServiceNow anchors it in the workflow and configuration management database. Each AI Passport is issued by the Super Intelligence Control Tower and grounded in the customer's CMDB, GRC policies, and workflow approvals. The company explicitly frames this as prevention rather than audit: "Unlike audit trails, which show what happened after the fact, AI Passport helps prevent AI agents from acting outside their scope by authorizing what they can do at runtime".
Competitive implications: The open-specification design is the most consequential detail. ServiceNow says AI Passport is built so that any governance product, identity provider, or enforcement point can issue and verify credentials for its own and third-party agents. If that holds in practice, ServiceNow becomes a standards participant rather than a walled garden — which is a stronger position for a company that does not own the enterprise identity record. But it also means the governance layer could commoditize if every platform adopts compatible credentials.
Business-model implications: AI Cost Management is the commercial hook. ServiceNow is integrating frontier model, embedded SaaS AI, hyperscaler and data center GPU costs into a single view, with alerts for budget overruns and spend grounded in business context. That gives the platform a CFO-facing argument alongside the CISO-facing one — a dual buyer motion that is difficult for point-solution governance vendors to replicate.
Deployment reality: Nothing here is generally available. AI Passport, shadow AI detection, and AI Cost Management are all Q4 2026 targets. The NVIDIA enforcement implementation is described as production-ready, but production availability of the combined offering is still ahead. ServiceNow's Q2 2026 AI-related contract value was disclosed at roughly the $1 billion level; whether AI Passport moves that materially is an open question.
What's next: Watch two things — whether third-party platforms actually implement compatible credential verification, and whether ServiceNow discloses AI Passport attach rates once it ships. Also note the October 20 launch of AI Workflow Factory and Autonomous Engineer, which extends the same "continuous improvement loop" logic into workflow engineering.
The CODEW Take: AI Passport is the most technically credible governance primitive announced this cycle because it addresses the actual failure mode — an agent doing something it was never authorized to do, in the middle of a multi-step workflow. The risk is adoption: cryptographic credentials only work if every enforcement point in a heterogeneous enterprise participates.
Schneider Electric's $22.6B PTC Deal Is a Bet That Industrial AI Needs the Design Data
What changed: Schneider Electric and PTC signed a definitive agreement on October 5, 2026, under which Schneider will acquire PTC in an all-cash transaction at $205 per share, valuing PTC's equity at approximately $22.6 billion (€20.1 billion) and implying an enterprise value of $23.7 billion. That represents a 42.3% premium to the last closing price and 21x EV/adjusted EBITA for 2027E. Schneider expects €250 million of annual run-rate cost synergies by year three and roughly €800 million of revenue synergies.
Strategic significance: The stated rationale is a "unified digital thread" — connecting PTC's product design, engineering and lifecycle data with Schneider's AVEVA industrial software and Cognite AI capabilities, creating what CEO Olivier Blum called "the industry's most complete Software & AI powerhouse". The logic is that industrial AI is only as good as the data it can reach, and design and engineering data is where the richest context lives.
Competitive implications: The deal places Schneider alongside Siemens in the software-heavy camp of European industrials, while ABB has explicitly chosen the opposite path. ABB chair Peter Voser told the FT: "We're not a company that will buy software companies or AI companies… we're not in favour of buying something where we're not completely at home". The FT reported Schneider shares fell 10% on the day the deal was announced, with Jefferies analysts questioning whether the transaction would "dilute" the energy management and data center business. That market reaction is a useful counterweight to the strategic narrative.
Business-model implications: This is fundamentally a recurring-revenue play. The deal adds high-quality recurring software revenue to a hardware-led business, and the 21x multiple only compresses to 13x if full run-rate synergies materialise. That is a meaningful execution dependency: the valuation assumes the synergies arrive.
Deployment reality: Signed, not closed. The transaction is subject to regulatory review, and the synergy targets are multi-year projections rather than demonstrated outcomes. Note also that this is an industrial-software acquisition, not an agentic-AI one — the AI rationale is forward-looking.
What's next: Watch regulatory timelines in both Europe and the U.S., and whether Schneider discloses customer adoption of the combined AVEVA–PTC–Cognite stack. Watch too whether Siemens responds with a competing software move.
The CODEW Take: The strategic logic is sound, but the market's skepticism is earned. Buying the design data layer is a defensible way to compete in industrial AI, but it is a multi-year integration of three separate software cultures — AVEVA, PTC and Cognite — inside a company whose core competence is electrification. The $22.6 billion question is whether Schneider can operate software better than it can buy it.
ERP Vendors Ship Agents — and Quietly Reopen the Customization Question
What changed: Three ERP leaders made concrete moves within weeks of each other. Oracle unveiled Fusion Claw runtime software in late September, providing 25 agentic AI applications with built-in governance and security. SAP used its Connect conference to declare that the agentic AI products first announced in May at Sapphire are now available this month, showcasing Nestlé, Novartis, PwC and Morgan Stanley as early adopters, and introduced SAP Pay alongside an agreement to acquire
TechWolf. Infor unveiled its Infor Industry AI architecture, with agents specialized in industry micro-verticals, a third-party-application layer, an adaptive UX, and a governance, risk and compliance layer that agents must pass through.
Strategic significance: The ERP vendors have largely finished building AI front ends. SAP's Joule has evolved from a chat client into Joule Work, a full web, mobile and desktop engagement layer, with Manoj Swaminathan describing the shift as moving away from "a static set of experiences" toward "a purpose-driven, targeted experience for an event, a situation or a particular task". That matters because ERP usability has been the category's oldest complaint, and the vendors are now claiming to have solved it.
Competitive implications: The more interesting competitive dynamic is defensive. As TechTarget's analysis frames it, these releases are partly designed to "assure buyers" and "deter frontier AI rivals" — meaning the ERP vendors are trying to prevent model providers and AI-native startups from inserting themselves between the enterprise and its system of record. The governance layers Oracle and Infor are building serve that purpose as much as any customer need.
Business-model implications: Infor's Velocity program offers "unlimited" access to its Velocity Suite of AI technologies so customers can deploy and customize industry-specific AI apps quickly — a bundled-access model rather than per-agent metering. That is a different monetization stance from Workday's Flex Credits or Salesforce's per-action pricing, and it suggests the ERP mid-market is competing on inclusion rather than consumption.
Deployment reality: Uneven. Oracle's Fusion Claw is available. SAP's agentic products are generally available this month, but analyst Joshua Greenbaum cautioned that the AI-driven ERP UX remains aspirational: "Those agents still have to be built, vendors have to build the infrastructure to support it, and they have to build security infrastructure". SAP's TechWolf deal was announced October 6 and is expected to close in Q4 2026, subject to customary conditions — it is not complete. Infor's architecture is shipping with named customers, but broad adoption is unproven.
What's next: Watch whether agentic customization reintroduces the over-customization problem that SaaS was supposed to solve. Greenbaum called this a risk, citing a Danish company that used Joule to build an app duplicating a standard SAP Ariba feature. Also watch whether TechWolf's Context Graph for Work becomes the context substrate for SAP's Joule agents after closing.
The CODEW Take: The ERP vendors are making the right architectural moves, but they are solving a governance problem with product announcements rather than with deployed controls. The customization question is the sleeper issue: if every enterprise builds its own agents on top of ERP, the governance surface expands faster than any vendor's control plane.
Salesforce Is Rewriting Its Own Pricing Model Before Someone Else Does
What changed: Salesforce's Agentforce has passed $1.5 billion in annual recurring revenue, up roughly 240% year over year, with customers running 3.2 billion AI "work units" in the quarter — up 97% sequentially, and 7 billion cumulative. The company is shifting from per-seat billing toward "agentic work units" plus Flex Credits, priced at $500 per 100,000 credits, with $0.10 per action and $0.15 per voice interaction.
Strategic significance: Salesforce is the clearest case of a SaaS incumbent deliberately dismantling the pricing model that built it. The reason is arithmetic. Per-seat pricing ties revenue to headcount; agentic work ties revenue to tasks executed. If agents do the work that seats used to be sold for, seat revenue compresses unless the vendor finds a new meter. Gartner has framed the stakes at scale: it identified $234 billion in enterprise application software spend at risk from "agentic arbitrage" by 2030, with managing vice president George Brocklehurst noting that agentic systems "deliver outcomes directly, bypassing traditional user experience-heavy applications and making the software invisible," which "breaks the link between user growth and revenue growth".
Competitive implications: Salesforce is hedging across all four competing pricing architectures — per-action credits, per-resolution, outcome-based, and hybrid seat-plus-consumption. That hedging is rational, but it also means the company has not committed to a thesis. Peers are picking different lanes: HubSpot charges $0.50 per conversation, Zendesk targets $1.50–$2.00 per automated resolution, and outcome-based startups are raising on the promise of charging for results.
Business-model implications: The disclosure gap matters. The $1.5 billion Agentforce figure — roughly $3.9 billion including Data 360 — is annualized contract value: what has been sold, not what has been recognized. Total revenue grew 11% to $11.3 billion, and about $456 million of the quarter came from the Informatica acquisition, not from agents. Full-year guidance is $46.1–$46.4 billion, up 11–12%. The AI business is real and growing fast, but it is still roughly 11% of a $46 billion base, and unit pricing margins remain undisclosed.
Deployment reality: Agentforce is live and generating contracted revenue — the strongest deployment position of any story in this edition. But the open question is whether work-unit economics beat the seats they replace. If Agentforce substitutes for licensed seats at a lower effective rate per worker, explosive volume could still shrink total revenue.
What's next: Watch for disclosed work-unit gross margins, the conversion of contracted ACV into recognized revenue, and whether Salesforce commits to a single pricing architecture or continues hedging.
The CODEW Take: Salesforce is doing the hardest thing an incumbent can do — cannibalizing its own meter before a competitor does it. The strategic logic is correct. The unresolved risk is that per-action pricing is a worse business than per-seat pricing if the actions are cheap and the seats are expensive.
Strategic Analysis: The Five Control Points of Agentic Enterprise Software
The five stories in this edition are not five separate product announcements. They are five vendors staking claims to different chokepoints in the same emerging architecture. Mapping them together makes the competitive structure visible.
| Control Point | The Claim | Who Is Contesting It |
|---|---|---|
| Identity | Who — and what — exists in the enterprise, and what that entity is entitled to | Workday, Okta, Microsoft Entra, SailPoint |
| Runtime authorization | What an agent may do at each step of a live workflow | ServiceNow, NVIDIA, cloud providers |
| Workflow ownership | Where the work actually happens, and therefore where the agent must plug in | Oracle, SAP, Infor, Salesforce, Zendesk |
| Data thread | The contextual data that makes an agent's decision defensible | Schneider/PTC, Siemens, ABB |
| Monetization | The meter that replaces the seat | Salesforce, Workday, Infor, outcome-pricing startups |
Why identity and authorization are separating. Workday and ServiceNow are attacking adjacent but distinct problems, and the distinction matters. Identity answers "who or what is this, and what is it entitled to?" Authorization answers "is this specific action permitted, right now, in this workflow?" An agent can have valid identity and still attempt an action outside its scope at step seven of a twelve-step process. That is why ServiceNow's runtime framing is architecturally stronger for the failure mode that actually occurs — but it is also why identity remains the upstream chokepoint. Workday's argument that HR is the source of identity is, in structural terms, correct: the employment relationship is the origin of most enterprise entitlements.
Why workflow ownership still wins the distribution battle. Governance layers only work if agents route through them. The ERP vendors own the workflows, which means they can make their own governance layer the default path and treat third-party governance as an integration. That is why Oracle embedded governance inside Fusion Claw and Infor built a GRC layer agents "must pass through" — it is defensively as well as functionally motivated. Workday and ServiceNow are strong enough to force integration, but neither can dictate terms to SAP or Oracle inside SAP's or Oracle's own workflows.
Why the monetization question is unresolved and probably will stay that way. Four pricing architectures are competing, and none has won: per-action credits, per-resolution, outcome-based, and hybrid seat-plus-consumption. Salesforce hedges across all four. Workday is moving to Flex Credits but explicitly says it is "neutral about which monetization model you choose". Infor bundles unlimited access. That inconsistency is not confusion — it is a market that has not yet established what an agent's work is worth. Until buyers can measure cost per completed task against a defensible baseline, vendors will keep offering multiple meters and discounting whichever one the customer resists least.
What's Next
- ServiceNow AI Passport general availability in Q4 2026. The first real test of whether cryptographic runtime authorization works across third-party agents and MCP servers in production environments.
- Workday/C1.ai closing by January 31, 2027. Watch whether Workday publishes agent-governance adoption metrics and how deeply C1.ai extends beyond Workday's own application estate.
- Schneider/PTC regulatory review. A $22.6 billion all-cash industrial software deal will face scrutiny in both Europe and the U.S.; the timeline shapes whether synergies begin in 2027 or later.
- Salesforce work-unit gross margin disclosure. The single most important missing number in the SaaS pricing debate. Without it, the bull and bear cases remain equally coherent.
- SAP TechWolf closing and Context Graph integration. Whether workforce context becomes the substrate for Joule agents, and whether competitors acquire equivalent capability.
- ERP agentic customization spillover. Whether enterprises build agents that duplicate standard ERP features, and whether vendors respond with governance controls or feature bundling.
- Adjacent entrants moving on the same control points. Zendesk's employee service agents launched October 20 with 6 million workflow executions in five weeks of early access; Pax8 launched an operating system for MSPs to govern and monetize agents. The governance layer is attracting players well beyond the five vendors in this edition.
The strategically important control points are runtime authorization and workflow ownership — in that order. Identity is a real chokepoint, but it is upstream and increasingly standardized; the defensible position is controlling what happens during execution, inside the workflow where the work actually occurs.
That ranking is uncomfortable for the two vendors making the loudest governance claims. Workday has the strongest identity argument and the weakest runtime position, because its agents operate across systems it does not control. ServiceNow has the strongest runtime architecture and the weaker identity claim, which is precisely why it built AI Passport on open specifications and partnered with NVIDIA rather than attempting to own the credential root itself. Both are hedging against their own structural weakness, which is a sign the market has not yet settled.
What evidence is still missing? Four gaps stand out, and none can be closed by a press release. First, there is no independent data on how many enterprise agents are actually operating under enforced runtime scope as opposed to registered but unconstrained. The 21%-mature-governance figure is a vendor-cited survey statistic, not an audited measurement. Second, ServiceNow's AI Passport has not shipped; its interoperability claims across third-party governance products are architectural intent, not demonstrated behavior. Third, Salesforce's work-unit margins are undisclosed, which means the central question of the SaaS pricing transition — whether per-action economics beat per-seat economics — remains unanswerable from public data. Fourth, Schneider's €800 million revenue synergy target is a projection attached to a deal that has not closed and whose own shareholders marked it down 10% on announcement.
What would change the assessment? If AI Passport ships in Q4 and a meaningful cohort of third-party platforms verifies its credentials, runtime authorization becomes a durable control point, and ServiceNow's position strengthens materially. If Workday closes C1.ai and demonstrates governance across third-party applications at scale, the identity argument becomes real rather than rhetorical. If Salesforce discloses work-unit margins above seat-equivalent economics, the SaaS transition is validated. Conversely, if enterprises keep building agents faster than they govern them — the pattern the current data suggests — then every governance product in this edition is selling into a demand curve that exists but is not yet being acted on.
The structural conclusion. Enterprise software is not being replaced by AI agents. It is being reorganized around them — and the reorganization is happening at the control layer, not the application layer. The vendors that win will be the ones that make their control point default rather than optional. That is a distribution problem disguised as a technology problem, and it is why workflow ownership still matters more than architectural elegance.
The CODEW Stat
21% — the share of large U.S. companies that report having a mature way to govern AI agents, against 95% that already run them, according to Deloitte research cited in Workday's acquisition announcement. It is the most consequential number in this edition because it defines the entire commercial opportunity being contested. Every vendor in these five stories is selling into the same gap: agents are already deployed, the controls are not. But the number cuts both ways. A 74-point gap between deployment and governance is a large market — or it is evidence that enterprises are not yet prioritizing the problem. Which reading is correct will determine whether agent governance becomes a durable software category or a feature that gets bundled into platforms that already own the workflow.
Sources: Primary sources: Workday newsroom and C1.ai announcement; ServiceNow investor relations and newsroom; PTC investor relations; SAP News Center; Oracle; Infor; Salesforce investor materials. Independent reporting and analysis: Constellation Research, CRN, TechTarget, CDO Magazine, Computerworld, Forkast, AInvest, Financial Times, TechWolf, IT Brief, Futurum Group. Coverage window: October 5–20, 2026. Where a development was announced before October 20, it is presented as a continuing development rather than as new October 20 news.
All factual claims regarding announcements, transaction terms, product availability, and financial figures are drawn from company disclosures and contemporaneous reporting. Editorial analysis is clearly distinguished from reported facts throughout. Signed acquisition agreements are described as signed, not completed; regulatory approval and customary closing conditions remain outstanding. Products described as expected in Q4 2026 or later are not generally available. Vendor-cited survey statistics, ARR figures, and synergy projections are labeled as company-reported and have not been independently verified. Salesforce Agentforce ARR is annualized contract value, not recognized revenue. This series does not provide investment advice.
Reviewed by Erwin Castro
on
Tuesday, October 20, 2026
Rating:
No comments: