Oracle Agentic Implementation – When Agents Draft The Setup, Review Becomes The Control

What Oracle Set Out To Fix: Implementations That Ask Too Much, Too Early

“Long ERP implementations ask too much of customers before they can start realizing value,” said Evan Goldberg, founder and executive vice president of Oracle NetSuite, as NetSuite launched SuiteExpress. At Oracle AI World, Oracle set out to shorten that wait across both of its application families, with agents that draft designs and configuration, a public registry of how Fusion should be set up, and an app builder that lets employees create applications with OpenAI’s Codex.

Taken together, the announcements treat implementation as work that agents can share. Oracle AI Success Navigator brings six agents, named Discovery, Design, Configuration, Testing, Integration, and Innovation. The Oracle Fusion Setup and Configuration Registry publishes Oracle’s setup knowledge for Fusion as a public, machine-readable source. SuiteExpress does similar work for NetSuite, and Oracle App Composer lets people “build applications with Codex, while Oracle manages deployment and operations on OCI.”

Greyhound Research reads this as the right target, and as a shift that moves risk rather than removing it. In its research report on Fusion Claw, Greyhound Research found that the authority an agentic application runs under is set in controls Fusion already holds, and that “a delegation written carelessly is exercised carelessly.” Those controls live in configuration. When agents draft the configuration, they help write the rules that later agents will obey.

The test for buyers is whether that speed can arrive without moving decisions past the people accountable for them. Passing it depends less on the agents than on four things buyers control: the readiness of their processes, data, and legal review, how the implementation contract is written, who signs off what the agents draft, and who owns what gets generated.

↑ Back to the section guide

What Oracle Announced: Five Components With Uneven Status

Oracle’s announcements share one idea but sit at different stages, and the stage decides what a buyer can put into a plan. An implementation plan is a dated commitment, and a component without a date cannot carry one. As Table 1 shows, only the registry and SuiteExpress carry availability wording a plan can rely on, and SuiteExpress only for one industry in two countries.

Read by status, the picture is uneven in a useful way. Any buyer can inspect the registry now, and an eligible NetSuite customer can start with SuiteExpress. The Navigator release describes a platform already in use by an early adopter and gives no dates, editions, or regions. App Composer and the services practice are statements of direction. Those three belong in a plan only once Oracle confirms their status in writing.

↑ Back to the section guide

What Buyers Are Asking For: Process, Data, And Legal Before Speed

Oracle’s customers at the event described an order of work that starts before any agent runs. Four Oracle HCM customers described the obstacles between them and agents in their HR operations as process, data, legal review, and cost, with technology only one part of the equation. They were speaking about agents in their HR applications, not about Oracle’s implementation agents. The parallel is still direct, because an implementation is where process, data, and legal positions are first written into the system.

On process, Matt McDonald, director of HR technology at IHG Hotels & Resorts, said “We’re really focusing on the process and letting that drive it.” Kelli Scroggin, director of HRIS and analytics at UL Solutions, said the company has “to revisit some of those decisions” from its earlier implementation to build the data AI needs.

Oracle’s registry release says AI assistants and coding agents using the registry can translate “business requirements, process documents, policies, and existing configurations” into implementation runbooks. Process documents record how work is currently done, workarounds included. Greyhound Research expects an agent that drafts from them to reproduce those workarounds faithfully, unless a process owner has first decided which ones to keep.

On data, Daniel Passarge, head of HR IT systems at Deutsche Bahn, advised “Start with data cleansing and look into your data.” That work began “before we even started the implementation project, and it took until the go-live.” Oracle’s tools treat data as implementation work too. SuiteExpress prepares and validates data from existing systems, and the Navigator release gives its Discovery Agent the job of identifying “scope, process gaps, risks, and modernization opportunities.” Agents can speed that analysis, but cleansing still depends on people who know what the data should say.

Set side by side in Exhibit 1, the customers’ statements ask an agent-led implementation for three things before speed: an owner for each process decision, data prepared as its own workstream, and legal and authority review before configuration is applied.

On legal review, Doug Johnivan, vice president of operations and apps support at Acosta Group, advised bringing legal teams in early: “get them involved from day one, because they will slow you down, unfortunately.” McDonald framed the question every agent has to answer: “Where’s the human in the loop?” In an agent-led implementation, that question has two answers. One is the reviewer who signs off what the agents draft. The other is the set of approval steps that the configuration keeps, or removes, for the agents that run later.

Customers want speed as well. Scroggin put it directly: “we need to be faster, we need fewer approvals, we need fewer handoffs.” Rob Dombkowski, senior director of business technology at Grupo Bimbo, valued a Fusion Claw demonstration for giving him “something tangible quickly that the business can get a benefit from.” In the SuiteExpress release, Steve Hehl, chief operating officer of Worksmith, says the process “required far less of our team’s time than I expected.”

The two sets of statements fit together once approvals are read as configuration. Approvals and handoffs of the kind UL Solutions wants to remove are set up inside the application. Removing one changes who may decide what, which makes it a change to authority as well as to process. Speed and control are therefore compatible in an agent-led implementation, provided the configuration that removes an approval is itself approved by the person accountable for that authority.

Survey data from technology leaders points the same way. Greyhound CIO Pulse 2026, across more than 1,000 enterprise technology leaders, finds 84 percent have shifted their focus from AI adoption to AI accountability, and fewer than one in three can point to measurable outcomes across most of their funded AI portfolio. For implementation agents, accountability means a record of what each agent drafted and who approved it. A measurable outcome means a configured, tested, and accepted process, not hours saved.

↑ Back to the section guide

The Case For It: A Shared Source For People And Agents

The least glamorous piece of the announcements, the registry, is also the strongest. Oracle’s registry release frames the problem plainly, describing manual setup and configuration that “can consume significant time during an implementation,” and answers it with a single source that “customers, partners, and AI agents can work from.” Greyhound Research regards that as a sound base, because an agent that drafts a configuration and the person who reviews it can then argue from the same text.

Because the source is “a public Git repository,” it keeps its history, so changes to Oracle’s recommended practice should become visible, and a buyer can compare its own configuration against what Oracle now recommends. Oracle describes the registry as “continuously maintained” alongside the product. If that holds, the registry also narrows the gap between how an application was configured at go-live and how the vendor would configure it several updates later.

The early evidence is narrow but specific. Lexington Health is the one customer the Navigator release names. Tom Tan, its executive director, IT, Business Services, reports that “right now, during implementation, the Oracle AI agentic delivery approach has allowed us to reduce certain design and configuration efforts by 80 percent.” The gain covers certain efforts only, and the figure is self-reported from a project still under way. Greyhound Research treats it as a credible signal about where agents help first, in turning agreed designs into configuration, and not as a forecast for a whole program.

In the registry release, Oracle quotes five implementation partners in support of the approach: Accenture Edge, Apex IT, Deloitte, Grant Thornton, and PwC. Oracle’s partner investment announcement also lists AI Success Navigator among tools “for partners to support customers beyond initial deployment.” For a buyer, that alignment means the firm implementing the system can work from the same source as Oracle’s agents, not from a private method the customer cannot inspect.

NetSuite’s version goes further on timing. Oracle NetSuite says SuiteExpress reduces “implementation time to weeks or even days,” and two customers in the release describe faster and less demanding projects. These are vendor-published claims under Oracle’s standard future product disclaimer, and SuiteExpress is currently available to only one industry. They show the direction Oracle NetSuite is taking for its own customers, and they give Fusion buyers a reference point to raise with their implementers.

Ranked by how far a buyer can check it in Exhibit 2, the evidence sits mostly in the middle. One item can be tested by any buyer now, and the strongest customer figure is self-reported from a project still under way. Results across a full Fusion program do not yet appear in the sources reviewed.

↑ Back to the section guide

Where The Risk Moves: Configuration Is Where Authority Is Written

In a conventional implementation, a design passes through workshops, a build, and testing at human pace, and errors surface slowly enough to be argued over. When an agent drafts the design and builds the configuration, the same error can be produced faster and applied more consistently. Consistency is a strength when the design is right, and it multiplies the cost when the design is wrong. The review step, not the drafting step, becomes the point where quality is decided.

That configuration also carries authority. Oracle’s Release 26D documentation says Fusion Claw “operates within the organization’s established policies, permissions, approvals, risk thresholds, and authority limits.” Those policies, approvals, and limits do not sit outside the application. They are set up inside it. Oracle’s Fusion Claw launch release says Claw applies an “Enterprise Operating Envelope,” and Greyhound Research’s report on Fusion Claw described that envelope as “the delegation matrix made executable.” Agentic implementation means agents now help write that matrix.

Oracle has designed for this at two points. In the Navigator release, the Design Agent produces “specific application designs for review and approval,” which keeps a person at the moment a business process becomes a system design. For configuration itself, the registry release says agents can “execute setup and updates directly in Oracle Fusion Applications” and that “expert reviewers can examine completed configurations, request adjustments, and approve them for deployment.”

Neither release yet says who that reviewer must be, under whose identity the agent acts, or how a record of drafts, changes, and approvals is kept. Greyhound Research interprets that as a gap in documentation rather than intent, and buyers can close it in their own project controls until Oracle documents it.

Across the six stages in Exhibit 3, Oracle’s releases describe a person approving the agents’ work at only two, Design and Configuration, and Configuration is where the authority Fusion Claw later runs within is set.

The other four stages carry no review language in the releases. That does not mean no review happens, and buyers should specify the review instead of assuming it.

Greyhound Research’s research report on Oracle’s agent runtime set four tests for agents that act inside enterprise applications: name the identity, read the log, bound the authority, and reverse the action. Applied to the implementation agents in Table 2, the releases describe review for authority but not identity, a log, or rollback of applied configuration, so buyers need their own controls for those until Oracle documents more.

↑ Back to the section guide

Roles And Responsibilities: Who Signs What The Agents Draft

Oracle’s releases describe what each agent does and, at two points, that a person approves its work. The rest of the allocation is left to the customer and its implementer. Every agent-led implementation still needs people who supply the decisions the agents draft from, people who approve what the agents produce, and someone who keeps the record. Left unwritten, those duties fall to whoever is nearest the tool, which can leave them with the implementation partner by default.

Oracle NetSuite describes one model. The SuiteExpress release says the AI carries out implementation work “with oversight from consultants,” with “consultants guiding key decisions and handling exceptions.” That allocation is clear for a NetSuite rollout. A Fusion program needs a stricter one, because its configuration sets the authority Fusion Claw later works within. Greyhound Research proposes an allocation built on two rules: the drafter never approves its own output, and the owner of the delegation of authority co-signs any change to who may approve what.

In the allocation Greyhound Research proposes in Exhibit 4, Oracle’s agents draft at all six stages and approve at none, and only the approvals at Design and Configuration are described in Oracle’s own releases.

Configuration carries the most signatures because it is where approvals, roles, and limits are set. The expert reviewer approves the configuration for deployment, the owner of the delegation of authority co-signs where authority changes, and the data lead confirms the reference data it relies on. Internal audit samples the record at that stage and at testing, and the program director keeps one record across all six stages, so any setting can be traced to the design and decision it came from.

Two of the stages without review language need the most attention. The Navigator release says the Testing Agent determines “the steps needed to align testing with configurations.” Tests aligned to a configuration confirm that the system does what it was configured to do. Only the process owner can confirm that it does what the business needs, so acceptance should be tested against the business requirement, not against the configuration alone.

The Integration Agent “develops the code needed to build” integrations with existing systems, according to the same release. That code moves data between systems, and the release describes no review for it. The enterprise architecture lead should approve agent-written integration code the way any code is approved before production, with the same testing, security review, and change control.

↑ Back to the section guide

Generated Applications: Fast To Build, Still Someone’s To Run

App Composer extends the same idea from setting up applications to creating them. The release pairs it with APEXlang, a new Oracle technology that App Composer supports, which Oracle says “gives customers the speed of generative development without producing opaque application code.” Greyhound Research credits Oracle for leading with readable output at announcement, before any audit has to ask for it.

The company works this way itself. Oracle chief executive Clay Magouyrk has said that “more than 70% of our software these days is written by AI.” No measurement basis appears in the coverage or in the Oracle releases reviewed for this research note, so the figure reads as a signal of internal practice and not a benchmark. It does tell buyers that the vendor building these tools relies on AI-written code in its own engineering.

For an enterprise, the harder question about a generated application is who owns it a year after it is built. An application an employee builds in an afternoon still touches enterprise data, still needs testing when the underlying system updates, and still has to be retired when the process it serves changes. Without a lifecycle, the speed that makes App Composer attractive also makes it easy to accumulate small applications that nobody is responsible for running.

Set against the six lifecycle decisions in Exhibit 5, Oracle’s announcement speaks to three, who may build, what the code looks like, and where it runs, and is silent on data access, testing at each update, and ownership through to retirement.

Timing works in the buyer’s favor. The release gives no availability for App Composer, and it carries Oracle’s standard future product disclaimer, which says the product direction it outlines “should not be relied upon in making purchasing decisions.” Organizations therefore have time to set their rules before the first generated application reaches production, which is far easier than retrofitting rules onto applications already in use.

↑ Back to the section guide

Partners And People: Who Captures The Time Saved

Agents that compress implementation effort change the economics for the firms that sell that effort. Greyhound Research sees Oracle’s partner commitments as bringing partners inside the change instead of around it. That choice serves customers who depend on those partners’ delivery capacity, and who need their implementer and their vendor working from the same method.

If an agent reduces certain design and configuration work by a large margin, the saving lands wherever the contract allows it to land. Under a time-and-materials contract, faster delivery reduces billed effort only if the plan is rebuilt around the agents. With a fixed price agreed before the agents were in scope, the saving can stay with the delivery side. Either outcome can be reasonable. A buyer should choose it deliberately, not inherit it from a contract template written for slower projects.

Two questions that any statement of work can answer decide where the saving lands: whether payment is tied to effort or to a tested outcome, and whether the contract declares which steps the agents perform. Exhibit 6 shows that only outcome-based payment with declared agent use makes the saving visible and shared by design.

The other three combinations can work, but each leaves the split to the delivery plan instead of the buyer’s choice.

If agents take on the drafting, a partner’s value moves to the work agents do not do. The registry release anticipates this, saying systems integrators “can extend this open foundation with their own methodologies, accelerators, automations, and industry solutions.” Greyhound Research expects the partners that gain most to be those that sell judgment: which process departures to keep, how to prepare the data, how to manage the change, and how to review what the agents draft. Those are the tasks customers at the event described as the hard ones.

Those changes belong in the statement of work. Table 3 sets out nine clauses for the next one, grouped as commercial, governance, and people and ownership. Four exist only because AI now does part of the work: the agent-use declaration, step-level pricing, the agent record, and generated-application ownership.

Skills decide whether a buyer can use any of this safely. Oracle’s OCI blog pairs the new AI Services Practice with a plan to train and certify one million professionals, and gives no timeframe for it. Greyhound Research’s research report on Oracle’s agent runtime cited Oracle’s count of 80,000 certified experts trained in AI Agent Studio. The two figures describe different programs, so the comparison is directional only. At more than twelve times that base, the plan describes a future workforce that a project staffed now cannot draw on.

The scarcer skill is review. Drafting scales with the agents, and approval does not, because each agent-drafted configuration still needs a reviewer who understands both the product and the customer’s delegation of authority. Oracle’s announcements do not say whether the practice or its certification prepares people for that role. A buyer can test the point directly by asking the partner who will review the agents’ work, what training those reviewers have had on the agents, and how many configurations each will sign off in a cycle.

↑ Back to the section guide

Risks To Watch: Where Agent-Led Implementation Can Go Wrong

Agent-led implementation fails in different ways from a conventional project, because the features that make it fast also make its errors consistent. Four risks stand out. Placed on the implementation stages in Exhibit 7, three of their four early warnings appear before go-live, while the fix is still cheap.

Best practice can flatten what differentiates. The Navigator release says the Configuration Agent builds on “product-defined best practices” available through the registry. Most processes gain from that. For the few that set a company apart, such as a pricing approval or a service commitment, a default can replace a deliberate difference without anyone deciding to remove it. Watch for a design that matches the registry everywhere. The remedy is a short register of intended departures, owned by process owners and checked at design review.

A registry the buyer does not control can drift from the buyer’s configuration. The registry’s value rests on Oracle keeping it current alongside the product, and a configuration built to track it closely also inherits its changes. When a recommended practice changes between updates, a configuration that was right at go-live can fall out of step. Watch for a registry change that touches an area holding approvals or limits. The remedy is to compare the registry’s history with the buyer’s own configuration at each Fusion update, within normal change control.

Review can slide into rubber-stamping. When agents draft quickly and well, the reviewer’s queue grows and the drafts look finished. Gains of the size early adopters report in drafting move the pressure onto review, which is where the control now sits. Watch for approval time per configuration falling while volume rises, or reviewers approving changes outside their own area. The remedy is a ceiling on what each reviewer signs in a cycle, set from the time a careful review takes, with internal audit sampling approved configurations.

Agents can carry bad data further, faster. Agents that prepare data and analyze current environments are only as good as the records they start from. An agent that maps a flawed supplier master or an outdated organization structure can reproduce the flaw consistently across every configuration that depends on it. Watch for test failures that cluster around reference data. The remedy is the data workstream that Deutsche Bahn started before its implementation began, run as a gate before agents draft configuration.

↑ Back to the section guide

The Greyhound CXO Runbook: Six Moves Before The Next Implementation Contract

Six moves follow from the evidence, written for the CIO and the head of enterprise applications, with procurement, legal, internal audit, and the owner of the delegation of authority alongside them. Each move names who decides. None waits for Oracle to date the agents it has announced, because the registry, the contract, the reviewers, and the record are already within the buyer’s control.

1/ Start with the registry, and plan on written status.

Use the registry now. Compare your current configuration with Oracle’s recommended practice, and brief your implementation partner from the same source.

Before putting any Navigator agent, App Composer, or the services practice into a project plan, get Oracle’s written status for your region and edition. A plan built on undated components inherits their uncertainty. The decision sits with the head of enterprise applications.

2/ Prepare process owners, data, and legal review before the first agent runs.

Name an owner for every process in scope, and have each one record which departures from Oracle’s recommended practice the business intends to keep. Start data preparation as its own workstream, gated before any agent drafts configuration.

Bring legal and the owner of the delegation of authority into design review from the first workshop. Agents draft from whatever decisions exist when they run, so the decisions should exist first. The decision sits with the CIO, together with process owners, the data lead, and the general counsel.

3/ Rebase the implementation contract on what the agents change.

Implementation contracts drafted before these agents existed say nothing about which steps an agent performs. Ask every bidder to state which steps Oracle’s agents or SuiteExpress will perform, how that changes the effort, and how the price reflects it.

Tie milestone payments to tested outcomes, such as a configured, tested, and accepted process, not to hours. Start from the nine clauses in Table 3, which put agent use, review, record, and ownership into one statement of work. The decision sits with the CIO and the procurement lead.

4/ Name the configuration reviewers, and have the delegation owner co-sign.

Agent-drafted configuration includes the approvals, roles, and limits that agentic applications later obey. Assign each role in Exhibit 4, where no drafter approves its own output, to a named person before the first agent run. Make one accountable person sign off configuration in each area, and require the owner of the delegation of authority to co-sign any change to who may approve what.

A configuration error in authority is not a one-time defect. It becomes an error in every agent run that relies on it, so set a ceiling on what each reviewer signs in a cycle. The decision sits with the head of enterprise applications, together with the owner of the delegation of authority.

5/ Keep the record Oracle has not yet described.

For every agent output, record what was drafted, who reviewed it, what changed, and when it was applied. Run agents under named project accounts with the least access each task needs, and keep a configuration baseline in a test environment before each agent run.

Auditors can be expected to ask how the configuration was produced, and the record should answer that before any agentic application runs on it. The decision sits with the program director, with internal audit agreeing the format before the project starts.

6/ Give generated applications a lifecycle before the first one exists.

Set the rules for App Composer now, while it is still arriving. Decide who may build, which data a generated application may touch, how each one is tested at every update, who owns it, and when it is reviewed for retirement.

Small applications built quickly are useful, and unowned ones accumulate quietly. The decision sits with the CIO and the enterprise architecture lead.

Sequenced in Exhibit 8, every move starts before the first agent runs, and four of them start before the implementation contract is signed.

↑ Back to the section guide

Questions For Oracle: What Buyers Will Ask Next

Q1 Buyers want visible value from the first phase, and early adopters are already working with the agents. Customers at the event valued AI that delivers a tangible benefit quickly, which starts with knowing what they can use now. Lexington Health is already using Oracle’s implementation agents mid-project. Which AI Success Navigator agents are generally available now, in which regions and editions, and are they included in existing subscriptions or metered separately?

Q2 Technology leaders have moved from adopting AI to accounting for it. Leaders who have made that shift are concentrating on whether AI can operate under the same governance standards that regulate their core data. For implementation agents, it starts with a record of what each agent built, which internal audit will expect to see before any agentic application runs on the result. Does each Navigator agent leave a record of what it drafted and applied, who reviewed it, and when, and can customers export that record?

Q3 Security teams will hold implementation agents to the standard Oracle sets elsewhere. In its own AI Security Mesh, Oracle says “each agent operates with a distinct identity, task-scope, and just-in-time permissions.” Teams that read that design will expect the same discipline from an agent that writes configuration into their production applications. Under whose identity does the Configuration Agent act when it builds configuration in a customer environment, and with what access?

Q4 Customers want fewer approvals, which is safe only where a change can be undone. Removing approvals and handoffs is one of the clearest gains customers expect from agents. It is a reasonable aim where a wrong change can be reversed quickly, and a risky one where it cannot. Can a customer roll back configuration applied by an agent to a prior baseline, and is that rollback recorded?

Q5 Confidence decides how fast customers move to agents that need heavy controls. Rondy Ng, Oracle’s executive vice president of applications development, expects applications that need many procedures, policies, and controls to “come after they have built up enough confidence in the solution.” That confidence is built earliest during implementation, when those controls are first written. Will the Enterprise Operating Envelope that governs Fusion Claw extend to agents acting during implementation, or apply only to agentic applications at run time?

Q6 Customers say the hard work before agents is readiness, and the scarce skill is review. Customers described data work that ran the length of a project and legal review that has to start on day one. Buyers will also need reviewers who can approve agent-drafted configuration, a skill the agents do not supply. When will App Composer and the AI Services Practice be available, and will the practice and its certification cover readiness work and the review of agent output, or the tools alone?

Read together in Table 4, four of the six questions concern controls Oracle’s releases do not yet describe, and each question maps to a Runbook move a buyer can make in the meantime.

↑ Back to the section guide

Agentic Implementation: The Final Greyhound Standpoint

Greyhound Research’s final judgment is that Oracle has pointed agentic AI at the part of enterprise applications that NetSuite’s own founder says asks too much of customers, and has started from a sound base. A public, machine-readable registry of setup knowledge, agents whose designs and configurations Oracle describes as subject to review and approval, and an app builder designed to produce readable code add up to a credible plan to shorten the distance between contract and value.

The evidence is early. One Fusion customer and two NetSuite customers report gains, most components carry no date, and Oracle’s releases do not yet describe identity, record, or rollback for agents that write configuration. The position would strengthen with written availability for each agent, a documented record of agent outputs and approvals, and named customers reporting results across a full program. It would weaken if agent-built configuration reached production without the expert review that Oracle’s registry release describes, or if programs cut readiness work to match the agents’ pace.

Of the five signals in Exhibit 9, two would show up first in a buyer’s own program rather than in anything Oracle publishes.

Until then, the buyers who gain most will be those who treat setup as governance. They will settle process, data, and legal positions before the first agent runs, start with the registry, and rewrite the next implementation contract around what the agents do. Above all, they will give the people who sign off configuration the weight of those who sign off authority, because where configuration sets approvals, roles, and limits, signing it off is signing off the authority that agents will later exercise.

↑ Back to the section guide

Sources And References: Every Document Behind Every Claim

Every fact, figure, and quotation in this research note traces to one of the sources below, presented in Harvard reference format and grouped by class. Each source is also linked in the sentence that carries it. Everything in the first three classes is Oracle’s own account of itself and proves what Oracle said rather than that the thing is true. Standing pages that carry no publication date are marked n.d. All sources were accessed on October 29, 2026.

Oracle announcements

Oracle Corporation (2026a) Oracle Announces $1 Billion Investment to Accelerate Partner Growth and AI-Driven Modernization for Customers. Press release, October 23. Available at: https://www.oracle.com/news/announcement/ai-world-oracle-announces-1-billion-commitment-to-partner-ecosystem-2026-10-23/

Oracle Corporation (2026b) Oracle Advances Enterprise Security for the AI Era. Press release, October 27. Available at: https://www.oracle.com/news/announcement/ai-world-oracle-advances-enterprise-security-for-the-ai-era-2026-10-27/

Oracle Corporation (2026c) Oracle Extends Fusion Agentic Applications with Introduction of Fusion Claw. Press release, September 29. Available at: https://www.oracle.com/news/announcement/oracle-extends-fusion-agentic-applications-with-introduction-of-fusion-claw-2026-09-29/

Oracle Corporation (2026d) Oracle Introduces AI Agent-Ready Fusion Applications Setup and Configuration to Speed Implementations. Press release, October 26. Available at: https://www.oracle.com/news/announcement/ai-world-oracle-introduces-ai-agent-ready-fusion-applications-setup-and-configuration-to-speed-implementations-2026-10-26/

Oracle Corporation (2026e) Oracle Reinvents Enterprise Computing for the AI Era with OpenAI. Press release, October 27. Available at: https://www.oracle.com/news/announcement/ai-world-oracle-reinvents-enterprise-computing-for-the-ai-era-with-openai-2026-10-27/

Oracle Corporation (2026f) Oracle Unveils Agentic AI Platform to Accelerate Implementation of Oracle Fusion Cloud Applications. Press release, October 26. Available at: https://www.oracle.com/news/announcement/ai-world-oracle-unveils-agentic-ai-platform-to-accelerate-implementation-of-oracle-fusion-cloud-applications-2026-10-26/

Oracle NetSuite (2026) NetSuite Introduces SuiteExpress to Accelerate ERP Implementation and Continuous Improvement with AI. Press release via PR Newswire, October 26. Available at: https://www.prnewswire.com/news-releases/netsuite-introduces-suiteexpress-to-accelerate-erp-implementation-and-continuous-improvement-with-ai-302919135.html

Oracle documentation

Oracle Help Center (n.d.) Use Oracle Fusion Claw to create autonomous outcomes with agentic applications. Release 26D Common Technologies and User Experience What’s New, feature f51374. Available at: https://docs.oracle.com/en/cloud/saas/readiness/common/26d/common26d/26D-common-wn-f51374.htm

Oracle blogs

Batta, K. (2026) Building the Next Decade of Oracle Cloud Infrastructure. Oracle blog, October 27. Available at: https://www.oracle.com/news/announcement/blog/ai-world-building-the-next-decade-of-oracle-cloud-infrastructure-2026-10-27/

Oracle executives and customers in named media

du Preez, D. (2026) Oracle AI World 26 – HCM customers want agents, but process, data and legal sign-off come first. diginomica, October 28. Available at: https://diginomica.com/oracle-ai-world-26-hcm-customers-want-agents-process-data-and-legal-sign-come-first

Essex, D. (2026) Oracle’s agentic AI outpacing customers’ ability to adopt it. TechTarget, October 27. Available at: https://www.techtarget.com/news/oracles-agentic-ai-outpacing-customers-ability-to-adopt-it/872938/

Shrivastav, U. (2026) ‘Understand The Security Threat Landscape From AI And Mitigate Against That’: Sam Altman. BW Businessworld, October 27. Available at: https://www.businessworld.in/article/understand-the-security-threat-landscape-from-ai-and-mitigate-against-that-sam-altman-630755

Research this note builds on

Greyhound Consilium (2026) The Delegation Dial – How Much Authority To Hand Fusion Claw, Process By Process. Practical guide, developed in partnership with Oracle, October 23. Available at: https://greyhoundresearch.com/the-delegation-dial-how-much-authority-to-hand-fusion-claw-process-by-process/

Greyhound Research (2026a) Panic Has A Shelf Life: Why Greyhound Research Remains Bullish On IBM After July 14. Research note, July 17, including findings from Greyhound CIO Pulse 2026. Available at: https://greyhoundresearch.com/panic-has-a-shelf-life-why-greyhound-research-remains-bullish-on-ibm-after-july-14/

Greyhound Research (2026b) Everyone Is Building AI Agents. Oracle Is Moving The Contest Inside The Runtime. Research report, July 20, including findings from Greyhound CIO Pulse 2026. Available at: https://greyhoundresearch.com/everyone-is-building-ai-agents-oracle-is-moving-the-contest-inside-the-runtime/

Greyhound Research (2026c) Oracle Fusion Claw – Autonomy Arrives With A Receipt. Research report, October 14. Available at: https://greyhoundresearch.com/oracle-fusion-claw-autonomy-arrives-with-a-receipt/

↑ Back to the section guide

Disclosures: Independence, Financial, Intended Use, Methodology, Authoring, And Copyright

The evidence in this research note is current through October 29, 2026. Any written response received from an organization named in it after that date is dated where it appears.

Independence Disclosure

Research under the Greyhound Research brand is independent, and no organization holds a right of approval over its content or publication. Greyhound Research did not offer any named organization a pre-publication factual review of this research note. Editorial control remained entirely with Greyhound Research throughout. Greyhound Consilium is the B2B thought-leadership arm of The House Of Greyhound and a sister brand to Greyhound Research. It builds on Greyhound Research’s independent research through collaborations that produce thought-leadership content, research, and practical toolkits for decision-makers. Every Greyhound Consilium publication names its partner on the face of the document. This research note is published under the Greyhound Research brand.

Financial Disclosure

Greyhound Research undertakes advisory and consulting engagements with enterprise buyers and technology providers, and no such engagement confers control over its research findings or its decision to publish them. Oracle is a client of Greyhound Research (part of The House Of Greyhound). The engagement covers only thought-leadership content published under the Greyhound Consilium brand and did not fund this research note. Greyhound Research holds no other financial interest in Oracle or its subsidiaries.

Intended Use

This research note is written for chief information officers, heads of enterprise applications, procurement leaders who contract implementation services, general counsel and internal audit leaders who oversee implementation governance, and the press. It reads Oracle’s implementation-related announcements at Oracle AI World 2026 to inform implementation, sourcing, and governance decisions. This research note is not investment advice, a recommendation to buy, sell, or hold any security, or a view on any company’s share price or valuation. Its assessments rest on the documents listed in Sources And References as they stood on October 29, 2026, and can change as Oracle and others publish more. Product availability, features, and terms are described as Oracle published them through October 29, 2026, so confirm them with your Oracle account team before acting on them.

Research Methodology

Greyhound Research’s analysis rests on source evidence verified against the document that supports each claim. The evidence comprises the sixteen sources listed in Sources And References, in five classes, from Oracle’s announcements, documentation, and blog to executives’ and customers’ statements in named media and the research this note builds on. A finding that something is not described is bounded to the releases the text names.

Every customer quoted is named and linked where the quotation appears, either in an Oracle release or in published coverage of Oracle AI World. No composite or invented customer appears. The Greyhound CIO Pulse 2026 figures are cited as published in Greyhound Research’s research note of July 17, 2026, and research report of July 20, 2026. No Greyhound Fieldnotes are cited, and no factual claim in this research note rests on unpublished material. Quotations are checked against the live source before publication.

Authoring Disclosure

This material reflects the analysis, viewpoint and editorial judgment of Greyhound Research. The position presented is developed from Greyhound Research’s research, advisory work and accumulated institutional knowledge. Depending on the assignment, AI may be used to retrieve relevant material; structure, draft or polish prose from analyst-provided concepts, dictation, notes, rough drafts and source material; support secondary research and factual cross-checking; and assist with editing, condensation and consistency with Greyhound Research’s house style. AI does not independently establish Greyhound Research’s viewpoint, analytical position or conclusions. Material factual claims are verified against source evidence before release. Greyhound Research reviews and approves the final material and retains responsibility for it.

Copyright Policy

All content in this document and on the Greyhound Research website is protected by copyright law and may not be reproduced, distributed, transmitted, displayed, published, or broadcast without the prior written permission of Greyhound Research or, in the case of third-party materials, the prior written consent of the copyright owner of that content. You may not alter, delete, obscure, or conceal any trademark, copyright, or other notice appearing in any Greyhound Research content. Quotation of up to one hundred words is permitted where it carries attribution to Greyhound Research and a link to the published page. Reproduction, redistribution or republication beyond that extent, in whole or in part, by email or in any medium including websites, newsletters and intranets, requires prior written permission, which may be requested at the address below. The published version of each content piece on the Greyhound Research website carries sharing tools for readers who wish to share it.

For questions arising from these disclosures, including licensing requests and media inquiries, contact the Greyhound Research Community Relations Team at connect@thofgr.com.

↑ Back to the section guide