When Agents Arrive with the Upgrade, Who's in Control?
"Most enterprises will meet their next agents without ever buying them. They'll arrive inside the ERP upgrade, the workflow platform, and the cloud contract. Oracle, ServiceNow and IBM all moved this month, and the pattern is the same: agents are becoming part of the tech stack you already run. That makes control a design job from day one. My advice to any leadership team: settle who approves an action, what an agent can access, where it runs, and what it costs before it switches on. The organisations that scale will be the ones that built control in. This issue maps who is already doing it."
CEO, Magure
Akhil brings 15+ years across growth, strategy, and transformation, and has led AI adoption programs for enterprises across the Middle East and global markets, with a focus on making AI accountable, observable, and cost-resilient at enterprise scale.
State of AI Play: In Figures
- of organisations expect to deploy agentic AI within the next 12 months
- 71%of organisations expect to deploy agentic AI within the next 12 monthsGartner
- of CEOs struggle to demonstrate AI-driven financial returns
- 44%of CEOs struggle to demonstrate AI-driven financial returnsGartner
- of IT leaders have limited or redesigned AI use cases because of rising token consumption
- 89%of IT leaders have limited or redesigned AI use cases because of rising token consumptionCloudera
- of security leaders identify attacks targeting AI systems as one of their biggest preparedness gaps
- 50%of security leaders identify attacks targeting AI systems as one of their biggest preparedness gapsPwC
- of IT leaders say data residency, sovereignty or jurisdiction-specific rules influence where AI workloads are deployed
- 84%of IT leaders say data residency, sovereignty or jurisdiction-specific rules influence where AI workloads are deployedCloudera
- of Global 2000 enterprises will be able to continuously verify their AI agents' decisions by 2028, up from 14% today
- 70%of Global 2000 enterprises will be able to continuously verify their AI agents' decisions by 2028, up from 14% todayIDC
The AI Deployment-Oversight Gap
| What Enterprises Report | Where It Breaks Down |
|---|---|
| 98% report having formal AI governance policies in place. | 47% say their organisation has previously not followed its AI governance process for urgent deployments. |
| 91% report their organisation uses agentic AI, in pilots or full deployment. | 49% of those say their governance framework hasn't been updated to include agentic AI requirements and risks. |
| 85% of agentic AI users say at least a handful of their systems execute actions without real-time human involvement. | 26% of them cannot detect unauthorised AI agents operating internally. |
| 98% conducted a formal AI assurance review at least annually. | 36% experienced an AI incident or failure that caused a materially negative impact. |
Source: EY
Global AI Moves
ServiceNow and OpenAI Connect Agents to Enterprise Workflows
On October 28, ServiceNow said OpenAI agents can reach its workflows through Action Fabric, and its MCP plugin in the OpenAI marketplace lets teams act on records from inside ChatGPT. Codex is expected to join the AI Workflow Factory in 2027.
Why it matters
Only the plugin is live. The question is who owns the workflow when an outside model triggers it. ServiceNow says it stays the control point, so test that in a pilot before agents touch production records.
Cognizant and Google Cloud Expand Agentic Enterprise Transformation Alliance
On October 19, Cognizant and Google Cloud launched a Gemini Frontier Capability Group, backed by more than 18,000 Cognizant associates. It pairs Google's AI platform with Cognizant's industry and engineering expertise, with first solutions planned for retail, healthcare and manufacturing.
Why it matters
Agentic AI delivery is shifting toward industry-specific implementation capacity, not generic model access. This matters for enterprises that need domain workflows, integration and change management, not just a model API. If you're buying from a services partner, ask how many trained people stand behind the platform and which solutions are live, not just planned.
OpenAI's GPT-6 Models Come to Google Cloud
On October 21, Google Cloud said GPT-6 models are coming to Model Garden in private preview for select customers. Usage is billed through Google Cloud and can draw down existing commitments.
Why it matters
This reduces platform lock-in and lets enterprises combine Google's data and infrastructure with OpenAI models inside an existing cloud estate. For buyers, it makes multi-model procurement more practical.
Oracle Puts 25 Agentic Applications Inside Fusion
Oracle announced that 25 Claw-powered agentic applications are available across Fusion, so agents now work inside the finance, HR, supply chain and customer experience apps enterprises already run.
Why it matters
Agents are arriving through the ERP upgrade cycle. Ask your vendors which will switch on by default, and who signs off before they act.
Google Introduces Gemini Agent for Work
Google introduced the Gemini agent as a single agent for knowledge work, media creation and writing and running code. It is in preview for Financial Services and Legal, with Government, Healthcare and Retail coming soon.
Why it matters
This points to a future where one assistant operates across everyday work rather than employees switching among specialised AI tools. For CEOs and functional leaders, the opportunity is productivity consolidation, but governance and access design become more important.
HPE and NVIDIA Target Secure Agentic AI in Hybrid Environments
HPE expanded its AI Factory with NVIDIA portfolio to help enterprises run autonomous AI across hybrid and sovereign environments, with controls over performance and data access. The release cites role-based access, machine identity support and audit logging for human and agent access. Some pieces are available now, while the self-service agent tools arrive in 2027.
Why it matters
Agentic AI adoption will depend as much on infrastructure, sovereignty and operational control as on model capability. For CTOs and CIOs, this is a signal that secure agent deployment must be designed into the estate from day one.
IBM Gives Agents Governed Identities
IBM's watsonx Orchestrate update, announced October 27, ties agents to governed identities through Microsoft Entra, IBM Verify and IBM Agent Identity, so each action traces to an accountable identity.
Why it matters
Enterprises cannot scale autonomous agents without knowing which agent owns which action, what it can access and how to audit its behavior. This makes agent identity and control a first-class enterprise requirement.
Around the World - In Brief
- Microsoft said it will scale its Azure Maia chips to multi-gigawatt deployments from 2027, making Maia its primary inference system. It expects Maia 300 to deliver 40% more tokens per megawatt than leading GPU systems, a company claim.
- HCLTech and Anthropic signed a global premier partnership to scale governed agentic AI and plan a dedicated Anthropic business unit and Centre of Excellence.
- SAP enhanced Joule Studio, Joule Work and the SAP Business AI Platform with agentic planning, richer business context and governance controls for multistep workflows.
- Snowflake and Oracle announced Snowflake as a native, first-party service on Oracle Cloud Infrastructure. Customers will be able to buy it through Oracle and apply Oracle Universal Credits to Snowflake usage, so data and AI spend can sit under one contract.
- Meta and Sierra are developing the Personal Agent Protocol, an open standard for how consumers' personal agents interact with businesses. It uses OAuth to authorise access, gives companies visibility into what agents do, and lets them set parameters for what agents can do. Genesys, Shopify, Stripe and Walmart are among the partners.
- Frontier-model releases - Google began a limited rollout of Gemini 4 Carbon for cybersecurity and long-horizon agentic work, OpenAI released GPT-6.1 Terra for efficient professional work and computer use, and Meta introduced Muse Wonder and Muse Wonder Full for agentic tasks.
UAE AI Watch: The UAE Is Taking Agentic AI Into the Private Sector
1. 11 Dubai Companies Pledge to Integrate Agentic AI
Eleven Dubai-based companies pledged on October 12 to move from exploring agentic AI to adopting it in service development, internal processes and decision-making, with Dubai Chambers' support. Signatories include Emirates, DP World, Dubai Islamic Bank, Aster DM Healthcare and Kitopi.
2. Emirates NBD Brings OpenAI Models In-Country
Emirates NBD and OpenAI announced a collaboration on October 19 that the bank says makes it the first in MENA to deploy OpenAI GPT models using in-country inference infrastructure in the UAE. It plans to use the models for customer engagement, risk and compliance analytics, knowledge-work automation and employee productivity.
3. Dubai Launches an Agentic AI Accelerator for Government Services
Dubai Future Foundation launched the Agentic AI for Government Services Accelerator on October 5. In its first phase, 33 Dubai government entities will assess more than 300 public-facing use cases, with technology companies certified under the Dubai AI Seal.
4. Dubai Opens an AI Sandbox for Regulator-Supervised Testing
Dubai Future Foundation launched an AI Sandbox on October 28 so companies can test AI solutions with regulators and work through the regulatory barriers to deploying them. Applications are open until December 18.
GCC AI Watch: GCC Governments Set the Rules and the Targets for AI Adoption
1. Qatar Creates a National Center for AI
Qatar's Amir ratified Cabinet Decision No. 34 of 2026 around October 19-20, establishing a National Center for AI under the Ministry of Communications and IT, covering the national AI strategy, regulatory frameworks and adoption support.
2. Saudi Arabia Opens a Personal-Data Sandbox for Regulated Testing
SDAIA opened a two-month registration in early October for a Personal Data Protection Track in its Data Regulatory Sandbox, so licensed private entities can test products that process personal data under supervision and align them with the Personal Data Protection Law.
3. Kuwait's Regulator Calls for Clear Permissions and Accountability in Agentic AI
At the October 7 launch of The Agentic AI Experience 2026, CITRA's chairman said agentic AI's permissions must be clearly defined and responsibility for its actions established, and that its results should stay under human oversight. Zain Kuwait's CEO stressed data protection and data sovereignty at the launch.
4. Oman Launches Phase Two of Its Digital Economy Programme
Oman announced phase two of its National Digital Economy Programme, covering 2026-2030. It targets OMR 1.5 billion in added GDP, OMR 300 million in technology investment and 3,000 new direct jobs a year, with accelerating AI adoption one of its four focus areas.
5. World Bank Calls Saudi Arabia's AI Governance a Benchmark
The World Bank's October MENA report gave SDAIA a case study and described the Saudi approach as a benchmark for institutional models for data and AI governance.
Asia-Pacific AI Watch: Asia-Pacific Turns AI Guidance Into Supervision, Standards and Infrastructure
1. Singapore's MAS Issues AI Risk Management Guidelines
MAS issued Guidelines on AI Risk Management on October 7 for all financial institutions, covering board oversight, AI inventories and third-party AI. Sections 3-4 apply from October 7, 2027, and Sections 5-6 by October 7, 2028.
2. Japan Publishes Agent Safety Standards as It Rolls Out Agents to Government Staff
Japan's Digital Agency began rolling out Agent Chat and published its GENAI Standards on October 20. The standards, developed with the AI Safety Institute, define compliance requirements for users, administrators and developers. About 180,000 government employees are expected to have access to government generative AI in fiscal 2026.
3. Singapore Backs Companies and SMEs to Put AI to Work
At SWITCH 2026 on October 27, Deputy Prime Minister Gan Kim Yong outlined new initiatives to help companies adopt and scale AI, including Champions of AI for a first cohort of 24 companies and agentic AI workshops for small businesses. They build on the four national AI missions in advanced manufacturing, finance, healthcare and connectivity.
4. Tata Communications Plans an AI-Ready Network Across India
Tata Communications announced a planned US$400-450 million investment in about 10,000 km of new AI-ready connectivity linking India's data-centre and AI hubs, starting with the Mumbai-Pune-Hyderabad-Chennai corridor in the first of three phases.
Your Move: Four Actions for this Month
CIOs & CTOs
Ask every vendor who signs off before agents switch on. Oracle is adding 25 agentic apps to Fusion, ServiceNow is opening workflows to OpenAI agents and IBM is giving agents their own identities. List which are live and which are roadmap, and who approves what before your next renewal.
AI & Innovation Leaders
Test one model swap on a real workflow. With OpenAI models heading to Google Cloud, model choice is becoming a procurement decision. Run your current model and one alternative on the same cases, and compare accuracy, time saved, human corrections and total cost before you switch.
Compliance & Risk
Start your evidence file before the deadline does. MAS is phasing in expectations on AI inventories and third-party AI from October 2027, and Kuwait's regulator wants agent permissions defined up front. List every agent, who owns it, and what it can touch.
Digital Transformation
Pick your next use case by what it can reuse. Another standalone tool adds another set of controls to build. Choose the workflow that can inherit the access, audit and approval steps you've already set up.
If you're testing a model swap, mapping agent permissions, or building your first governed workflow, MagOneAI Studio is where teams are doing this in production - with the access controls, audit trail and cost tracking built in.
Build + Manage + Govern AI Lens
This month, agents stopped being something you buy and started being something you inherit. Oracle embedded 25 inside Fusion. ServiceNow opened its workflows to OpenAI. IBM gave agents their own governed identities. The pattern is the same: agents are arriving in the platforms you already run, whether you planned for them or not.
That changes the job. You are no longer deciding whether to build an agent - you are deciding how to deploy the ones arriving in your next upgrade, how to manage their cost and performance across a stack you did not design for them, and how to govern what they are allowed to do before they switch on. The enterprises that move fastest will be the ones that treat this as one connected problem, not four separate ones.
- BUILD
- Choose the right models, data and infrastructure for the work.
- DEPLOY
- Put agents into real workflows with clear boundaries and human checkpoints.
- MANAGE
- Track performance, usage, cost and business value once they are live.
- GOVERN
- Control access, decisions and actions with auditability and accountability built in.
A governed agent in one workflow teaches more than ten pilots without controls. Build the first one so the second does not mean rewiring the rules.
Resource of the Month
Start Building: Your Agentic AI Learning Path
Every enterprise AI deployment starts with someone who builds the first one. Whether you are a developer inside an enterprise team, a student exploring agentic AI, or a team lead upskilling your squad - here is where to start:
The fastest way to prove agentic AI works for your organisation is to show it running. Forward this to the builder on your team who will make it happen.
That’s AI Radar · Issue 05
AI Radar is published monthly by Magure.
Next issue: November 2026, covering the latest in enterprise AI, agentic deployments, and regional signals.
Story suggestions? Reply to the issue email. We read every response.
How Did You Find This Issue?
Your feedback shapes the next issue.