{"id":139550,"date":"2026-10-08T07:25:19","date_gmt":"2026-10-08T06:25:19","guid":{"rendered":"https:\/\/www.intelligentcio.com\/me\/?p=139550"},"modified":"2026-10-08T07:25:19","modified_gmt":"2026-10-08T06:25:19","slug":"the-next-phase-of-enterprise-ai","status":"publish","type":"post","link":"https:\/\/www.intelligentcio.com\/me\/2026\/10\/08\/the-next-phase-of-enterprise-ai\/","title":{"rendered":"The next phase of enterprise AI"},"content":{"rendered":"
\n
\"\"<\/figure><\/div>\n\n\n

At AI Everything Abu Dhabi, Aarti Mohan, Vice President \u2013 Business Development, Middle East and Africa, Oracle, discusses what is holding back production deployments, the rise of agentic AI and why cost, culture, governance and sovereignty will increasingly shape enterprise AI strategies.<\/em><\/p>\n\n\n\n

Why are so many organisations still struggling to move AI from experimentation into production at scale?<\/strong><\/p>\n\n\n\n

I personally think that companies have moved beyond the pilot phase and are now trying to move into production. There are a couple of barriers. One is understanding where the business outcome is and identifying impactful use cases that can go into production. Not all use cases will make it into production.<\/p>\n\n\n\n

The second is the cultural shift within the organisation and among the people involved. That is another element that may be delaying the move into production.<\/p>\n\n\n\n

The third aspect is cost sensitivity around how we are consuming tokens. All the agentic applications and large agentic workflows that we are trying to automate use tokens, and I think organisations want greater transparency around those costs before putting things into production.<\/p>\n\n\n\n

So, it is a matter of time. It will happen very soon, but I think these are the top three barriers today.<\/p>\n\n\n\n

Customer service and chatbots were among the early, low-hanging use cases for AI. Are you now seeing organisations apply AI to more complex, business-critical processes?<\/strong><\/p>\n\n\n\n

Yes, we are. A couple of our large customers, across both the public and commercial sectors, have implemented a number of use cases around back-office automation. These include finance reporting, sourcing and supplier qualification.<\/p>\n\n\n\n

In HR, we have seen a lot of uptake because these tend to be text-heavy use cases. These include performance reviews, approving expenses and providing justification, or looking at goals and objectives for team members. Those are some of the top use cases.<\/p>\n\n\n\n

We are also starting to see use cases around document understanding. For example, when you receive a large number of files from suppliers as part of the qualification process, you can use AI to read and understand those documents. Similarly, organisations can scan invoices and expenses using AI and put them in the right place. We are seeing more and more uptake in these areas.<\/p>\n\n\n\n

At Oracle, we have implemented AI in several areas as well, starting with customer service, of course. Many of our support responses are automated, at least at the first level of support, and the turnaround time has gone from a few minutes to a few seconds. We are not talking about hours. We are talking about reducing a few minutes to a few seconds.<\/p>\n\n\n\n

We have also opened up AI and LLMs to our developers. More than 26,000 developers are now using LLMs to code, assist with coding and documentation, help with bug fixes and so on.<\/p>\n\n\n\n

As foundation models become increasingly commoditised, where will meaningful differentiation come from for enterprise customers?<\/strong><\/p>\n\n\n\n

I think the differentiation going forward will come from the regulatory element. For example, I may want an LLM in my language and in my country.<\/p>\n\n\n\n

Does that mean sovereignty and localisation could ultimately become more important differentiators than the underlying model itself?<\/strong><\/p>\n\n\n\n

I think so. I think it will come more from a sovereignty perspective and a language perspective. If I am in this region, for example, I may want an Arabic LLM provider that is based here. If I am in another country, I may want a country-specific language model based in that country.<\/p>\n\n\n\n

That’s where it is heading. Of course, as regulation evolves, we will increasingly see which providers are able to scale globally versus providers that are based in one region or continent while serving customers globally.<\/p>\n\n\n\n

Building dedicated AI infrastructure can require significant upfront investment, particularly in GPUs. Are customers therefore becoming more receptive to shared AI infrastructure, and how do they reconcile that with sovereignty and compliance requirements?<\/strong><\/p>\n\n\n\n

Yes, there are customers looking at shared infrastructure. We have our sovereign cloud offering. The sovereign cloud essentially serves customers within the region we are trying to address, and the infrastructure is built with an anchor provider.<\/p>\n\n\n\n

In the UAE, for example, we have Core42 serving the government. We also have du serving some public sector and commercial entities in the UAE. They use shared infrastructure but serve each of these clients and customers individually.<\/p>\n\n\n\n

It helps with cost, scale and time to market. It is also located within the country, so the sovereignty requirements can be met.<\/p>\n\n\n\n

Agentic AI takes automation considerably further by allowing AI to act rather than simply provide answers. Have agents reached the point where enterprises can confidently entrust them with business-critical processes?<\/strong><\/p>\n\n\n\n

I think there is very little choice. The agents are already here.<\/p>\n\n\n\n

Which business processes are mature enough today to be handed over, at least partially, to AI agents?<\/strong><\/p>\n\n\n\n

There are many. We have launched around 20 agentic applications, covering processes across customer service, supply chain, finance and HR, so they span end-to-end processes.<\/p>\n\n\n\n

For example, if I wanted to automate my financial closing process, I would only want to share the exceptions with the user. I would want to automate all the reconciliations, automate the subledger closing and then publish the results. The person preparing the reports would simply review them and see whether there is anything that requires further attention.<\/p>\n\n\n\n

As agents gain greater autonomy and begin acting across enterprise systems, how do you govern their actions and ensure organisations retain visibility and control?<\/strong><\/p>\n\n\n\n

Yes, we have to. Governance has to be part of it. We have solutions around governance, risk and compliance that are also enabled with AI models. These not only cover our applications but extend beyond them.<\/p>\n\n\n\n

I think every customer today has a heterogeneous environment. It is not all Oracle. There may be Oracle technologies as well as non-Oracle technologies. The governance layer therefore has to span the customer’s entire technology footprint.<\/p>\n","protected":false},"excerpt":{"rendered":"

At AI Everything Abu Dhabi, Aarti Mohan, Vice President \u2013 Business Development, Middle East and Africa, Oracle, discusses what is holding back production deployments, the rise of agentic AI and why cost, culture, governance and sovereignty will increasingly shape enterprise AI strategies. Why are so many organisations still struggling to move AI from experimentation into […]<\/p>\n","protected":false},"author":7289,"featured_media":139552,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[809],"tags":[3211,711],"class_list":["post-139550","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-more-news","tag-ai","tag-oracle"],"acf":[],"publishpress_future_workflow_manual_trigger":{"enabledWorkflows":[]},"_links":{"self":[{"href":"https:\/\/www.intelligentcio.com\/me\/wp-json\/wp\/v2\/posts\/139550","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.intelligentcio.com\/me\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.intelligentcio.com\/me\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.intelligentcio.com\/me\/wp-json\/wp\/v2\/users\/7289"}],"replies":[{"embeddable":true,"href":"https:\/\/www.intelligentcio.com\/me\/wp-json\/wp\/v2\/comments?post=139550"}],"version-history":[{"count":1,"href":"https:\/\/www.intelligentcio.com\/me\/wp-json\/wp\/v2\/posts\/139550\/revisions"}],"predecessor-version":[{"id":139553,"href":"https:\/\/www.intelligentcio.com\/me\/wp-json\/wp\/v2\/posts\/139550\/revisions\/139553"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.intelligentcio.com\/me\/wp-json\/wp\/v2\/media\/139552"}],"wp:attachment":[{"href":"https:\/\/www.intelligentcio.com\/me\/wp-json\/wp\/v2\/media?parent=139550"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.intelligentcio.com\/me\/wp-json\/wp\/v2\/categories?post=139550"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.intelligentcio.com\/me\/wp-json\/wp\/v2\/tags?post=139550"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}