MANILA, Philippines — PLDT Inc. is pushing artificial intelligence beyond chatbots and office productivity tools, deploying AI agents that can retrieve company knowledge, accelerate customer responses and assist with complex risk assessments as demand for digital connectivity continues to climb.
The Philippine telecommunications group has expanded its use of UiPath’s automation and business-orchestration platform across several operational workflows, putting so-called agentic AI—systems designed not only to generate answers but also to perform parts of a task or workflow—to work inside the organization.
The most striking numbers show just how aggressively PLDT is trying to compress work that once took hours or even days.
Its KAI knowledge system can reportedly turn a research process that once required as long as five days into a contextual response within one to three seconds. PLDT estimates the system can eliminate 25,000 to 30,000 hours of manual work annually, creating productivity capacity roughly equivalent to 12 full-time employees.
But KAI is only one part of the story.
Three AI systems, three very different jobs
KAI, short for Knowledge, Automation and Intelligence, is designed to retrieve information from approved company sources for tasks including proposal development and customer support.
Unlike a general-purpose chatbot pulling information from across the internet, the system is grounded in sources approved by PLDT, an important distinction for a telecommunications company handling internal operational and commercial information.
TechNode Global reported that the assistant was originally developed by PLDT’s network team using Microsoft Copilot before being migrated to and enhanced through UiPath and integrated with Microsoft Teams. Its users include enterprise sales, pre-sales and network teams.
Then there is Ellie, or the Enablement and Intelligence Engine.
Designed for PLDT Enterprise employees, Ellie connects internal information and workflows so staff can move more quickly from finding information to acting on it.
According to figures released by PLDT and UiPath, Ellie has reduced manual effort by roughly 25% to 35%, saved an estimated 18,000 to 25,000 work hours annually, and shortened certain customer response times by as much as 80%.
The third system may be the clearest illustration of what makes agentic AI different from conventional workplace chatbots.
PLDT’s ERICA, or Enterprise Risk Intelligence Companion Agent, can synthesize information, score risks and recommend possible treatments for enterprise risk assessments.
Deployed in February 2026, ERICA has reportedly compressed evaluations that previously required anywhere from two to 10 days into periods ranging from about five minutes to one day.
PLDT estimates that this represents a 97% to 99% reduction in manual effort for the affected processes.
PLDT’s AI strategy is becoming much bigger than UiPath
The latest rollout is not an isolated technology experiment.
PLDT and wireless subsidiary Smart Communications have been steadily assembling an AI ecosystem involving several of the world’s largest enterprise-technology companies.
In August, the group disclosed that it had deployed 1,000 Microsoft 365 Copilot seats, incorporating AI into activities such as drafting communications, summarizing meetings, analyzing information and supporting internal decision-making. Microsoft portrayed PLDT and Smart as examples of companies attempting to move AI from limited experiments into everyday workflows.
Customer-facing operations are also being automated.
In June, Amdocs and Amazon Web Services announced that PLDT Home had expanded an agentic AI customer-service system called Store Genie after an earlier deployment at Smart Communications.
The companies said the platform was operating across Smart and PLDT Home, supporting more than 750 frontline agents at 149 locations. They reported that some processes previously measured in hours had been reduced to minutes—including certain order inquiries that dropped from around 10 hours to about two minutes.
Those deployments suggest PLDT is pursuing AI on several fronts at once: employee productivity, enterprise knowledge, customer service, network-related support and corporate risk management.
Why telecom operators are racing toward AI
The timing is significant.
Telecommunications companies are dealing with more complicated product portfolios as fiber broadband, 5G, cloud services, cybersecurity, satellite connectivity and enterprise digital services become increasingly intertwined.
Every additional service can also create more customer questions, technical procedures, internal documentation and operational dependencies.
That makes the economics of automation increasingly attractive: rather than simply adding employees every time workloads increase, companies can attempt to automate repetitive research and routine service processes while directing human workers toward exceptions, customer relationships and higher-value decisions.
PLDT and UiPath specifically linked the current expansion to growing demand associated with 5G adoption, broadband usage and satellite connectivity.
The company’s broader infrastructure strategy is also increasingly tied to AI demand. In August, Manila Bulletin reported that PLDT’s VITRO data-center business was studying a new facility in Southern Luzon with at least 100 megawatts of operational capacity, as the company positions its infrastructure for greater computing requirements from hyperscalers and artificial intelligence workloads.
The numbers are impressive—but they need context
There is an important qualification behind PLDT’s eye-catching productivity figures.
The estimates of 25,000 to 30,000 hours saved through KAI, the 18,000 to 25,000 hours attributed to Ellie and ERICA’s 97% to 99% reduction in manual effort come from PLDT and UiPath.
TechNode Global noted that the companies have not publicly disclosed enough detail to independently reproduce every calculation, including the full measurement methodology, user volumes and whether some productivity benefits attributed to the different systems overlap.
That does not make the gains insignificant.
It means the figures should be presented as company-reported operational results rather than independently audited performance measurements.
And that distinction could become increasingly important as corporations race to demonstrate a return on billions of dollars being spent globally on generative and agentic AI.
The bigger bet
What makes PLDT’s strategy worth watching is no longer whether the company is experimenting with AI.
It clearly is not.
From Microsoft Copilot and UiPath agents to Amdocs and AWS-powered customer-service automation, PLDT is beginning to distribute AI across different layers of the business rather than relying on one giant system to do everything.
The more consequential test will come next: whether the massive reductions in processing time eventually translate into improvements that customers can consistently feel—faster repairs, quicker answers, fewer escalations and better service.
Because cutting a five-day internal task to three seconds sounds dramatic.
The real measure of PLDT’s AI transformation will be what happens during all the time it just freed up.
Leave a Reply