PLDT claims AI agents save tens of thousands of work hours annually

A telecom AI story measured in work hours

PLDT’s latest automation milestone is not framed as a chatbot launch or a one-off proof of concept, but as a productivity claim: the Philippine digital services provider says a fleet of in-house AI agents is saving tens of thousands of manual work hours annually across business functions [1]. The systems, built with UiPath Agent Builder, Integration Service and Robots, are deployed in enterprise sales, knowledge retrieval, risk management and lead-generation workflows, according to UiPath’s October 5 announcement [1].

The company also received a Silver Award at the 2026 UiPath AI Breakthrough Awards for the implementation, a recognition UiPath says is tied to enterprise-scale use of agentic AI, business orchestration and intelligent workflows [1]. TNGlobal’s report on the same announcement described the deployment as spanning enterprise sales, knowledge retrieval and risk management, while noting that the productivity figures were supplied by PLDT and UiPath and were not independently verified by the publication [2].

That caveat matters. The numbers are striking, but they remain company-reported metrics rather than audited productivity data. Still, the level of detail released by PLDT and UiPath gives a useful view into where large organizations are currently finding value from AI agents: not in replacing entire departments, but in compressing slow, repetitive, information-heavy workflows.

Ellie: proposal work, CRM context and sales follow-through

The most clearly defined agent in the rollout is Ellie, short for Enablement and Intelligence Engine [1]. PLDT describes Ellie as an AI-powered assistant for enterprise sales that brings together internal knowledge and live customer relationship management data, giving sales teams usable intelligence inside proposal and presentation workflows [1].

According to UiPath, Ellie automatically generates tailored, customer-ready proposals and presentations, reducing time-intensive work for sales support teams [1]. PLDT estimates that this workflow alone is reclaiming about 18,000 to 25,000 hours of work per year [1]. TNGlobal reported the same annual savings range and said sales teams saw a 40 percent to 60 percent reduction in time spent developing standard proposal and presentation materials [2].

The company also reported a 40 percent to 50 percent reduction in proposal error rates and a 20 percent to 50 percent increase in upsell rates [1]. InsiderPH highlighted the same set of metrics, emphasizing Ellie’s role in cutting standard proposal and presentation development time and supporting higher upsell performance [3].

The sales use case is important because it sits at the intersection of data retrieval, content generation and customer context. In many enterprise environments, proposal work is repetitive but not simple: teams must locate the right product details, align pricing or service information, reflect the customer’s current relationship with the provider and avoid errors. PLDT’s claim suggests that an agentic layer can do more than answer questions; it can assemble usable work products while connecting to operational systems.

Ellie is also described as a strategic co-pilot that monitors customer-health indicators [1]. UiPath said the agent uses predictive analytics to detect churn risks based on signals such as declining product utilization or delayed payments, then prompts relationship managers with alerts, retention recommendations and upsell opportunities [1]. InsiderPH also reported that Ellie monitors those customer health indicators and alerts relationship managers to retention and upsell opportunities [3].

That makes the system more than a document generator. If it functions as described, Ellie connects pre-sales productivity with account management, using customer data to influence timing and next actions. For a telecommunications company serving enterprise accounts, that kind of workflow can matter because sales, retention and service quality often depend on whether account teams can act before a customer issue becomes a lost contract.

KAI: turning days of research into seconds

PLDT’s second named system is KAI, or Knowledge, Automation, Intelligence [1]. UiPath says KAI transforms knowledge retrieval from a manual research process that could take up to five days into contextual responses delivered in one to three seconds [1]. TNGlobal reported the same shift from multi-day research to near-instant contextual responses [2].

The stated productivity gain is substantial. UiPath estimates that KAI reduces 25,000 to 30,000 hours of manual effort annually and creates around 12 full-time-equivalent units of annual productivity capacity [1]. InsiderPH also reported that KAI saves an estimated 25,000 to 30,000 hours annually by reducing knowledge retrieval times [3].

The key phrase in UiPath’s description is that KAI is “grounded exclusively in approved sources” [1]. In a telecom setting, that detail is not cosmetic. Knowledge retrieval can touch product specifications, service commitments, compliance information, technical documentation and customer-support material. If an AI assistant retrieves or synthesizes information from unapproved or outdated sources, the speed gain can quickly become a risk.

KAI’s value, then, rests on two claims at once: speed and control. A system that returns answers in seconds may improve responsiveness, but a system grounded in approved sources is what makes the answer usable inside a regulated, customer-facing business process. PLDT’s reported 99 percent faster response rate suggests the company is measuring the tool not only as a convenience for employees, but as a capacity engine for proposal creation and customer support [1].

ERICA: risk assessment as an agentic workflow

The third named agent, ERICA, stands for Enterprise Risk Intelligence Companion Agent [1]. UiPath says ERICA automates structured risk statements and scoring against compliance standards [1]. The company reported that ERICA has reduced manual effort by 97 percent to 99 percent and compressed risk assessments from two to 10 days down to five minutes to one day [1].

TNGlobal and InsiderPH both reported the same 97 percent to 99 percent reduction in manual risk-assessment effort [2][3]. TNGlobal also noted that ERICA shortens risk assessments from two to 10 days to between five minutes and one day [2].

This is one of the more consequential parts of the announcement because risk assessment is not merely administrative. It is the kind of process where speed, documentation and consistency all matter. Manual risk reviews can be slow because they require teams to gather evidence, interpret standards, score exposure and translate findings into structured statements. An AI agent that can draft, score and organize that work may not remove human accountability, but it can change where the human effort is concentrated.

That distinction is central to understanding the PLDT example. The company is not simply saying that AI produces text faster. It is saying that agents are being inserted into operational workflows where employees previously spent days collecting, checking and formatting information. In ERICA’s case, the productivity story is about compressing the preparation and structuring of risk work so human teams can focus more on judgment and treatment decisions.

Why the announcement matters beyond PLDT

The PLDT case is significant because it shows agentic AI being evaluated in operational terms: hours reclaimed, response times reduced, errors lowered and assessment cycles shortened. UiPath framed the implementation as part of a shift from conventional automation toward AI agents, business orchestration and intelligent workflows at enterprise scale [1].

That framing reflects a broader enterprise question: what happens when AI is embedded not at the edge of work, but inside the systems that coordinate work? PLDT’s agents appear to sit between employees, internal knowledge bases, CRM data, compliance standards and automation tools. The value comes from orchestration: gathering the right data, triggering the right workflow, producing a usable artifact and prompting the right employee action.

For PLDT, the immediate claim is productivity. Ellie is tied to 18,000 to 25,000 reclaimed hours per year, while KAI is tied to 25,000 to 30,000 hours of reduced manual effort per year [1][2]. Together, those ranges alone point to a sizeable internal capacity gain before counting ERICA’s reported reduction in manual risk-assessment effort.

For other enterprises, the more important lesson may be the selection of use cases. PLDT is applying AI agents where information is fragmented, cycle times are long and employees need outputs they can act on. Sales proposals, knowledge retrieval and risk scoring are not identical processes, but they share a common problem: too much valuable employee time is spent searching, compiling and reformatting information.

The unresolved questions

The announcement also leaves several questions open. TNGlobal explicitly noted that the figures were supplied by PLDT and UiPath and were not independently verified [2]. The reports do not provide a full methodology for calculating hours saved, error-rate reductions or upsell changes. They also do not specify how PLDT measures human review, exception handling or long-term quality control across the agent outputs.

Those omissions do not negate the announcement, but they shape how it should be read. The story is best understood as a company-reported case study with concrete operational metrics, not as an independently audited benchmark for all telecom AI deployments.

The more durable point is that PLDT is presenting agentic AI as part of its operating model. Gilbert Gaw, PLDT and Smart first vice president and head of IT and transformation office, said the recognition reflects the teams’ work on Ellie, KAI and ERICA and the company’s drive to keep expanding what agentic AI can do for the business and customers [1][2]. UiPath’s Karl Crowther described PLDT as using AI agents and agentic business orchestration to change how work gets done [1].

If the reported gains hold up over time, PLDT’s example will stand less as a story about a single award and more as a sign of how enterprise AI is maturing. The next phase is not simply whether companies can deploy agents, but whether they can govern them, measure them and connect them to workflows where speed and accuracy both matter. On that test, PLDT is now offering one of the clearer telecom-sector claims: tens of thousands of work hours saved annually through AI agents embedded in core business processes.

Sources from the last 72 hours

1. [1]PLDT Wins Silver Award at 2026 UiPath AI Breakthrough Awards for Agentic AI-Driven Enterprise TransformationOct 5, 2026, 2:00 PM
2. [2]PLDT says AI agents save tens of thousands of work hours annuallyOct 5, 2026, 2:00 AM
3. [3]WINNING | PLDT's AI agents slash thousands of work hoursOct 5, 2026, 8:36 AM

AI-generated article based on recent web research, then preserved as a dated editorial snapshot.