From Hours to Outcomes: The Business Model That Survives AI

Somewhere in a tower in Bonifacio Global City or Cebu IT Park, someone is making a decision that no algorithm has been asked to make.

The caller’s account is technically current, but the record is broken — last payment posted to the wrong account, system hasn’t caught it, and the customer is furious about a service interruption they technically caused and technically didn’t. The worker on the phone reads the situation in about four seconds. She knows this caller type. She knows which way this goes if she follows the script and which way it goes if she uses judgment. She uses judgment. The call ends well. The ticket is closed. The system logs it as a resolved interaction, duration four minutes and eleven seconds.

The client has guardrails for this: scripts, escalation paths, approval limits all designed to ensure that judgment stays within bounds and liability stays manageable. But the contract does not pay for the four-second read, the pattern recognition, the accumulated knowledge of a thousand similar calls. It only pays for the hour.

That is not an oversight. It is the structure of the bargain. The client buys capacity — an hour of output, delivered within spec, at a price lower than they could source at home. What happens inside that hour, the routine and the edge case alike, is the provider’s problem to manage. The unit of sale was the seat. The person in the seat is, commercially speaking, the hour.

That equation built something enormous. IBPAP reports that the Philippine IT-BPM sector closed 2025 on track for roughly 1.9 million jobs and $40 billion in export revenue.¹ It is also an equation AI is now exposing.

AI does not have to eliminate the BPO industry to damage the labor-hour model. It only has to make the hour less central to the value being delivered. If software can resolve part of a customer-service workflow, triage a claims-intake queue, or let a QA team review exceptions instead of every transaction, the client eventually asks the obvious question: why am I still buying this mainly as headcount?

The industry priced around labor input. AI gives buyers new ways to measure output. That gap is the problem providers now have to solve. Ultimately, they have to change what they sell.

The hourly model punishes its own future

In the hourly model, the provider wins by keeping utilization high, wages controlled, attrition manageable, and the account staffed. AI destabilizes that structure.

If an agent resolves the same work 20 percent faster, the client sees a reason to buy fewer hours. If an AI assistant reduces after-call work, the provider may improve delivery while weakening the commercial basis of the account. If automation handles the simple volume, the remaining human work becomes harder, more variable, and less compatible with factory-style productivity metrics.

Efficiency starts as a delivery improvement and turns into a revenue threat.

The industry already sees the pressure. The BSP’s 2025 economic newsletter on GenAI and Philippine IT-BPM says 67 percent of IT-BPM member firms had incorporated AI tools into their operations — while roughly 1.6 million employees, or 88.5 percent of the IT-BPM workforce, sit in contact center and business-process services, the segment most exposed to workflow automation and repricing.²

Near-term evidence remains mixed, and it should be read honestly. The same BSP paper says GenAI is expected to augment rather than displace jobs in the near term, and that full-scale adoption is unlikely within three years because of integration, scaling, and regulatory limits.² I think that timeline is too relaxed. The capex pattern of the large operators, the spread of labor arbitrage into Africa, and the pace of AI capability all point to a shorter fuse.³

Even if I am wrong about the timing, that does not make the model safe.

Gartner predicts agentic AI will autonomously resolve 80 percent of common customer-service issues by 2029.⁴ Cisco’s 2025 survey of 7,950 global business and technical decision-makers found respondents expected 68 percent of customer-service and support interactions with technology vendors to be handled by agentic AI by 2028.⁵ Forecasts like these will be wrong in detail. They usually are. But they are useful as market signals.

When buyers, vendors, and analysts begin treating routine service automation as a planning assumption rather than a demo curiosity, headcount stops being the natural unit of value.

Outcome pricing is not new. AI makes it harder to avoid.

Outcome-based BPO has been a recurring idea for more than a decade. What is new is that AI changes both sides of the equation simultaneously, and that is what makes the current shift different from previous cycles.

On the delivery side, software lets providers redesign work around the result rather than the staffing plan: automate the routine path, route exceptions to humans, have QA evaluate the system rather than check individual calls.

On the buying side, the same tooling makes effort harder to defend as the thing being purchased. Clients want resolved tickets, approved claims, reconciled invoices, qualified leads. They may still pay a base fee. They may still require service levels. But the commercial center of gravity moves toward the outcome.

That shift is already visible in the market.

Concentrix describes its Next Generation Customer Care offering as an outcome-driven managed service. The page explicitly states: “our outcome-based pricing model aligns our success with yours.”⁶ Everest Group’s 2025 CXM assessment cites that work as a strategic differentiator.⁷ TaskUs is using similar language — AI Services revenue grew more than 30 percent for the sixth consecutive quarter, and on the Q1 2026 earnings call CEO Bryce Maddock was direct: the goal is outcome-based pricing arrangements combining technology and talent “in a single price per solution” and “getting paid for resolving cases.”⁸⁹

That language matters because it describes the deliverable, not the staffing. The provider sells a result, and people and AI both have defined roles inside the system that produces it. Everest Group puts it plainly: outcome-based metrics are the new value currency in BPO, and clients do not just want a process completed; they want measurable business impact.¹⁰

But the outcome has to be real

Outcome pricing fails when the outcome is vague.

“Better customer experience” is a wish, not an outcome. “AI transformation” is a vendor slide. Even “Productivity improvement” means very little unless the baseline, the measurement window, and the attribution rules are spelled out for both parties.

A real outcome has to be observable, valuable, and contractible.

For customer support, that might mean first-contact resolution for a defined class of tickets, with exclusions for product defects, billing disputes, and policy exceptions the provider cannot control.

For claims processing, it might be complete intake and initial adjudication within 24 hours, with accuracy thresholds and a human-review lane for ambiguous cases.

For finance operations, invoice reconciliation to a defined confidence level, with exceptions queued for review and error rates measured against an agreed audit sample.

For lead operations, every inbound lead enriched, qualified, routed, and followed up within a fixed window, with unqualified leads marked with reason codes.

If the metric is loose, the contract becomes a fight. If the metric is sharp, the workflow can be designed around it. And that discipline is what elevates outcome pricing above a commercial preference. It becomes the operating model.

The practical model is hybrid first

The cleanest version of outcome pricing is simple: the client pays when the result is delivered.

Most real BPO transitions will not start there, because pure outcome alignment requires something most client relationships do not yet have — shared data, clear ownership, and enough mutual trust to apportion risk fairly.

They will start hybrid: a base fee to cover committed capacity, governance, technology, and fixed operating cost, plus an outcome-linked component tied to the result the client actually cares about. Everest makes this point bluntly: most current BPO deals remain hybrid models because the infrastructure for true risk-sharing takes time to build.¹⁰

For most providers, that bridge is more credible than a pure pay-per-result contract.

A Filipino operator trying to move from labor hours to outcomes should not begin by taking unlimited downside risk on a broken client process. Start narrower.

Pick one workflow with visible boundaries and one outcome that can be measured without a consulting archaeology project. Instrument the process, build the automation around the routine path, and define the human exception lane up front. Price a base fee with a performance component on top. Review the data every month, and tighten the contract as trust and measurement improve.

Less exciting than saying “AI will replace BPO,” but much closer to how this transition will actually work.

What changes inside the provider

An outcome-based firm is not an hourly BPO with a new price sheet.

The roles that matter are different: workflow designers who can map the process from intake to result; domain operators who can spot exceptions before they become failures; data teams who can build evaluation sets, QA loops, and escalation rules; commercial people who can price risk instead of only quoting seats; and managers who can read unit economics by outcome rather than staffing variance by account.

The frontline role changes too.

In the old model, the baseline worker handled volume. In the new model, software takes more of the routine path, and the human worker moves toward exception handling, evaluation, training, client-specific judgment, and process improvement. That does not mean fewer workers in every account immediately. It does mean the shape of the workforce changes.

Here is what the Philippines tends to underestimate about itself: the country has almost two million people who have lived inside customer support, claims, healthcare administration, finance operations, content moderation, workforce management, QA, training, and escalation. That knowledge is not generic. It is full of edge cases: the broken customer record, the ambiguous claim, the missing document, the angry caller who is technically wrong but operationally important, the compliance rule that looks simple until the real case hits it.

AI systems need that knowledge. They do not replace it cleanly. They need it to define the workflow, label the data, catch the failure, and improve the next version.

The work is turning that domain knowledge into owned capability instead of letting it remain rented labor.

The opportunity for founders is smaller — and more practical — than people think

People overestimate how large the first company has to be.

A Filipino founder does not need to build a full-service outsourcing company to participate in this shift. The first offer can be narrow:

claims intake for one kind of clinic or insurer;
lead qualification for one kind of real estate team;
invoice matching for one kind of supplier network;
customer onboarding for one kind of SaaS company;
content moderation QA for one kind of online marketplace;
renewal follow-up for one kind of subscription business.

The offer should not start with “we provide agents.”

It should start with the result.

“We process the intake queue within 24 hours and return only the exceptions your senior staff actually need to see.”

“We qualify every inbound lead, enrich the record, draft the first response, and route the opportunity before the day ends.”

“We reconcile the invoices, flag the mismatches, and give your finance team a clean exception list every morning.”

Those are not headcount offers. They are promises about a missing operational job — and that distinction is what changes the buyer’s posture. The client is not buying forty hours. The client is buying fewer dropped balls, and relief from having to think about that workflow at all.

Domain knowledge is the wedge. AI gives it reach. The accumulating asset is the process and data that improve each month.

At that point, a service starts becoming a productized capability.

The hard part is ownership

The risk is not that Philippine firms fail to adopt AI. The risk is that they adopt the language of outcomes while foreign platforms own the actual system.

If the local provider only resells someone else’s AI tool and supplies cheaper exception handlers, the structural problem has not changed; it has only been repackaged. The margin still migrates upward into the platform, the model, the customer relationship, or the workflow standard. The Philippines gets a different version of the same bargain: less labor, not enough control.

The better target is narrower and harder: use foreign tools where they help, but own the operating layer.

Use Salesforce, Zendesk, OpenAI, Anthropic, Decagon, Regal, or whatever else works. But do not confuse tool access with capability ownership. The owned layer is the workflow definition, the domain-specific data, the evaluation set, the operating playbook, the client relationship, and the ability to deliver the outcome repeatedly without becoming captive to one vendor’s product roadmap.

Bessemer Venture Partners makes the same point from the vendor side: AI companies are no longer just selling access, they are selling outcomes, pricing around completed workflows or human-equivalent output.¹¹ If that is where the software vendors are going, Filipino operators should not stay trapped selling only the human labor underneath them.

What the Philippines should build now

The next version of the Philippine services industry should not be framed as “BPO plus AI.” That frame is too small, and it keeps the Philippines in a reactive position relative to the technology.

A better frame is AI-native services built from Philippine operational knowledge — some inside existing BPO firms, others built by former team leads, QA managers, trainers, workforce planners, nurses, accountants, claims processors, and customer-support operators who understand a vertical well enough to package a better outcome.

That bridge is practical, not rhetorical: it turns rented labor into owned capability.

Outcome pricing will not absorb every workflow, every client, or every provider. In regulated and complex work, humans will remain central for a long time. But the value chain is moving away from headcount as the default unit, and the Philippines should not wait to learn that from clients who have already figured it out.

The old question was headcount: how many people can we deploy at what cost? The new question is harder and more interesting: what result can we guarantee, what system delivers it, what data improves it, and who owns the capability when the margin expands?

That question should organize Filipino operators, founders, investors, and policymakers.

I am not writing this as a neutral forecast. This is the bet behind HackManila: the data is clear enough to act on before the consensus catches up.

That means doing the work in public: a room where each operator picks one workflow, names the customer outcome, defines the human exception lane, prices the result, and walks out with a testable offer.

That is what HackManila is for. Community Night #1 — From Hours to Outcomes — is coming soon at Vantis in PBCom Tower, Makati. A short talk on outcome pricing, then a working session where each person sketches the first version of their own outcome-based offer. Bring a workflow you understand. Leave with something you can quote.

The hour-based model built the bridge. It should not be mistaken for the destination.

IBPAP, “Philippine IT-BPM Industry Closed with Growth, Moves Toward Higher-Value Work,” December 30, 2025; reports 2025 targets of 1.9 million jobs and USD 40 billion export revenues, with 2026 projections of USD 42 billion and nearly 1.97 million jobs. IBPAP Newsroom ↩
Bangko Sentral ng Pilipinas, Kristhel Anne M. Caynila and Genna Paola S. Centeno, “Adopting Generative Artificial Intelligence: Opportunities and Challenges in the Philippine IT-BPM Industry,” BSP Economic Newsletter No. 25-01, July 2025; reports 67% of IT-BPM member firms incorporating AI tools, 1.8 million FTEs in 2024, 1.6 million / 88.5% in contact center and business process services, and near-term augmentation rather than full-scale displacement. BSP EN25-01 (PDF) ↩ ↩²
This is a reading of capital-allocation and delivery-footprint signals, not a precise forecast. Concentrix’s 2024 Form 10-K describes the Webhelp acquisition as expanding its footprint in Europe, Latin America, and Africa, and names emerging-market growth opportunities including Egypt and South Africa. Teleperformance has reported large India expansion plans and Africa workforce scale above 50,000; TaskUs’s 2025 annual report shows AI Services as its fastest-growing service line and provides country-level performed-revenue and property-and-equipment disclosures. Concentrix Annual Reports; Business Standard — Teleperformance India hiring; Frost & Sullivan — TP Africa Report (PDF); TaskUs 2025 Annual Report ↩
Gartner, “Gartner Predicts Agentic AI Will Autonomously Resolve 80% of Common Customer Service Issues Without Human Intervention by 2029,” March 5, 2025. Gartner Newsroom — Agentic AI by 2029 ↩
Cisco, “Agentic AI Poised to Handle 68% of Customer Service and Support Interactions by 2028,” May 27, 2025; based on a survey of 7,950 global business and technical decision-makers across 30 countries. Cisco Newsroom — Agentic AI by 2028 ↩
Concentrix, “Next Generation Customer Care Technology,” describing NGCC as an outcome-driven managed service with generative AI frameworks, a value-based commercial model, and outcome-based pricing. Concentrix — Next Generation Customer Care ↩
Concentrix, “Concentrix Named a Leader in CXM Services PEAK Matrix Assessment 2025 — Global,” citing Everest Group recognition and strategic initiatives including outcome-based pricing models. Concentrix — Everest Group CXM PEAK Matrix 2025 ↩
TaskUs, “TaskUs Announces Fiscal First Quarter 2026 Results,” May 6, 2026; reports AI Services revenue growth above 30% for the sixth consecutive quarter and describes a combination of AI agents and human talent. TaskUs — Q1 2026 Results ↩
The Motley Fool transcript of TaskUs Q1 2026 earnings call, May 7, 2026; Bryce Maddock says the goal is to drive AI consulting engagements toward outcome-based pricing arrangements combining technology and talent in a single price per solution, “getting paid for resolving cases.” Motley Fool — TaskUs Q1 2026 Earnings Transcript ↩
Everest Group, “Outcome-based metrics: the new value currency in BPO.” Everest Group — Outcome-based metrics in BPO ↩ ↩²
Bessemer Venture Partners, “The AI pricing and monetization playbook,” discussing outcome-based charge metrics and AI-enabled services priced around completed workflows, outcomes, or human-equivalent output. Bessemer — AI Pricing and Monetization Playbook ↩