Samsung Just Committed $1 Billion to Helix. Here's Why the World's Biggest Conglomerate Is Betting Its AI Future on KKR and Nvidia's Infrastructure Venture.
Futurum Research's 1H 2026 survey found 43% of enterprise decision-makers prefer consumption-based AI pricing, with 27% favoring outcome-based structures. Gartner sees $234 billion of enterprise SaaS exposed to this shift by 2030. The vendors charging per resolution are already winning the procurement table.
In the first half of 2026, Futurum Research surveyed enterprise software decision-makers on their preferred AI pricing models. The result rewrote the assumptions underlying ten years of SaaS contract negotiation: 43% of enterprise decision-makers said they now prefer consumption-based pricing for AI software, and 27% said they prefer outcome-based structures — meaning vendors charge only when the AI delivers a verifiable business result. Combined, that is 70% of enterprise buyers who want AI pricing tied to usage or outcomes, and only 30% who are content paying for access regardless of what the AI actually accomplishes.
That preference shift has a number behind it. Gartner estimated in July 2026 that up to $234 billion of enterprise application spending — roughly 20% of the total enterprise SaaS market — could be exposed to outcome-based pricing models by 2030. That is not a forecast of revenue disappearing. It is a forecast of revenue migrating: from per-seat subscriptions to per-outcome billing, from predictable annual contracts to variable consumption invoices, from a pricing architecture built around access to one built around demonstrable value.
For SaaS vendors, this is the pricing disruption they watched coming for three years and are now running out of time to prepare for. For enterprise buyers, it is the negotiating leverage they have wanted since AI agents first started handling real work in production.
The Structural Reason Outcome-Based Pricing Is Winning
Outcome-based pricing is not a new concept. Legal firms have offered contingency arrangements for decades. Management consultants have piloted gain-sharing models. But it has always been a niche option rather than a default, because traditional software does not have a naturally measurable outcome — "was this spreadsheet worth $X per month?" is not a tractable question.
AI agents change the measurement problem fundamentally. A customer service AI agent either resolved the ticket or it did not. A lead qualification AI either produced a qualified meeting or it did not. A document processing AI either extracted the correct data fields or it did not. The binary measurability of agentic task completion creates natural outcome-based pricing anchors that were not available to traditional SaaS vendors.
The second structural driver is unit economics. When an AI agent resolves a customer support ticket for $2 and a human agent costs $15–40 to handle the same ticket, the ROI calculation is so obvious that traditional per-seat licensing fees become a credibility problem for the vendor. Charging $40 per seat per month for a customer service platform that includes an AI agent capable of handling 60–70% of tickets automatically is defensible only as long as customers do not do the math. In 2026, they are doing the math.
Signal's analysis of usage-based pricing's 10-point NRR advantage over seat-based models documented the macro version of this dynamic: SaaS companies using consumption-based pricing post 108% NRR versus 98% for seat-based models, a gap that is structural and growing. Outcome-based pricing is the next step in this evolution — from billing for consumption to billing for results.
The Zendesk Blueprint: $2 Per Resolved Ticket
Zendesk's outcome-based pricing model is the most clearly articulated enterprise implementation in production today. The model is straightforward: AI-automated resolutions bill at approximately $2 per ticket under pay-as-you-go terms, and approximately $1.50 per ticket under committed-volume contracts. Tickets that the AI attempts but cannot fully resolve — routing them to a human agent — are not billed.
That structure creates a direct alignment between Zendesk's revenue and the customer's value realization. Every automated resolution represents a ticket that would otherwise have cost $15–40 in human agent time. The $2 charge represents a 85–95% cost reduction for the customer on that ticket. The vendor captures a fraction of the savings delivered; the customer captures the rest. Both sides benefit from every successful AI resolution.
The operational complexity lies in the definition of resolution. Zendesk defines a successful resolution contractually — a customer replying with a satisfaction confirmation, a ticket closed without reopening within 48–72 hours — and the master service agreement specifies how edge cases are handled. This definitional work is the primary upfront investment in implementing outcome-based billing, and it is the reason most SaaS vendors have been slow to move: building the instrumentation and contractual framework to measure outcomes reliably takes 12–18 months and requires integrating billing systems with product analytics at a granularity that traditional SaaS companies have rarely needed.
Sierra AI and Intercom's Fin product have built similar resolution-based billing models, confirming that the Zendesk structure is becoming the architectural standard for AI customer-service agents. The Intercom Fin customer lifecycle data Signal analyzed earlier this month showed that outcome-aligned pricing structures correlate with significantly higher customer retention in the AI customer service category — buyers who understand they pay only for successful resolutions churn at lower rates than buyers on flat-fee plans.
The Five Pricing Models: Where Each One Fits
The migration from seat-based to outcome-based pricing is not a single step. Enterprise AI vendors are moving through a spectrum of models, and where they land depends on the measurability of their AI's outcomes and the maturity of their instrumentation.
| Model | How It Bills | Best For | Enterprise Adoption (1H 2026) |
|---|---|---|---|
| Per seat / user | Fixed monthly per user | General productivity tools with diffuse value | 30% of enterprise AI contracts |
| Credit / prepay | Purchased credits consumed per task | Developers, high-variance workloads | 12% |
| Usage-based | Per token, API call, or compute hour | Infrastructure, APIs, raw model access | 15% |
| Consumption-based | Per task completed (regardless of outcome) | Automation workflows, RPA replacement | 43% preferred |
| Outcome-based | Per verified business result (resolution, lead, document) | Customer service, sales qualification, document AI | 27% preferred |
| Hybrid | Base platform fee + outcome billing above a floor | Enterprise-wide AI deployments | Growing: ~20% of new contracts |
Source: Futurum Research 1H 2026 Enterprise Software Decision Makers Survey
The progression from usage-based to outcome-based is constrained by measurement infrastructure. A vendor can implement usage-based billing with standard API telemetry. Outcome-based billing requires the vendor to know what happened downstream of the API call — whether the support ticket was resolved, whether the lead was qualified, whether the document data was accurate. That requires deeper product instrumentation and often a bidirectional integration with the customer's CRM, ticketing system, or ERP.
The Six-Step Playbook for Building Outcome-Based Pricing
For SaaS vendors evaluating the transition to outcome-based models, the path from seat-based pricing to outcome-based billing requires investment at six distinct levels.
1. Define the outcome precisely. Before building any pricing infrastructure, the product team must answer: what event in the product logs constitutes a successful outcome for this customer? The definition must be binary, auditable, and defensible in a contract dispute. "Customer satisfaction" is not an outcome; "ticket closed without reopening within 72 hours after AI resolution" is an outcome.
2. Build the instrumentation stack. Outcome measurement requires event tracking at the task level — not just "user session started" but "task of type X was attempted, succeeded/failed, with these contributing factors." This typically requires a dedicated events infrastructure separate from standard product analytics, with higher data retention guarantees than typical session logging.
3. Pilot with high-measurability customers. The first outcome-based contracts should be with customers where the outcome is cleanest — customer service operations with high ticket volumes, document processing teams with structured data extraction requirements, sales teams with defined lead qualification criteria. High-measurability customers generate the historical data needed to price the model correctly and surface the edge cases that need contractual treatment.
4. Build the billing-product integration. Outcome-based billing cannot live in a spreadsheet. The billing system must ingest outcome events from the product in real time, apply the contracted per-outcome rate, apply any volume discounts or spend caps, and produce invoices that show the customer exactly what they paid for. This integration is the most common implementation failure point: vendors who underinvest in the billing-product bridge produce invoices customers cannot audit, which generates disputes and churn.
5. Address the accounting treatment upfront. Deloitte's June 2026 technology spotlight on outcome-based pricing for AI agents flagged a material challenge for vendors: revenue recognition under ASC 606 and IFRS 15 requires identifying the performance obligation and the transaction price at contract inception, which is straightforward for fixed subscriptions but complex when both the price per outcome and the number of outcomes are variable. SaaS vendors transitioning to outcome-based models need to engage their auditors early — before the first contract is signed — to establish their revenue recognition approach.
6. Launch with a hybrid option. Pure outcome-based pricing faces buyer resistance during the transition from seat-based procurement: finance teams need a minimum spend floor to forecast budgets. Launching the model with a hybrid option — a base platform fee plus outcome billing above a monthly floor — lowers the procurement barrier and allows buyers to convert to pure outcome-based billing once they have 6–12 months of outcome volume data to model against.
The $234 Billion Exposure: What It Means for Traditional SaaS
Gartner's $234 billion estimate is a calculation of exposure, not a prediction of loss. Traditional SaaS vendors do not lose that revenue when buyers shift to outcome-based pricing — they reprice it. The question is whether the reprice is margin-accretive or margin-compressive.
For vendors whose AI agents perform at high outcome rates — 65%+ automated resolution on customer service tickets, 80%+ accuracy on document extraction — the shift to outcome-based pricing can be margin-accretive. High performance means high outcome volumes, and high outcome volumes at $1.50–$2 per outcome can exceed the per-seat revenue the vendor was capturing on the same customer base.
For vendors whose AI agents perform at mediocre rates — 30–40% automated resolution, frequent escalations — outcome-based pricing is a direct margin compress. Low outcome rates mean low billing, and the vendor's cost base (compute, storage, engineering) does not scale down with billing volumes. The vendors with AI performance problems will face a brutal repricing when enterprise buyers shift contracts to outcome-based terms.
This is why the 70% buyer preference for consumption- and outcome-based models in Futurum's survey is not just a procurement preference — it is a performance accountability mechanism. Buyers who shift to outcome-based contracts are simultaneously negotiating for better pricing and imposing a performance test on the AI vendor. The vendors that survive the transition are the ones whose AI actually works at scale.
Signal's earlier analysis of OpenAI's three-tier model family documented the model provider side of this dynamic: the 50% price reductions on GPT-6 Sol and Luna are partly a response to the enterprise buyer pressure for cost structures that allow outcome-based business models to remain profitable at scale. If the underlying model costs $10 per million tokens, $2 per ticket is not a viable pricing floor. At $0.50 per million tokens, it is.
The Hybrid Structure That Most Vendors Will Land On
Neither pure outcome-based pricing nor pure seat-based pricing is likely to be the stable equilibrium for enterprise AI contracts. The destination is a hybrid model that most enterprise AI vendors are already converging on.
The hybrid structure combines a platform commitment floor — a minimum monthly or annual spend that guarantees the vendor minimum predictable revenue and gives the buyer budget predictability — with outcome-based billing above it. The floor is set at a level that covers the vendor's infrastructure and support costs for the customer's deployment. The outcome billing above the floor charges per verified result, with volume discounts kicking in above defined thresholds.
For enterprise buyers, this structure offers the budget predictability of a subscription with the value-alignment of outcome-based pricing. For vendors, it offers minimum revenue floor protection with the upside of high-performing AI delivering high outcome volumes.
The practical effect is to turn the AI vendor's per-outcome price into a performance guarantee. If the AI performs well, both sides win: the vendor captures high outcome volumes at the per-resolution rate; the customer captures the cost savings from high automation rates. If the AI performs poorly, the customer pays only the floor commitment while the vendor bears the reputational and churn cost of underperforming against the customer's automation rate expectations.
The vendors winning enterprise AI contracts in 2026 are the ones who have already built this hybrid model, because buyers who have seen the Gartner analysis and the Futurum survey data are now walking into procurement conversations asking for it.
Takeaway: The $234 billion outcome-based pricing disruption is not a prediction — it is already happening at the contract level. 43% of enterprise decision-makers now prefer consumption-based AI pricing; 27% prefer outcome-based structures; and the vendors who have built the instrumentation, the contractual framework, and the hybrid pricing architecture to support both are winning procurement conversations that seat-based competitors are losing. The vendors who are not yet measuring outcomes in their product logs need to start today, because the enterprise buyer who walks in 12 months from now asking for per-resolution billing is not going to wait while you build the instrumentation stack. The performance accountability test that outcome-based pricing imposes is the best thing that has happened to enterprise software quality in a decade — and the worst thing that will happen to mediocre AI vendors who have been hiding behind flat subscription fees.
Frequently Asked Questions
What is outcome-based pricing for AI agents?
Outcome-based pricing for AI agents charges customers only when the AI delivers a specific, measurable result — a successful support ticket resolution, a qualified lead generated, an invoice processed, a workflow completed — rather than charging for access, seats, or raw token consumption. Unlike per-seat SaaS pricing (which bills regardless of usage) or usage-based pricing (which bills per API call or token consumed), outcome-based pricing aligns the vendor's revenue directly with the customer's value realization. The customer pays nothing when the AI fails or escalates to a human; they pay a defined unit price when the AI succeeds. Zendesk is the clearest enterprise example: their AI agents bill approximately $2 per automated resolution under pay-as-you-go terms and approximately $1.50 per resolution under committed volume arrangements, with no charge for tickets that require human intervention. Outcome-based pricing is more complex to implement than seat-based or usage-based models because it requires clear definition of what constitutes a successful outcome, robust instrumentation to measure it, and contractual agreement on how edge cases and partial outcomes are handled.
Why is Gartner estimating $234 billion of SaaS spending is at risk from outcome-based AI pricing?
Gartner's July 2026 estimate that up to $234 billion of enterprise application spending could be exposed to outcome-based AI pricing by 2030 reflects a structural threat to traditional SaaS licensing economics. Traditional enterprise SaaS is priced on access (per seat) or capacity (storage, API calls), which means customers pay whether or not the software delivers value. AI agents that perform the same tasks as human SaaS users — processing tickets, generating reports, updating CRM records — create a natural comparison: if an AI agent resolves a support ticket for $2 and a human agent costs $15–40 for the same resolution, the economic case for per-seat licensing fees for the human workflow erodes rapidly. Gartner's $234 billion figure represents roughly 20% of total enterprise application SaaS spending, scoped to the categories most exposed to agentic task completion: customer service, document processing, sales automation, HR workflows, and IT service management. The risk to traditional SaaS vendors is not that customers stop paying — it is that the pricing model shifts from predictable subscription revenue to variable outcome-based revenue, compressing margins and increasing revenue volatility for vendors who have not built the infrastructure to measure and bill per outcome.
How does Zendesk's outcome-based pricing model work in practice?
Zendesk's AI agent billing model is the most visible enterprise implementation of outcome-based pricing in production today. Under Zendesk's model, AI-automated resolutions — support tickets fully resolved by the AI without human intervention — are billed at approximately $2 per resolution under pay-as-you-go pricing and approximately $1.50 per resolution under committed-volume contracts. Tickets that the AI attempts but cannot fully resolve, routing them to a human agent instead, are not billed. This creates a direct alignment between Zendesk's revenue and the customer's cost savings: every automated resolution represents a ticket that would otherwise have cost $15–40 in human agent time, so the $2 charge is compelling on unit economics alone. The model requires Zendesk to define 'resolution' precisely and contractually — does a customer replying 'thank you' count? Does a ticket reopened 48 hours later reclaim the resolution credit? These edge cases are resolved in Zendesk's master service agreements and are the primary operational complexity of implementing outcome-based billing. The Sierra AI platform and Intercom's Fin product use similar resolution-based billing models, indicating that outcome-based pricing is becoming the standard architecture for customer-service AI agents.
What is the difference between outcome-based pricing and usage-based pricing for AI?
Usage-based pricing bills for inputs — tokens consumed, API calls made, compute hours used — regardless of whether those inputs produced a valuable output. Outcome-based pricing bills for verified outputs — a resolved ticket, a closed lead, a processed document — regardless of how many tokens or compute resources were required to produce it. The distinction matters because AI workloads are highly variable in their input consumption. A simple support ticket may require 500 tokens to resolve; a complex one requiring multi-step reasoning and document retrieval may require 50,000 tokens. Under usage-based pricing, the complex ticket costs 100 times as much as the simple one. Under outcome-based pricing, both cost the same — the defined per-resolution price — because the customer is paying for the result, not the work required to produce it. This makes outcome-based pricing more predictable for buyers and more margin-compressing for vendors in cases where complex tasks require disproportionate compute. The Futurum Research 1H 2026 survey found that 43% of enterprise decision-makers prefer consumption-based pricing and 27% prefer outcome-based structures, with the remaining 30% split across seat-based, credit-based, and hybrid approaches.
What is the hybrid pricing structure most enterprise AI vendors will adopt?
Most enterprise AI vendors are converging on a hybrid structure that combines a minimum commitment floor with outcome-based billing above it. The floor — typically a base platform fee or minimum monthly spend — gives the vendor predictable minimum revenue and gives the buyer a cost cap on periods of low activity. The outcome-based component above the floor charges per verified result: per resolution, per lead, per processed document. This structure addresses the two primary objections to pure outcome-based pricing: buyers dislike the open-ended spend exposure of pure outcome-based models during high-activity periods, and vendors dislike the zero-revenue outcome of pure outcome-based models during system onboarding and low-volume periods. The hybrid approach, described in Gartner's July 2026 analysis and Deloitte's technology spotlight on outcome-based accounting, is expected to become the dominant enterprise AI pricing architecture by 2027. It allows vendors to retain the revenue predictability of subscription models while offering the value alignment that enterprise procurement teams increasingly demand. Zendesk, Sierra, and Intercom Fin all offer hybrid variants alongside their pure outcome-based options.
How should enterprise buyers negotiate outcome-based AI contracts?
Enterprise buyers negotiating outcome-based AI contracts should focus on four contractual elements that determine whether the model actually delivers the cost savings promised. First, negotiate the definition of 'outcome' with precision: what constitutes a successful resolution, lead qualification, or document processing event must be defined in writing before the contract is signed, not discovered after invoices arrive. Second, insist on measurement transparency: the vendor must provide real-time or near-real-time reporting on outcome volumes, outcome rates, and the cases where the AI attempted but failed to achieve an outcome — buyers need this data to audit their bills and identify performance issues early. Third, negotiate caps: outcome-based pricing can produce unexpected spend spikes during high-volume periods (seasonal customer service demand, campaign launches); monthly or quarterly spend caps protect against budget overruns while leaving the outcome-based incentive intact. Fourth, build in performance floors: if the AI's outcome rate falls below a defined threshold (for example, below 60% automated resolution rate for a customer service deployment), the per-outcome price should be renegotiated or the contract should include a remediation clause. These four elements are the difference between an outcome-based AI contract that delivers budget predictability and one that creates procurement headaches.