Tired of reading? You can now listen to this article in a PodCast format. Video and image formats are at the bottom of this article.
Twelve months ago, I argued that Results‑as‑a‑Service (RaaS) would be the next evolution of software pricing: pay for outcomes, not licenses (read my old post). Enterprises wouldn’t just rent software; they’d rent results. That prediction was directionally right—but the way RaaS is actually showing up in contracts looks very different from the 2025 hype cycle.
Over the past year, AI agents moved from “cool demo” to “embedded in production workflows.” Agentic AI is now a multi‑billion‑dollar market, with forecasts pushing it into the tens of billions by the early 2030s. At the same time, some of the loudest voices in enterprise AI are now calling outcome‑based pricing “the most expensive myth” in the space.
So which is it?
The reality is simple: pure outcome‑based RaaS is rare and hard; hybrid RaaS is what’s actually shipping.
This Part 2 is about the implementation reality—what worked, what broke, and how the smartest teams are evolving their pricing and contracts to make RaaS sustainable in 2026.
Too long of a read? Just watch our 6 min video below.
1. From Theory to Reality
The Three Models That Actually Ship
In the 2025 article, I framed RaaS as the next step beyond SaaS: instead of paying for seats or usage, you pay for concrete business outcomes (qualified meetings, resolved tickets, closed deals). Vendors like Sierra were already leaning into this, charging per resolved support interaction instead of per agent or per seat.
Over the last year, what’s emerged is a three‑model spectrum:
| Model | How It’s Priced | Vendor Risk | Buyer Risk | Where It Shows Up |
| Traditional SaaS | Fixed subscription (per seat, per account, per feature) | Low | High (pay regardless of value) | Most legacy tools |
| Pure RaaS (Theory) | 100% outcome‑based (only pay on results) | Very high (front‑loaded work, back‑loaded revenue) | Very low | Rare pilots, niche deals |
| Hybrid RaaS (Reality) | Base fee (30–50%) + outcome component (50–70%) | Managed | Managed | Most “outcome‑based” deals in practice |
A few concrete examples:
Sierra AI publicly champions “pay for outcomes, not for software,” getting paid when an AI agent resolves a customer issue instead of charging for seats or generic usage. Under the hood, that looks more like a structured per‑resolution fee pegged to the cost of a human‑handled call (typically $10–$20 per interaction), often layered into broader commercial relationships.
Enterprise pricing for AI agents is increasingly a layer cake: a foundational platform fee, a usage layer (minutes/tokens/interactions), and an outcome‑based layer for high‑value events.
In other words, the headline is often “outcome‑based,” but the contract is almost always hybrid.
2. The Attribution Problem: RaaS’s Biggest Operational Hurdle
On paper, outcome‑based pricing sounds clean: “We only get paid when we deliver X.” In practice, step into an enterprise environment and everything gets messy.
When a support ticket is resolved or a deal closes, who gets credit?
- The AI agent that handled most of the interaction?
- The human who stepped in at the end?
- The marketing campaign that sourced the lead?
- The seasonal spike in demand that lifted everyone’s numbers?
Analysts working on outcome‑guarantee frameworks for AI agents have highlighted the same thing: multi‑variable attribution is the number one implementation barrier. Everything from market conditions to concurrent internal initiatives can inflate or suppress the metrics you attach to payment.
This creates three predictable failure modes:
Over‑paying for ambient lift
The AI vendor gets credit for results driven by broader market tailwinds or internal changes (pricing changes, new hires, etc.). Finance teams quickly notice they’re paying outsized fees for outcomes their existing stack or team could have achieved anyway.
Under‑crediting the agent
The inverse: internal stakeholders argue the AI didn’t “really” create the outcome; it merely assisted. Legal, procurement, RevOps, and finance end up spending cycles arguing over dashboards instead of improving performance.
Perpetual reconciliation meetings
Because outcomes cannot be perfectly defined or attributed, enterprises get stuck in recurring cycles of measurement debates and invoice verification sessions.
That’s exactly the critique emerging from operators and from Forbes’ January 2026 piece: outcome‑based pricing embeds ongoing governance friction and turns ROI conversations into billing arguments.
What’s Working Instead
The teams that are making RaaS work do a few things differently:
Define “outcome” with edge‑case examples
Outcomes are written like product specs. A “resolved ticket,” for example, must meet clear criteria: no human escalation, specific CSAT threshold, no re‑open within X days.
Bake in attribution rules up front
Contracts explicitly define:
- When the AI gets full credit
- When credit is split between human and AI
- When no outcome fee applies (e.g., escalations)
Instrument “outcome audit trails”
Vendors and customers run shared dashboards with real‑time metering across interactions, outcomes, and escalations, often leveraging agent‑driven SLA monitoring.
Use outcome pricing only where attribution is tight
Discrete, well‑bounded workflows (like support resolution, lead qualification, document drafting) are far better candidates than multi‑touch domains like “brand awareness” or “overall revenue growth.”
The punchline: RaaS fails where attribution is fuzzy, and succeeds where the workflow is discrete, instrumented, and heavily automated.
3. The Legal & Compliance Minefield You Can’t Ignore
In 2025, most RaaS discussions treated legal and compliance as a footnote. In 2026, that’s no longer possible.
Legal practitioners evaluating AI performance guarantees have pointed out that certain outcome‑based arrangements can start to look a lot like insurance products. That comes with its own regulatory regime.
Three zones are especially risky:
3.1 Insurance‑like Guarantees
When you guarantee outcomes that depend on factors outside your direct control, regulators may see that as underwriting risk.
- Some legal analyses argue that if performance guarantees cover non‑incidental risks—things not purely tied to the vendor’s operations—the structure can trigger insurance licensing requirements.
- This risk doesn’t hit every RaaS model equally, but broad, open‑ended guarantees (“we’ll hit X business outcome regardless of context”) are far more exposed than narrow, operational promises tied to clearly defined services.
What smart teams are doing:
- Keeping guarantees narrow and operational (e.g., “resolution time for tickets handled by the agent” vs. “overall NPS or revenue lift”).
- Aligning coverage limits with contract economics rather than offering “blanket” guarantees.
- Involving legal early, especially across multi‑state or multi‑country deployments where the “incidental” test for insurance varies.
3.2 Healthcare and Finance Revenue‑Sharing Rules
In verticals like healthcare, outcome‑based fees can collide with value‑based reimbursement rules, anti‑kickback statutes, and revenue‑sharing prohibitions.
- Healthcare attorneys note that outcome‑or revenue‑linked fees can look uncomfortably close to disfavored revenue‑sharing models if they touch clinical outcomes or regulated reimbursements.
- Finance has similar issues when fees start to look like a share of financial performance rather than a clear service fee.
Practical pattern: RaaS in these verticals tends to favor:
- Operational metrics (e.g., speed of claims processing, administrative throughput) rather than direct clinical or financial outcomes.
- Hybrid structures where the outcome component is capped and clearly framed as a performance bonus within a broader service relationship.
3.3 Warranty & Representation Law
Any time you say “we guarantee X result,” you’re now in warranty territory. Under U.S. federal warranty law (Magnuson‑Moss), businesses must be explicit and transparent about the terms of any warranty they offer.
Combined with the probabilistic nature of AI systems, this means:
- Guarantees must account for the fact that AI performance is distributional, not binary.
- Clauses need to spell out exclusions and limitations (data quality, third‑party system failures, customer process changes, etc.).
- Allies in legal are pushing enterprises to use structured performance guarantees instead of vague promises.
The high‑level lesson: RaaS isn’t just a pricing innovation; it’s a legal design problem. If you ignore that, contracts will die in procurement and legal review long before they reach production.
4. Cash Flow & Revenue Recognition: The Vendor Side of the Story
In 2025, most RaaS narratives focused on the buyer’s upside: reduced risk, aligned incentives, “pay only when it works.” The vendor’s financial reality got less airtime.
Two forces are colliding here:
- Front‑loaded costs
RaaS vendors still incur significant upfront investment: solution design, integrations, implementation, data cleanup, training, support. These costs hit before any outcome revenue shows up. - Back‑loaded, variable revenue
Under pure outcome pricing, vendors only recognize revenue when outcomes are achieved and verified. That can be monthly, quarterly, or even slower depending on the contract and verification process.
From a revenue recognition standpoint, outcome‑based contracts stack complexity on top of the existing ASC 606 model most SaaS companies already struggle with. Finance teams have to:
- Identify performance obligations
- Set transaction prices that include uncertainty and risk premia
- Allocate prices across base and outcome components
- Recognize revenue only when specific obligations (outcomes) are satisfied
Several accounting and finance guides now flag outcome‑linked fees as a special case that requires extra care in modeling and documentation.
Why Pure RaaS Breaks Most P&Ls
When you combine front‑loaded costs with back‑loaded, uncertain revenue, you get:
- Long payback periods
- Working capital strain
- Difficulty forecasting cash flows
- Increased risk in downturns or performance dips
This is one reason why many outcome‑based deals either:
- Fail to scale beyond pilot phases, or
- Quietly revert to hybrid structures where a base fee covers a meaningful portion of fixed costs.
Hybrid RaaS as Financial Engineering
Hybrid RaaS isn’t just a compromise; it’s financial engineering that makes these deals viable:
- Base Fee (30–50%)
Covers platform, implementation, and ongoing support. Revenue is more predictable; costs are covered earlier in the relationship. - Outcome Component (50–70%)
Tied to clearly defined, auditable metrics. Keeps incentives aligned and gives buyers the “skin in the game” story they want for internal stakeholders. - Caps and Floors
Outcome components are often capped so finance teams can model worst‑case budgets, while vendors secure minimum commitments to avoid building a bespoke solution that never breaks even.
Newer rev‑rec and billing platforms are already publishing playbooks for hybrid and evolving AI pricing models, explicitly calling out the need to balance risk while retaining outcome‑based upside.
The net effect: Hybrid RaaS is what lets vendors stay solvent while still selling the “pay for results” story.
5. Vertical Reality: Where RaaS Actually Works
Tier 1: High‑Fit Verticals (Discrete, Measurable, Automated)
These are the categories where RaaS is starting to look like a new default.
- Customer Support
- Clear unit of value: resolved conversation or deflected call.
- Measurable cost baseline: a live agent call typically costs $10–$20.
- RaaS implementation: vendors like Sierra charge per successful resolution, pegged to a fraction of the avoided cost.
- Outcome pricing works because the outcome is discrete, auditable, and directly tied to cost savings.
- Sales/SDR
- Unit of value: qualified meeting, accepted opportunity.
- Common structure: base fee + per‑opportunity or per‑meeting bonus tied to agreed qualification criteria.
- Still messy in attribution, but far more tractable than revenue‑share on closed/won.
- Legal Operations
- Unit of value: drafted document, redlined contract, completed review.
- Firms like Harvey have shown that AI can materially compress drafting and review time for large law firms.
- RaaS here often shows up as a blend of per‑matter fees plus usage, framed around throughput and speed to completion.
Tier 2: Emerging Verticals (Measurable but Complex)
These categories see a lot of experimentation, but pure RaaS is rare.
- Healthcare Operations
- High agentic AI growth; healthcare is projected as one of the fastest‑growing segments for AI agents.
- But outcome fees tied to clinical results or reimbursements land in regulatory hot water.
- Successful models instead focus on operational metrics: time to process claims, administrative throughput, scheduling efficiency.
- Financial Services Ops
- Similar story: operational outcomes (e.g., KYC speed, fraud‑check throughput) are viable; anything that looks like shared financial upside is fraught.
- Hybrid pricing with clear fee‑for‑service and capped performance bonuses is more common.
- Supply Chain & Logistics
- Outcomes like on‑time delivery or inventory optimization are attractive but heavily influenced by external factors (supplier reliability, macro shocks).
- Many deals end up as SLA‑linked service credits rather than pure outcome‑based fees.
Tier 3: Poor Fit (Fuzzy Outcomes, Multi‑Touch)
These categories are still mostly theoretical for RaaS.
- Brand marketing
- Broad “revenue growth” programs
- R&D / innovation projects
The pattern: RaaS works best where you can point at a discrete event, instrument it end‑to‑end, and minimize the number of human and external dependencies. Everywhere else, it either devolves into arguments or quietly reverts to traditional pricing.
6. Implementation Playbook: A 90‑Day RaaS Pilot That Doesn’t Blow Up
If you’re a founder, operator, or buyer trying to test RaaS in 2026, you don’t need a grand theory. You need a contained experiment.
Here’s a 90‑day pilot framework I recommend, informed by both vendor best practices and the emerging research around outcome guarantees.
Phase 1 (Days 1–30): Design for Measurability
- Pick the right use case
- Discrete workflow
- Clear unit of value
- High volume
- Low to moderate regulatory complexity
Customer support deflections, SDR qualification, and document drafting remain the best starting points.
- Define outcomes and non‑outcomes
- Write a one‑page spec that defines what counts as a success, what counts as a partial success, and what doesn’t count at all.
- Include edge cases: escalations, partial AI assistance, human overrides.
- Agree on attribution and audit
- Decide how you’ll split credit when humans and AI collaborate.
- Define the data sources you’ll trust (CRMs, ticketing systems, logs).
- Set up shared dashboards for real‑time monitoring.
- Design hybrid pricing, not pure RaaS
- Base fee covers platform and support.
- Outcome fee kicks in only on clearly defined successes.
- Add caps so Finance can model worst‑case spend.
Phase 2 (Days 31–60): Launch and Over‑Instrument
- Start with a narrow slice of traffic
- Route a subset of tickets, leads, or matters to the AI‑driven workflow.
- Keep the control group (legacy process) for comparison.
- Instrument everything
- Track volume, success rate, escalations, CSAT/NPS, time‑to‑resolution, and impact on adjacent metrics.
- Log and categorize disputes over what counts as an outcome.
- Weekly reviews, not quarterly surprises
- Vendor + customer meet weekly to review metrics, disputes, and edge cases.
- Adjust definitions and routing rules as needed (with change tracking).
Phase 3 (Days 61–90): Refine and Decide
- Analyze outcomes vs. counterfactual
- Compare the RaaS workflow to the control group on cost per outcome, speed, and quality.
- Identify where attribution friction was highest.
- Tune the pricing mix
- If the vendor is upside‑down on cash flow, increase base fee or lower outcome amounts.
- If the buyer sees limited upside, rebalance so they capture more of the efficiency gains.
- Codify the learnings into contract v2
- Roll successful pilots into a one‑year contract with refined specs, dispute resolution mechanisms, and clear remedies for SLA breaches.
The goal of this 90‑day process isn’t to prove RaaS universally works. It’s to answer a simpler question: “Does RaaS work for this use case, in this environment, with these metrics?”
If the answer is “yes,” you scale. If not, you either revert to consumption‑based pricing (minutes, calls, tokens) or a more traditional subscription model.
7. The Skeptics’ Case—and the Hybrid Rebuttal
In early 2026, a widely shared Forbes piece labeled outcome‑based pricing “the most expensive myth in enterprise AI.” The core critique:
- Efficiency gains often flow to the vendor, not the customer, because the vendor’s fee structure captures a share of the improvement.
- Outcome models embed a permanent risk premium based on uncertainty, even after systems stabilize.
- They generate governance overhead as enterprises pour time into validating invoices instead of improving performance.
That critique is valid—for badly designed outcome models.
The counterpoint emerging from vendors and operators who are making this work is more nuanced:
- When outcomes are narrow, discrete, and tightly instrumented, outcome fees can create real alignment—especially in high‑volume, cost‑center functions like support.
- Hybrid models can be structured to keep most efficiency upside with the customer (e.g., vendor’s fee is a fraction of avoided cost, with caps and stair‑step discounts over time).
- For CFOs, outcome‑linked components can make it easier to tie AI spend directly to KPIs they already track, if variability is bounded.
In practice, the market seems to be converging on a middle ground:
- Pure outcome pricing will remain rare and experimental.
- Hybrid RaaS—base + bounded outcome component—will spread across high‑fit workflows.
- Consumption‑based pricing (minutes, calls, tokens) will continue to dominate for general‑purpose AI assistants and broad, multi‑use agents.
8. Where RaaS Goes Next
Looking ahead to 2027, the directional trends are clear:
- Agentic AI continues to grow at strong double‑digit CAGR, with forecasts into the tens or even hundreds of billions.
- More enterprise apps ship with embedded agents, not just APIs, which naturally pushes conversation toward value and outcome, not just features.
- At the same time, procurement and finance are becoming more vocal about predictability, transparency, and governance, pushing back on any model that feels like a black box.
Put all of that together, and the most likely future looks like this:
- RaaS doesn’t replace SaaS; it layers on top of it.
- Outcome components become one of several levers in a more sophisticated pricing stack: subscription, consumption, and outcomes coexisting.
- The vendors who win aren’t the ones shouting “pay only for results” the loudest; they’re the ones who quietly design contracts that finance, legal, and operations can live with—and then actually hit the numbers.
If Part 1 of this series was about the promise of RaaS, Part 2 is about its constraints—and how the most pragmatic teams are working inside those constraints to build something sustainable.
The takeaway for 2026 is simple:
- If you’re a buyer, don’t chase pure RaaS for the story. Look for hybrid models with clear definitions, caps, and dispute processes.
- If you’re a vendor, stop trying to be an insurance company. Charge enough base to survive, price outcomes where attribution is tight, and use transparency as a competitive edge.
- For everyone else watching from the sidelines: RaaS isn’t dead. It’s just growing up.
Related Assets
Presentation style
Infographic 01
Infographic 02
Discover more from Fornelli by Design | FxD
Subscribe to get the latest posts sent to your email.
