Back in Week 2, we flagged an AWS funding motion called Business Value Realization and promised to unpack it fully once the ground was laid. That ground is laid. Business Value Realization, announced June 16, 2026, ties partner funding to customer outcomes, tracked through defined adoption stages, with customers attesting to the business outcomes achieved and AI agents in AWS Partner Central generating weekly adoption reports on each engagement. Funding is disbursed automatically through the AWS Partner Funding Portal as each stage is completed. Partners that also earn the companion BVR Competency unlock $50,000 in marketing development funds in each of 2026 and 2027. AWS built the plumbing to prove an outcome occurred before money moves.
What that plumbing does not do, and was never built to do, is say whose outcome it was when more than one partner touched it. A deployment a systems integrator delivered, on a product a technology partner built, that a marketplace listing surfaced, that a co-sell agent qualified first, clears exactly one milestone in Business Value Realization. The verification is solved. The credit is not even asked.
Deal registration answered this question once, for one shape of deal: a single partner sources a single opportunity at a single identifiable moment. Episode 8 found marketplace transactions already making that shape obsolete for a growing share of deals. This episode is about the deals that were never that shape to begin with, the ones where an agent surfaces the lead, a second partner's product gets embedded in the delivery, and a third orchestrates the workflow that makes the outcome land at all. The plumbing is solvable, as we said closing out Act II. The money is not, and this episode is where Act III starts building the answer.
Four ways a deal gets made now
Every attribution system in production today, including the ones built by BlueThread's own approved vendors, recognizes exactly two categories: sourced, meaning a partner originated the opportunity, and influenced, meaning a partner helped move it without originating it. Crossbeam's own product description is explicit about this. Its Attribution Activity Timeline tags each qualifying activity as sourced or influenced and sends that single tag into Salesforce, drawing in part on Gong call-recording data to detect when a partner gets mentioned on a sales call. It is a well-built two-category system, and two categories are what it recognizes.
Two categories described most partner-touched revenue reasonably well when a partner was a person doing one identifiable thing. They stop describing it once a deal can involve a product embedded in the delivery itself, or a workflow an orchestration layer ran end to end without a person originating or influencing any single step of it in the traditional sense. This series needs names for those two additional shapes, so here they are: embedded revenue, where a partner's product or capability is built into what the customer actually receives, and orchestrated revenue, where a partner's system runs the workflow that produces the outcome, coordinating other parties' contributions rather than contributing one identifiable touch of its own. Sourced, influenced, embedded, orchestrated. Four types in total, and no vendor's data model currently has fields for the last two.
What the plumbing solves, and what it still can't answer
AWS extended its own automation on July 20, 2026, giving Partner Central's agents eligibility validation, documentation review, and auto-population across every funding program, including Strategic Collaboration Agreements and the AWS Growth Initiative. Read the announcement closely and the omission is the finding. Nowhere does it describe how credit gets split when a funded outcome involved more than one partner. AWS automated the administrative half of the funding problem completely and left the attribution half exactly where deal registration left it a decade ago.
This is the honest state of the entire market. TSIA's State of Channel Partnerships 2026 research, last updated January 29, 2026, found only 41 percent of vendors have defined what partner success looks like in their own business and which partner actions drive it, before any agent enters the picture. Attribution disputes were already a definition problem, the kind dashboards can't fix, because nobody had agreed on the definition in the first place.
What today's tools actually recognize
The pattern in Episode 7 was multi-seller credit among humans, and it was already unsettled. PartnerStack and Wynter's State of Partnerships in GTM 2026 survey of 100 revenue, marketing, and partnerships leaders at B2B SaaS companies found 42 percent attribute partner revenue with multi-touch models, 31 percent with first-touch, and 19 percent with last-touch. Three different answers to one question, at companies that all have a CRM. Adding embedded and orchestrated revenue on top of that does not create the disagreement. It exposes how much of it was already there.
Forecastable's own guidance to partnerships teams on this exact fight names the actual mechanism plainly: a CFO who asks how a number is defined, gets hesitation instead of an answer, and mentally writes the number off from that point forward. Its own prescribed workaround for a deal touched by more than one partner is to document the secondary partners in a notes field and resolve credit allocation manually at compensation time. That is the honest state of practice in 2026: the fix for a four-party deal is a text box a human reads later. No approved BlueThread vendor, and no vendor we found in this research, has replaced that text box with a structured field.
Our own State of Partnerships survey adds a piece of texture worth sitting with. Across four separate questions asking partnership professionals how comfortable they are letting AI make a consequential call on its own, without a human checking it first, "not without human review" was the most common answer on three of the four, and on the hardest of them it was chosen by exactly half of the 78 respondents. Even on the question where people were most comfortable delegating, nearly one in four still said the same thing. Handing a structured taxonomy to an automated system to referee is a much smaller ask than handing it a live credit decision, and even the smaller ask does not yet have the field's trust secured.
The case against it, and where it holds
Three objections come up whenever a four-type taxonomy meets an actual comp conversation, and each deserves a real answer.
Gaming is the first. A partner who knows embedded revenue pays out will describe every integration as embedded, whether or not it was load-bearing to the deal. The fix is the same qualifying-conditions discipline this series has recommended before: embedded status requires a documented technical dependency, verified the same way a support ticket or an integration log already gets verified today.
Complexity is the second, and it is the one smaller teams raise fastest. A five-person partnerships function does not have the bandwidth to run four categories with the same rigor a five-hundred-person RevOps org can. The fix is scope: apply the four-type split only above a deal-size or partner-count threshold where the disagreement actually has stakes, and default everything below it to the simpler sourced-or-influenced call teams already make.
The third is the one worth taking most seriously, because published guidance on attribution already flags it independently of this series. Journeybee's own 2026 guidance on partner attribution states the risk directly: AI's role is to prompt a human review and improve data quality. A four-type taxonomy is a classification scheme. The moment it starts assigning dollar amounts without a person confirming the classification, the same trust gap our own survey found becomes the whole system's single point of failure. The taxonomy earns its keep by giving humans a shared vocabulary for a disagreement they were already having, and it settles that disagreement only with a person still signing off on the answer.
The Operator Move: run one deal through the four types
Twenty minutes. Pick one deal that closed last quarter and involved more than one partner or tool.
- Which partner sourced it, meaning which one originated the opportunity before anyone else touched it?
- Which partner or product is embedded in what the customer actually received, separate from whoever sourced the deal?
- Did any workflow, human or agent-run, coordinate the other pieces into a single result?
- If you had to defend this deal's credit split to your CFO in one sentence per party, could you do it today?
If question four stalls you, you have found the exact gap this episode is about.
By the end of 2027, at least one major CRM or PRM platform will ship a native data field for embedded or orchestrated revenue, beyond the sourced-and-influenced categories every major platform, including Crossbeam, currently offers. Logged to the public scorecard.
- AWS, Business Value Realization. A funding motion that ties partner funding to customer-attested outcomes, announced June 16, 2026, with funding disbursed automatically once a stage clears. This episode's hook and the payoff of a breadcrumb planted in Week 2, it proves AWS can verify an outcome happened. It says nothing about whose outcome it was when more than one partner was involved.
- AWS Partner Central, agents expanded to all funding programs. Automated eligibility validation, documentation review, and fund request population extended across every AWS partner funding program, including Strategic Collaboration Agreements, announced July 20, 2026. The administrative half of funding is now fully automated. The attribution half, splitting credit among the partners who produced a funded outcome, is not mentioned once.
- TSIA, State of Channel Partnerships 2026. Research finding that only 41 percent of vendors have defined what partner success looks like and which partner actions drive it, last updated January 29, 2026. The definition problem this episode is about predates agentic AI by years. Agents are simply the ones making it urgent.
- Level 1. We still only track sourced and influenced revenue, and multi-partner deals get resolved informally, if at all.
- Level 2. We know embedded and orchestrated contributions exist on some deals, but we track them in notes fields and resolve credit by memory at compensation time.
- Level 3. We apply a documented four-type framework to our highest-stakes deals, with clear qualifying conditions, but smaller deals still default to the old two-category call.
- Level 4. All four types are documented with verification standards a CFO would accept without asking who defined them, and any AI involved in the process improves the documentation rather than assigning the credit itself.
Business Value Realization proves AWS can verify an outcome happened. It has nothing to say about whose outcome it was. Next week we turn from who gets paid to who's on the hook when the outcome doesn't land at all. Episode 11: Who's Accountable When the Outcome Doesn't Land?
One ask. We are building an operator-led benchmark on how partnership teams actually use AI. Seven minutes, anonymous, results published publicly. Add your voice.