The most likely next milestone in agentic commerce is not a feature release. It is a governance announcement. Signals point to at least one of the two dominant agent-checkout standards, OpenAI and Stripe’s Agentic Commerce Protocol (ACP) or Google and Shopify’s Universal Commerce Protocol (UCP), moving to neutral foundation stewardship or a formal multi-vendor governance body by NRF 2027 in mid-January.
The tell is not what either camp has shipped since April 2026. It is what neither has shipped.
In short
- The prediction: one of ACP or UCP likely announces neutral stewardship or a formal multi-vendor governance council by NRF 2027 (mid-January 2027), roughly five months out.
- Signal 1: both specifications have been frozen since April 2026. ACP’s latest is 2026-04-17; UCP’s latest tagged release is v2026-04-08. Neither has cut a spec release in four months, after cadences of four to six weeks.
- Signal 2: the adjacent agent protocol that already sits under neutral governance, MCP, shipped its largest revision since launch on July 28, 2026, under Linux Foundation stewardship. The neutrally governed spec moved; the corporately stewarded ones did not.
- Signal 3: Salesforce’s second Agentic Enterprise Index, published in August 2026, reports that businesses deploying agents on digital commerce channels saw 4x higher online sales growth. Commercial stakes are now high enough that fragmentation is expensive.
- The counter-case: the freeze may simply be pre-holiday implementation lockdown, and no rational steward hands over the rails weeks before the first genuinely agentic peak season.
Why this matters now
Agentic commerce spent 2025 and early 2026 in a land-grab phase. Three protocols launched inside seven months: AP2 in September 2025, ACP later the same month, and UCP at NRF in January 2026. Each arrived with a coalition, an Apache 2.0 licence, and a GitHub repository.
Land-grab phases end in one of two ways. Either one standard wins outright and the others are absorbed, or the standards stop competing on features and start competing on legitimacy. The pattern in adjacent infrastructure suggests the second outcome is more common when no single vendor controls enough distribution to force the issue.
Neither ACP nor UCP controls enough distribution. ACP has OpenAI’s conversational surface and Stripe’s merchant base. UCP has Google Search, Gemini, and Shopify’s merchant estate. Both are large; neither is decisive.
That standoff has a predictable resolution, and the timing signals are unusually legible right now. For retailers deciding which rails to wire into their stack before the 2027 planning cycle, the governance question is not academic. It determines how much integration work is durable and how much is disposable.
There is also a narrower reason the question is live in August rather than October. Retail technology budgets for the following year are typically locked between September and November, and integration commitments made in that window are difficult to unwind. A merchant approving direct protocol work now is committing engineering capacity through a period in which the standard itself may change owners.
The stakes differ by merchant type. A brand selling through a single dominant surface can reasonably pick that surface’s standard and accept the concentration risk. A multi-surface retailer cannot, and it is the multi-surface retailers who fund most of the integration work that gives either protocol its claim to legitimacy.
Signal 1: both specifications froze in April 2026
The clearest signal is an absence. Public repositories for both protocols show no new specification release for four months, and the cadence before that was brisk.
ACP, maintained by OpenAI and Stripe as Founding Maintainers, shipped 2025-09-29 at launch, then 2025-12-12, 2026-01-16, 2026-01-30, and 2026-04-17. That is five releases in under seven months, with the January pair landing two weeks apart. The 2026-04-17 release added Cart, Feed, Orders, Authentication, and MCP integration to the protocol surface, which was a substantial expansion. Nothing has been tagged since.
UCP’s record is similar in shape. Its tagged releases run v2026-01-11 at the NRF launch, v2026-01-23, and v2026-04-08 (published April 9, 2026), which added cart capability, multi-parent support, deterministic schema resolution, error handling, and identity linking via OAuth 2.0. Nothing has been tagged since April either.
The UCP case is more pointed because the maintainers pre-announced the follow-on. A repository discussion posted on April 2, 2026 flagged a new version as coming “within two weeks” and described catalog and cart support, protocol hardening around versioning and error handling, and enhancements to payment handlers. The April 8 release delivered much of that. The roadmap language implied continued momentum that has not materialised in tag form.
This is not repository abandonment. UCP’s main repository shows activity through July 30, 2026, and satellite repositories for conformance tests and samples were updated across late June and July. Work is happening. It is simply not being cut into versioned specification releases, which is exactly the pattern you would expect when the question on the table is who gets to sign off, not what goes in.
| Protocol | Launched | Steward | Latest spec | Gap to today |
|---|---|---|---|---|
| ACP | Sept 29, 2025 | OpenAI + Stripe (Founding Maintainers) | 2026-04-17 | ~4 months |
| UCP | Jan 11, 2026 | Google + Shopify coalition | v2026-04-08 | ~4 months |
| AP2 | Sept 16, 2025 | Google, moved to FIDO Alliance (April 2026) | v0.2 (Human-Not-Present) | handed off |
| MCP | Nov 2024 | Linux Foundation | 2026-07-28 | ~3 weeks |
Read the bottom two rows against the top two. The two protocols that moved to neutral bodies are the two that have shipped or handed off cleanly. The two still under corporate stewardship are the two that stopped. That correlation is not proof of causation, but it is the kind of pattern that tends to precede a governance announcement rather than follow one.
It is worth noting that ACP’s own governance documentation names the destination explicitly. The repository describes a consensus-based decision model with escalation procedures and states the intended path as “neutral foundation stewardship as ecosystem matures.” The maintainers have already written down where this goes. The open question is when they judge the ecosystem mature, and the freeze suggests that judgement is being made now. The specification and its governance files are public on the ACP repository for anyone tracking the handoff.
Signal 2: MCP shipped its biggest release under neutral governance
On July 28, 2026, the Model Context Protocol released specification version 2026-07-28. It was the largest revision since the protocol launched, and the shape of it matters more than the fact of it.
The release moved MCP’s base protocol to stateless, self-contained requests with per-request capability negotiation, replacing the prior stateful model. It formalised an Extensions mechanism for optional, opt-in functionality negotiated during initialisation. It shipped Tasks for asynchronous long-running operations with polling, mid-flight input, and durable handles, plus MCP Apps for interactive UI elements rendered inline in conversations.
Every one of those changes is the kind that is hard to make under single-vendor stewardship and comparatively easy under a neutral body. A stateless rewrite breaks implementations. An extensions mechanism formally cedes surface area to third parties. Both require buy-in from parties who will not accept a competitor’s unilateral call.
MCP made those calls under Linux Foundation governance. That is the template, and it is a template both commerce protocols are already touching: ACP’s April release explicitly added MCP integration to its own surface, and MCP Apps give agent surfaces a rendering primitive that commerce flows will want.
The precedent is reinforced by AP2. Google launched the Agent Payments Protocol in September 2025 and moved it to the FIDO Alliance by April 2026, roughly seven months later. AP2 continued to develop after the handoff, reaching v0.2 with Human-Not-Present support. The payments layer of the agentic stack has already completed the transition that the commerce layer has not.
| Precedent | Origin steward | Neutral body | Time to handoff | Outcome after |
|---|---|---|---|---|
| AP2 | FIDO Alliance | ~7 months | continued releases, v0.2 | |
| MCP | Anthropic | Linux Foundation | ~14 months | largest-ever revision, July 2026 |
| ACP | OpenAI + Stripe | stated intent, unnamed | ~11 months and counting | frozen since April |
| UCP | Google + Shopify | none stated | ~7 months and counting | frozen since April |
ACP is now eleven months past launch with a stated intent and no named destination. UCP is seven months past launch, which is precisely the interval at which AP2 handed off. Both are inside the window the precedents describe.
Signal 3: the commercial stakes crossed a threshold in August
Governance handoffs need a forcing function, and the third signal supplies one. Salesforce published the second edition of its Agentic Enterprise Index in August 2026, built on aggregated Agentforce platform usage across thousands of businesses plus a proprietary study of 4,689 respondents fielded in May 2026.
The headline commerce finding is the one that changes the calculus: businesses that deployed agents on their digital commerce channels saw 4x higher online sales growth than those that did not. That is no longer a pilot-stage result. It is a number a board will act on.
The supporting metrics describe an adoption curve steepening rather than plateauing. The average Salesforce business ran five agents in February 2025 and thirteen by April 2026, a 7% compound monthly growth rate across fifteen months. Time to create and deploy a new agent fell 53%, from roughly four days to 1.9 days. Skills available per agent rose from two at the start of 2025 to six by year-end, and actions executed per account grew at a 31% compound monthly rate.
Faster deployment and broader per-agent capability both increase the number of protocol touchpoints a merchant maintains. When agents were slow to build and did two things each, supporting one checkout standard was a reasonable bet. When they take under two days to stand up and do six things each, every additional standard is a recurring tax on the same engineering team.
The deployment-speed figure deserves particular attention because it changes who makes protocol decisions. When standing up an agent took four days, the work sat with a platform team that thinks in architecture. At 1.9 days it moves toward individual product and merchandising teams, who will reasonably choose whichever standard their surface of interest already speaks. Decentralised choice accelerates fragmentation inside a single company, not only across the market.
That tax is already visible in who is paying it. Shopify, Salesforce Commerce Cloud, and Adobe Commerce each support all three protocols. Klarna, which had already integrated Google’s AP2 alongside Google Pay and Chrome autofill, adopted UCP in February 2026.
The platform and provider layer is absorbing fragmentation on the merchant’s behalf, which works but concentrates the cost with exactly the parties who have the most leverage to demand it stop. We covered the same dynamic when agentic commerce settled toward retailer-controlled checkout rather than agent-controlled checkout.
What the pattern suggests
Stack the three signals and a coherent reading emerges. Feature competition between ACP and UCP effectively stopped in April. The neutral-governance template has been demonstrated twice in the adjacent stack, once by AP2 in April and once, emphatically, by MCP in July. And the commercial cost of maintaining parallel standards became quantifiable in August.
The most likely explanation for a simultaneous four-month freeze at two competing standards is not coincidence and probably not exhaustion. It is that both camps have reached the limit of what unilateral stewardship can deliver. The next tranche of work for either protocol involves the hard interoperability questions: shared identity primitives, cross-protocol cart semantics, dispute and refund handling across agent boundaries, and settlement responsibility when an agent transacts across several merchants.
None of those can be specified credibly by one party’s maintainers. A UCP cart standard written solely by Google and Shopify will not be adopted by OpenAI’s surface, and the reverse holds equally. The technical roadmap has run into a governance wall, and the freeze is what hitting that wall looks like from outside.
If that reading is right, the sequencing likely runs: quiet coalition-building through Q3 and Q4 2026, protocol work continuing in repositories without tagged releases, then an announcement timed to a venue with retail authority. NRF in mid-January is the obvious candidate, not least because UCP itself was announced from the NRF 2026 keynote stage. Announcing a governance handoff at the same venue one year later would be a deliberate symmetry.
There is a supporting detail in how ACP is structured that fits this reading. The project requires Contributor License Agreements from both individual and corporate contributors, which is the standard legal preparation for eventually assigning a specification to a third party. CLAs are not proof of intent to transfer, but a project that has collected them has removed the main obstacle to doing so.
A secondary possibility worth holding: the handoff may not be to an existing body. A purpose-built commerce standards council, seeded by both coalitions, is plausible given that neither the Linux Foundation nor FIDO is an obvious home for merchant-side commerce semantics. The prediction is about neutrality of stewardship, not about which specific institution receives it.
Wider context: the surface layer keeps expanding
The pressure on protocol governance is not coming only from merchants. It is coming from the growth in agent surfaces that each need a checkout path.
UCP is already wired into AI Mode in Search and the Gemini apps, where US shoppers can buy from participating retailers without leaving the conversation. Google introduced Universal Cart at I/O 2026, a cross-merchant cart spanning Search, the Gemini app, YouTube, and Gmail, with UCP-powered checkout at Nike, Sephora, Target, Ulta, Walmart, Wayfair, and Shopify merchants. In August 2026 Google extended agentic capability into Maps, where the Ask Maps assistant now handles food ordering, hotel discovery, and real-time price comparison.
Each new surface multiplies the integration matrix. A merchant that wants presence across conversational search, a mapping assistant, a video platform, and an email client is not integrating once. Under fragmented standards it is integrating repeatedly against subtly different cart and identity semantics.
Geographic expansion compounds this. UCP is currently available to selected US merchants with rollout to Canada, Australia, and the UK slated before the end of 2026. Each market adds regulatory and payment-method variation on top of the protocol variation, and the EU AI Act’s high-risk system requirements take full effect in August 2026, adding a compliance dimension that neutral governance handles more comfortably than corporate stewardship does.
Amazon sits outside both coalitions with its own approach, having moved to open its marketplace to external AI agents on terms it controls. A large holdout weakens the case that either standard can win by distribution alone, which strengthens the case for a negotiated settlement.
Implications for retailers, platforms, and investors
For retailers, the practical read is that direct protocol integration carries more rework risk over the next two quarters than it did in the first half of 2026. Integrating through a platform or payment provider that already abstracts multiple standards is likely the lower-variance path into the 2026 holiday season and the 2027 planning cycle.
That does not argue for waiting. Agent readiness is a separate problem from protocol choice, and it is the one with immediate revenue attached. Catalogue structure, inventory accuracy, and machine-readable policy data determine whether an agent can transact at all, and none of that work is wasted regardless of which standard prevails. The emergence of scoring systems that assess whether AI agents can actually shop a given store suggests the market is already pricing readiness independently of protocol allegiance.
A useful test for any current integration proposal is to ask what survives a governance change. Catalogue feeds, structured product data, policy machine-readability, and inventory truth all survive. Bespoke cart-session handling written against one protocol’s current semantics probably does not. Weighting a roadmap toward the first category costs little and hedges the prediction either way.
For platforms and providers, triple-protocol support is becoming the entry ticket rather than a differentiator. The differentiation is moving to conformance tooling, fraud handling across agent boundaries, and dispute resolution when the buyer was not a human. Those are the areas where a neutral specification would define the floor and vendors would compete above it.
For investors, a governance handoff would be mildly negative for narratives built on proprietary protocol control and positive for infrastructure that monetises volume regardless of standard. The retail-technology vendors positioned to benefit are the abstraction layers, which is a familiar dynamic from earlier platform consolidations such as the SAP Hybris migration window that pushed retailers toward Shopify, Adobe, and commercetools.
| Scenario | Rough likelihood | What you would see first | Merchant response |
|---|---|---|---|
| Neutral handoff announced by NRF 2027 | Base case | coalition membership news, a named foundation, joint working-group posts | hold direct integration, deepen platform abstraction |
| Feature war resumes, both ship Q4 specs | Plausible alternative | tagged releases in September or October, competing cart semantics | pick the standard matching your dominant traffic surface |
| One standard wins on distribution | Less likely near term | a major surface deprecating support for the rival | migrate, accept rework cost |
| Permanent fragmentation, abstraction wins | Plausible | continued freeze past Q1 2027 with no governance news | never integrate directly, treat protocols as vendor concern |
Caveats: what could go wrong
The strongest counter-argument is timing, and it deserves weight. No rational steward hands control of checkout rails to a committee in the weeks before the first holiday peak where agentic purchasing is commercially material. If the freeze is pre-peak stabilisation, the correct inference is that both camps are locking specifications so implementers can ship against a stable target, and releases resume in Q1 2027 with governance unchanged.
This reading fits the evidence nearly as well as the governance reading. Repository activity continued through July at both projects, which is consistent with implementation support work rather than deadlock. A spec freeze before a peak season is ordinary engineering discipline, not a crisis.
The second counter-signal is that abstraction may relieve the pressure entirely. If Shopify, Salesforce Commerce Cloud, Adobe Commerce, and the major payment providers absorb fragmentation competently, no merchant ever feels the cost, and the forcing function for neutral governance evaporates. Fragmentation persists indefinitely because the parties who suffer from it have been insulated from it.
Third, strategic incentives cut against handoff on both sides. Protocol control shapes which surfaces get preferential merchant coverage, and both coalitions have spent real capital acquiring that position. Precedent from AP2 and MCP is suggestive, but neither of those protocols sat as close to transaction revenue as ACP and UCP do.
Fourth, venue risk. Even if a handoff happens, NRF is a retail trade show rather than a standards forum, and the announcement could land at a developer conference, in a foundation press release, or quietly in a repository governance file. A reader checking only NRF 2027 coverage could score this prediction wrong on a technicality. The falsifiable claim is the handoff itself inside roughly five months, not the stage it is announced from.
Finally, the Salesforce index is vendor research. The 4x commerce figure is drawn from Agentforce customers, who are self-selected for agent adoption, and the direction of causation between agent deployment and sales growth is not established. It is a reasonable directional indicator of commercial stakes and a poor basis for a precise elasticity.
How to track this
The prediction is cheap to monitor, which is part of why it is worth making. Four observable checkpoints would confirm or break it well before January.
- A tagged specification release at either repository before December 2026 would weaken the governance-deadlock reading substantially, especially if it lands unilaterally.
- Changes to governance files, CONTRIBUTING documents, or CLA terms in either repository would be an early and unambiguous tell, since foundation transfers require exactly those edits.
- New corporate names appearing as maintainers or in working-group membership, particularly names from the rival coalition, would indicate coalition-building ahead of an announcement.
- Joint appearances or co-authored technical posts between the two camps, which have so far been absent, would be the clearest public precursor.
The base rate for infrastructure protocols reaching neutral governance within eighteen months of launch is not high in general, but it is high within this specific cohort. Two of four agentic-era protocols have already done it. That is the strongest quantitative support the prediction has, and also its weakest point, because a cohort of four is a small sample.
Frequently asked questions
What exactly is being predicted, and how would someone check it?
The claim is that at least one of ACP or UCP will likely announce neutral foundation stewardship or a formal multi-vendor governance body by mid-January 2027. Verification is a check of both repositories’ governance files and public announcements. If both remain solely under their founding maintainers on January 31, 2027, the prediction failed.
Why does protocol governance matter to a retailer at all?
It determines whether integration work is durable. A specification controlled by one commercial party can change in ways that serve that party, and a merchant that wired directly into it absorbs the rework. Neutral stewardship historically slows change and raises the bar for breaking revisions, which is what makes integration investment safer.
Is the four-month spec freeze really unusual?
For these two projects, yes relative to their own history. ACP shipped five releases in under seven months and UCP shipped three in three months, then both stopped in April. Four months of silence would be unremarkable for a mature standard and is notable for standards this young and this contested.
Could this be read as agentic commerce losing momentum?
The evidence points the other way. Surfaces expanded through the freeze, with Universal Cart at I/O 2026 and Maps agentic ordering in August, and adoption metrics in the Salesforce August index steepened rather than flattened. The specifications stopped moving while deployment did not, which is closer to a governance bottleneck than to fading interest.
What is the strongest argument against this prediction?
Pre-holiday stabilisation. Freezing a specification so implementers can ship against a stable target ahead of peak season is normal practice and requires no governance explanation. If Q1 2027 brings tagged releases and no governance news, that reading was correct and this one was not.
Which protocol is more likely to move first?
ACP appears likelier on the documentary evidence, because its governance files already name neutral foundation stewardship as the intended path and it is eleven months past launch. UCP has a shorter track record and a broader founding coalition, which cuts both ways: more parties to satisfy, but also more parties who already expect shared control.
Does a handoff mean the protocols merge?
Not necessarily, and probably not initially. Neutral stewardship of two separate specifications is a common intermediate state, and interoperability bridges usually precede any merger. Converged cart and identity semantics under shared governance would deliver most of the practical benefit without either camp conceding its specification.
Should a merchant delay agentic integration until this resolves?
Delaying the readiness work would be a mistake, since catalogue structure, inventory accuracy, and machine-readable policies pay off under any standard. Delaying direct, protocol-specific integration in favour of a platform or provider that abstracts several standards looks like the lower-risk posture for the next two quarters. The distinction between readiness and integration is the useful one here.
How does this connect to agentic checkout becoming its own sales channel?
Governance and channel definition are two sides of the same maturation. As agentic checkout becomes a separately named sales channel with its own attribution and reporting, the standards underneath it face pressure to behave like utilities rather than competitive weapons. Channels that matter to finance teams tend to get standardised infrastructure.
What would make this prediction land early?
A governance-file commit in either repository, or an unexpected joint technical post between the two camps, would likely precede any formal announcement by weeks. Both are visible to anyone watching the repositories, which makes this one of the cheaper predictions in commerce infrastructure to track.