Product1 distinct publisher2 min readUpdated
Agent listings on AWS Marketplace went from about 1,000 to more than 4,000 in a year, and searches for them now rank third. The sales motion above the self-service line has not moved.
The Product Desk · Product desk

Compiled by The Product DeskSomething wrong?How this is made
A keyword's rank is ordinal, and ordinal numbers hide volume. Third place means the phrase passed 61 others in the search log [1][21]; it does not say how many buyers typed it, or what any of them signed. The listing count is the sturdier number: about 1,000 agent offerings a year ago against more than 4,000 now [2], a net 3,000 additions in twelve months [11] and roughly a tenth of a catalogue that is approaching 40,000 applications [3][4].
Supply that grows fourfold in a year [12] runs straight into how the discovery layer behaves. Shashi Bellamkonda of Info-Tech Research Group says agent-driven search "may only look at less than 70% of the solutions available", and attributes that to model behaviour rather than anything AWS built [8]. His remedy is manual: reprompt, add detail, and treat what comes back as a high-quality draft rather than a finished decision, reviewed by a person [10]. That is the discovery gain and the discovery tax in the same paragraph. The buyer stops reading category pages and starts doing recall work on the model's behalf.
The boundary between machine work and paid work is drawn fairly precisely in the reporting. Bret Greenstein of West Monroe Partners lists what marketplaces now absorb: discovery, vendor comparison, licensing, entitlement management, auditing, renewals, and parts of contracting and procurement [7]. AWS's Matt Yanchyshyn says the support extends past discovery into deployment, with due diligence a heavy adoption area beyond pricing and transactions [5], and describes the agentic layer as being about customer success, private pricing and supporting bigger deals [13]. Note the verb. Supporting bigger deals is not closing them, and the plain self-service click-and-go path still covers much of the marketplace [18].
The money forecast is worth doing by hand. Market Intelo puts agentic procurement software at $1.2 billion in 2025, growing 24% a year to nearly $9 billion by 2034 [14]. Compound 24% across the nine years from 2025 and $1.2 billion lands at about $8.3 billion [15]; getting to $9 billion needs about 25.1% [16]. The gap is small, and it is the kind of slack that suggests the headline figure and the growth rate were fitted to each other rather than derived from the same data. Either way, take the rate at face value and the spend on this category grows about a quarter a year while one marketplace's agent shelf grew fourfold in one [17]. Sellers are arriving faster than the budget is, which is a crowding story, and crowding is settled by whatever ranks answers.
Follow any of these and your For You feed starts watching them — no settings page required.
Ranked by verification strength, evidence, and original report placement.
A year ago "AI agents" ranked 64th among keyword searches across AWS Marketplace; it now ranks third, according to Matt Yanchyshyn, vice president of AWS Marketplace & Partner Services.
AWS Marketplace had about 1,000 agent offerings last year and has more than 4,000 this year, Yanchyshyn told ZDNET.
Yanchyshyn says the marketplace supports not just discovery but deployment, and that due diligence is an area seeing a lot of adoption, beyond traditional pricing and transaction support.
AI agents handle the administrative and due diligence tasks of software acquisition and deployment, but larger enterprise applications still require human sales teams.
Bret Greenstein, chief AI officer at West Monroe Partners, says agent-powered marketplaces have been reducing IT overhead by helping with application discovery, vendor comparison, licensing, entitlement management, auditing, renewals and even portions of contracting and procurement.
Shashi Bellamkonda, principal research director at Info-Tech Research Group, says AI agents "may only look at less than 70% of the solutions available" and that this is not an AWS issue but the way the models work.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Single trade source, vendor-reported metrics
All quantitative claims come from one ZDNET article, and the load-bearing figures (agent listing counts, keyword rank, catalogue size) are self-reported by an AWS executive with no listing-level data, methodology, or independent verification. The critical and governance claims are named-expert opinion rather than measurement, and the one third-party market figure is internally inconsistent. That supports directional reading but not verification.
Real supply-side traction, buyer-side usage undisclosed
There is concrete adoption evidence on the supply and interest side: 4,000+ agent listings from about 1,000, a keyword rank move from 64th to third, and at least one named vendor deployment (CharmHealth's Ria) using an agent for pre-sales evaluation. What is missing is demand-side depth - no transaction counts, spend, attach rates, or named enterprise buyers - and AWS itself says self-service still covers much of the marketplace while large deals stay with human sellers.
Mildly overstated: supply counts standing in for demand
The article's own framing is comparatively restrained - it repeatedly says human reps and engineers are not being displaced - but the 'rising demand' narrative is carried by listing counts and keyword rank, which measure supply and curiosity rather than usage or spend. The gap widens where the cited definition of agentic procurement (autonomous negotiation and end-to-end contract execution) is far ahead of what AWS describes shipping, and where a 300% listing jump sits against a ~24% projected spend CAGR whose endpoint does not even reconcile.
Promotional sourcing throughout
Nearly every voice benefits from the trend being real: an AWS vice president describing his own marketplace's growth, a consultancy chief AI officer who sells agentic transformation and governance work, an analyst firm principal whose research covers software selection, a SaaS CEO promoting her company's own AI assistant, and a market-research firm whose product is category forecasts. The caveats are meaningful precisely because they come from the two sources with the least direct stake in AWS listing counts.
Directionally credible, weakly verified
The direction of travel - more agent listings, more agent-shaped buyer queries, human sellers retained above the self-service line - is coherent and consistent within the source, and the arithmetic derivations are checkable. But with one publisher, vendor-sourced counts, no demand-side metrics, and a forecast that fails its own math, confidence in the magnitudes should stay well below confidence in the direction.
leadership
Slack Code makes the chat window a coding surface, and a platform call for engineering leaders1 distinct publisher
security
ToxicPanda 2.0 Widens From 16 Apps to 140, and From Overlays to ADB Shell3 distinct publishers
build
Two Actions, One Loose Policy: The Bedrock Wildcards That Widen A Least-Privilege Grant1 distinct publisher
security
Leaked AWS keys keep working: 526 root, 242 admin, and a rotation rate of 13.7 percent1 distinct publisher
Distinct publishers with included, body-backed reporting in this cluster.
1 article · August 21, 2026