Strategic Intelligence Dossier · Volume 2 · 2026 Research

The Agent-Mediated Customer

In Volume 1, you built the case for your organisation's AI agents: the decisioning, the orchestration, the autonomous next best action. This volume addresses the other side of that equation. Your customer increasingly has AI agents of their own, searching, comparing and buying on their behalf. Are you ready to be found, trusted, and chosen by them?

AuthorSatya Upadhyaya, Martech Practitioner & Advisor
Data vintage2026 research
AudienceCMO · CDO · CTO · Board
ContinuationPicks up from Volume 1's closing reflection
$20.9B
AI platform retail spend in 2026, nearly 4x 2025
↑ eMarketer, Dec 2025
10–20%
Of e-commerce already captured by AI agents: $190 to $385B
$15T
B2B purchasing volume agent-intermediated by 2028
↑ Gartner forecast
90%
Of B2B buying will be AI agent-intermediated by 2028
Picking Up From Volume 1's Closing Reflection

The Dark Funnel Has Arrived

Your customers are already starting to shop without visiting your website. Instead of browsing categories and reading reviews, they ask an AI assistant for a recommendation, and the agent completes the purchase on their behalf. This isn't a future scenario. ChatGPT's Instant Checkout has been live since late 2025. Google's Universal Commerce Protocol is rolling out. Visa, Mastercard and PayPal have all launched agent commerce infrastructure in 2026.

The diagram below shows what changes. Toggle between the journey your Martech investments were built for, and the journey that is starting to run in parallel.

Discovery
Search engine
Types keywords, scrolls results
Asks an assistant
"Find me a winter jacket under $150"
Agent queries feeds
Reads structured product data via protocol
Evaluation
Browses product pages
Looks at images, reads copy
Reads reviews
Builds trust through social proof
Agent compares attributes
Price, specs, availability, trust score
Checks verified credentials
Machine-readable trust signals
Decision & purchase
Adds to cart
Manual checkout flow
Enters payment details
Forms, address, card number
Agent recommends & confirms
The click becomes approval, not exploration
Delegated payment executes
Pre-authorised, within set limits
Right now, most customers still follow this path. Every touchpoint above is one your brand has spent years designing: your site, your reviews, your checkout flow, your loyalty program. This is the journey almost every Martech investment to date has been built to serve.
The reframe at the heart of this dossier
Your Martech maturity and your agent-readiness are two different scores

You can be at Stage 3 or Stage 4 of the decisioning maturity ladder from Volume 1, with unified data, predictive NBA, even autonomous agents on your side, and still be invisible to the agents now shopping on your customers' behalf. Decisioning maturity optimises for people. Agent-readiness optimises for machines reading your data. Both matter, and neither substitutes for the other.

Sources: NEURONwriter, How to Prepare Your E-commerce Site for AI Shopping Agents, May 2026 · Ekamoira, How AI Agents Are Changing E-commerce, Feb 2026 · eMarketer Dec 2025 · Opascope Agentic Commerce Guide, April 2026

Visual 1 of 3 · Market Intelligence

The Agentic Commerce Imperative

Agentic commerce went from protocol announcements in 2025 to live, scaling infrastructure in 2026. The shift from SEO to AEO, or Answer Engine Optimisation, isn't a rebrand of digital marketing. It's a fundamentally different objective: not a high ranking on a search page, but a high inclusion rate within an AI system's recommendation logic.

Discoverability
From SEO to AEO

AEO is designed to provide an AI agent with structured, verifiable facts it can ingest and recommend as the "absolute answer." A brand's goal shifts from a high search ranking to a high inclusion rate within an AI system's recommendation logic.

Structured dataSchema markupAEO
Zero-click commerce
The click becomes approval

Agentic commerce is delegated shopping: a customer sets intent and guardrails, then an AI agent compares options and completes steps on their behalf. The "click" is no longer exploration. It's an approval of a decision the agent already made.

Delegated shoppingZero-clickAgent UX
New metrics
The agentic conversion rate

Traditional metrics like sessions and pageviews become irrelevant when a user never visits your site. The Agentic Conversion Rate, the percentage of AI agent inquiries resulting in a completed transaction, is emerging as the 2026 equivalent of A/B testing.

Agentic conversion rateQuery logsDark funnel
Protocols
A common language for agents

Standardised protocols, including UCP (Universal Commerce Protocol), ACP, and payment layers like AP2, allow any AI agent to talk to any merchant store. Bespoke APIs are too slow and expensive to maintain at the scale agentic commerce requires.

UCPACPAP2

The scale of the shift: 2026 to 2030

$900B–$1T
US retail revenue from agentic commerce by 2030
McKinsey
$3–5T
Global agentic commerce value by 2030
McKinsey
4.4×
Higher conversion for AI-generated product recommendations vs traditional search
McKinsey via MetaRouter 2026
60%
Of shoppers expect to use AI agents to shop within 12 months
Kearney 2026
63%
Of European shoppers already use AI in their shopping journey
commercetools 2026
1–4%
Of global digital transactions could be agentic by 2029
Tredence 2026

The practitioner's lens: Every one of these statistics describes the same underlying shift, from a brand competing for human attention to a brand competing for machine inclusion. The brands moving first aren't necessarily the biggest. They're the ones whose product, pricing and trust data was already structured enough to plug into this new layer with minimal rework. Data discipline, not size, is the early advantage.

Sources: Stord, State of AI in E-Commerce 2026 · commercetools, 7 AI Trends Shaping Agentic Commerce 2026 · NEURONwriter 2026 · Opascope, AI Shopping Assistant Guide 2026 · MetaRouter, Agentic Commerce Trends 2026 · Tredence, Agentic Commerce in Banking 2026

Visual 2 of 3 · Market Intelligence

Trust Infrastructure: The New Competitive Layer

This is the section every banking, telco, insurance and government leader should read most closely. While retail leads on commerce volume, financial services is building the trust and identity infrastructure that the entire agentic economy will run on, and it's happening now, not in 2030.

The problem
How do we know an AI agent is doing exactly what was asked, and nothing more?
Without a verified link between a person and their agent, fraud, misrepresentation and unauthorised transactions all become easier
Regulators expect explainability. Why did an autonomous system initiate, modify or reject a transaction?
Banks have always verified human identity. They've never had to build reputation infrastructure for non-human actors, until now
What's being built, right now, in 2026
Experian Agent Trust, a "Know Your Agent" (KYA) framework linking AI agents to verified consumer identity
Mastercard Agent Suite & Verifiable Intent, provable user authorisation for agent-led commerce
Visa Intelligent Commerce, with live pilots with DBS and Santander, and Europe's first end-to-end AI agent payment completed
FIS Agentic Commerce Platform, giving issuing banks the ability to identify and authorise agent-initiated transactions

Why this matters beyond banking

$15T
B2B purchasing volume routed through AI agent exchanges by 2028
Gartner
20%
Of monetary transactions will be programmable by 2030
Gartner
2.3/5
Average "AI Trust Maturity" score across 500 organisations surveyed
McKinsey 2026
€35M
Maximum EU AI Act penalty (or 7% of global turnover) for non-compliance
IMF Working Note, April 2026
44%
Of finance teams will use agentic AI in 2026, a 600%+ increase
Wolters Kluwer
11%
Of organisations have agents in production, out of 99% planning to
Neurons Lab 2026

The cross-industry signal: A 2.3 out of 5 AI Trust Maturity score, and only 11% of organisations with agents actually in production despite 99% planning to, tells the real story. The technology isn't the constraint. Trust, governance and verified identity are. For telco, insurance, and government-adjacent organisations watching retail and banking move first, the trust infrastructure being built by Visa, Mastercard and Experian today is the layer your agent-readiness will eventually plug into, regardless of your industry. Building data discipline and verification readiness now isn't premature. It's sequencing.

Sources: Medium/ACHIVX, Credit of Trust: How Banks Should Score AI Agent Reputation, May 2026 · Tredence, Agentic Commerce in Banking 2026 · Experian Agent Trust announcement, April 2026 · Mastercard Verifiable Intent, March 2026 · IMF Notes Vol. 2026 Issue 004 · Neurons Lab, Agentic AI in Financial Services 2026

Visual 3 of 3 · Market Intelligence

Are You Legible to a Machine?

This is the same product, viewed two ways. The first is how it appears to a person browsing your website. The second is how it appears to an AI agent reading your underlying product data. The gap between them is the gap this dossier exists to close.

What a human sees
Alpine Trek Insulated Jacket
★★★★☆ 4.3 (212 reviews)
$129.00
Stay warm on the trail this winter. Our best-selling jacket combines lightweight insulation with a durable outer shell, perfect for everything from city commutes to weekend hikes.
Looks great. Clear, persuasive, on-brand.
What an AI agent reads
product.name"Alpine Trek Insulated Jacket"
product.price129.00 USD
product.availabilitynot specified
product.gtin / skumissing
product.attributes.warmthRatingmissing
product.attributes.materialmissing
review.aggregateRatingnot in schema
merchant.trustSignal (KYA)absent
Mostly unreadable. Likely skipped by the agent.
Present & structured Missing or unstructured

The same product. Two completely different experiences. The human sees a polished, persuasive page: four-and-a-half stars, a warm description, a clear price. The agent sees a name and a price, and almost nothing else it can verify or compare. No availability, no attributes, no trust signal. To a human, this product looks great. To an agent comparing twelve jackets across six merchants, it's statistically invisible, not penalised, simply never considered.

Illustrative example based on common schema gaps identified across e-commerce catalogues. Sources: Opascope AI Shopping Assistant Guide 2026 · NEURONwriter, Preparing for AI Shopping Agents 2026 · ucphub.ai, AI in E-commerce 2026

Cross-Industry Calibration · Market Intelligence

Vertical Adoption Timeline: When Does This Reach Your Industry?

Not every vertical moves at the same pace, and not always in the order you'd expect. Select an industry to see the anticipated maturity curve from now through 2030, and the specific signals driving that timing.

How to use this timeline: The "now" marker isn't a future prediction. It describes infrastructure and pilots that are already live. The further-out horizons depend on regulatory clarity, consumer trust, and how quickly competitors in your vertical move. The organisations that treat this as "watch and wait" risk discovering their competitors became the default recommendation while they were still deciding whether the trend was real.

Sources: Gartner B2B forecast via Digital Commerce 360 · Tredence Agentic Commerce in Banking 2026 · commercetools 7 AI Trends 2026 · Mirakl, Agentic Commerce: The Next Revolution 2026 · MetaRouter 2026

Part II · Readiness Framework

The Agent-Readiness Ladder

This is a new instrument, distinct from the decisioning maturity ladder in Volume 1. That ladder measured how well you understand and act on your customer. This one measures how well agents acting on your customer's behalf can find, understand, and trust you. Click each stage to explore.

Stage 1
Invisible
Unstructured
Product, service and pricing data exists only as human-readable web content. No schema markup. Agents cannot parse or compare your offers.
Stage 2
Visible
Structured
Core product data is structured (schema.org, product feeds). Agents can find and list your offers, but cannot yet verify or fully trust them.
Stage 3
Legible
Protocol-connected
Integrated with commerce protocols (UCP/ACP). Trust signals such as reviews, policies and guarantees are exposed as structured data agents can weigh in comparisons.
Stage 4
Trusted & Transactable
Verified
Verified credentials (KYA-equivalent) in place. Agents can complete delegated, payment-authorised transactions directly. Agentic conversion measured and optimised.
Select a stage above to explore

How this relates to Volume 1: A Stage 4 decisioning-maturity organisation (Agentic AI, per Volume 1) could still be Stage 1 on this ladder (Invisible). Their own agents are sophisticated, but their product and trust data was never structured for outside agents to read. Conversely, a Stage 2 organisation (Triggered Journeys) with disciplined product data and clean schema markup could already be Stage 2 or 3 here. The two ladders are related but independent, and most organisations haven't yet assessed where they sit on this one at all.

Framework developed by Satya Upadhyaya, informed by: Stord State of AI in E-Commerce 2026 · NEURONwriter 2026 · Ekamoira, AI Agents and Open Protocols 2026 · Experian Agent Trust 2026

Part II · Readiness Framework

Agent-Readiness Diagnostic

Score your organisation across five dimensions, each one genuinely new and not a re-skin of the Volume 1 decisioning diagnostic. This measures whether your organisation is discoverable, legible and trustworthy to the AI agents now shopping, comparing and transacting on your customers' behalf.

Structured data coverage
Are products, services or offers described in schema markup an agent can parse, not just human-readable copy?
Protocol integration
Is the organisation connected to emerging commerce/agent protocols (e.g. UCP, ACP, AP2) or still API-only on a case-by-case basis?
Verified credentials & identity
Can your organisation, products or offers be cryptographically verified? This is the "Know Your Agent" equivalent for your brand.
Machine-readable trust signals
Are reviews, ratings, return policies, guarantees and pricing promises exposed in a form an agent can read and weigh, not buried in marketing copy?
Agentic channel measurement
Can you currently detect, attribute and measure traffic, queries and conversions coming from AI agents acting on customers' behalf?
Agent-readiness score
·
Score all five dimensions
This is a new instrument. It measures readiness for the agent-mediated customer, separate from your Volume 1 decisioning maturity stage.
Where to focus first

Score each dimension on the left to see your priority gaps.

Part III · Investment Case · Startups & SMB
For startups & small organisations

Business Case: Getting Agent-Ready

For organisations at Stage 1 (Invisible) on the agent-readiness ladder, which is most organisations today. The goal isn't to chase every protocol on day one. It's to get the foundational structured data and trust signals in place that make you legible to agents at all, using accessible tools and a focused first project.

Current state: Invisible
·Product/service information exists only as web copy, images and PDFs
·No schema markup, so search engines and agents see an unstructured page
·Reviews and ratings exist on third-party platforms but aren't linked or structured
·No visibility into whether AI platforms are already citing, or ignoring, the brand
Target state: Visible
Core schema markup live across key pages (Product, Organization, FAQPage, Service)
Pricing, availability and core attributes exposed in structured, machine-readable form
Reviews and trust signals connected and structured where the platform supports it
A baseline AI visibility/citation check established, so you know where you stand today
920%
Average lift in AI-driven traffic reported by AEO agency clients
AEO Engine 2026
41%
Of structured data now uses JSON-LD, the preferred AEO format
LoudFace 2026
5
Schema types that compound for AEO: Organization, FAQPage, Article, BreadcrumbList, Service
LoudFace 2026
<$30/mo
Entry-level AEO monitoring tools now accessible to small teams
Stackmatix 2026

The honest starting point: Most small organisations don't need a multi-protocol integration strategy yet. They need their existing site to stop being invisible to the AI systems already answering questions about businesses like theirs. Getting core schema right, even just five types done well, is a half-day-to-week fix with disproportionate impact, because so few competitors have done it.

Structured data audit & baseline
$1.5K–$4K
Audit current schema coverage, run a baseline AI citation/visibility check across 2-3 platforms (ChatGPT, Perplexity, Google AI Mode).
Core schema implementation
$2K–$6K
Implement the five compounding schema types: Organization, Product/Service, FAQPage, Article, BreadcrumbList, across priority pages.
Trust signal structuring
$1.5K–$4K
Connect and structure reviews, ratings, return policies and guarantees so they are machine-readable, not just displayed.
Monitoring setup & advisory
$1K–$3K
Set up ongoing AI citation tracking and a simple monthly review. Hands-on guidance so the team can maintain schema independently.
Total investment range
$6K – $17K
Programme duration
4–6 weeks
Ongoing monitoring cost
$0–$30/mo
Week 1

Baseline & audit

Run AI citation checks across 2-3 major platforms for the brand and key products/services. Audit current schema coverage. Identify the highest-value pages to prioritise.

Weeks 2–3

Core schema implementation

Implement Organization, Product/Service, FAQPage, BreadcrumbList and Article schema on priority pages. Validate against schema.org and platform-specific requirements.

Weeks 4–5

Trust signal structuring

Structure reviews and ratings as aggregateRating data. Expose return policies and guarantees as structured Service/Offer attributes where supported.

Week 6

Re-test & handover

Re-run the citation check to measure change from baseline. Set up a simple monthly monitoring routine. Train the team to maintain and extend schema as new products/services launch.

High

Incorrect schema triggers platform penalties

Incorrect or misleading schema markup can trigger manual actions if it violates platform spam policies. Mitigation: validate every implementation against current platform guidelines before publishing, and don't copy-paste templates without per-page customisation.

Medium

Schema implemented but not maintained

Schema can drift out of sync with actual product/service data over time. Mitigation: build schema updates into the existing content publishing workflow, not as a separate task.

Low

Limited visibility into actual AI citation impact

AI citation monitoring is still partly manual. Mitigation: start with a small number of tracked prompts most relevant to the business, and accept directional rather than precise measurement initially.

Recommended tools: entry level

Schema.org + Google Rich Results Test
Free
The foundation: schema.org defines the vocabulary, and Google's free testing tool validates implementation before publishing.
Best for: every organisation, regardless of size, as the starting point
Otterly AI / Peec AI
From under $30/mo
Entry-level AI visibility and citation monitoring, tracking whether ChatGPT, Perplexity and others mention your brand on relevant queries.
Best for: solo marketers and small teams establishing a baseline
Surfer SEO (AI Tracker)
Existing SEO tool tier
If already using Surfer for SEO, the AI Tracker module adds LLM visibility monitoring alongside existing content scoring, with no new platform needed.
Best for: teams with an existing SEO tool subscription

The five schema types worth getting right first

  • Organization: establishes who you are, unambiguously, for entity recognition
  • Product or Service: price, availability, and core attributes in structured form
  • FAQPage: 5 to 8 real questions customers actually ask, answered directly
  • BreadcrumbList: clarifies your site's taxonomy and category structure
  • Article: for any content marketing, positions it as a citable source

Sources: AI Advantage Agency, AEO Agency Pricing 2026 · AEO Engine, Best AEO Agencies for Ecommerce 2026 · LoudFace, Schema Markup for AEO 2026 · Stackmatix, Best AEO Tools 2026 · Yotpo, Master Ecommerce AEO 2026

Part III · Investment Case · Enterprise
For established & enterprise organisations

Business Case: Agent-Ready at Scale

For organisations moving from Visible to Legible and Trusted & Transactable: protocol integration, verified credentials, and measurement infrastructure across a large product catalogue, multi-market footprint, or regulated environment. This is where banking, telco, insurance and large retail need to be building now.

Current state: Visible
·Core schema in place on key pages, but coverage inconsistent across the full catalogue or product set
·No integration with emerging commerce/agent protocols (UCP, ACP, AP2)
·No verified credential or "Know Your Agent"-equivalent infrastructure
·No measurement of agent-originated traffic, queries or conversions
Target state: Legible & Trusted
Structured data and trust signals consistent across the full catalogue/product set, multi-market where relevant
Integrated with at least one major commerce/agent protocol in production
Verified credentials in place, aligned to emerging KYA-equivalent frameworks for the sector
Agentic conversion rate and agent-originated traffic measured and reported alongside traditional metrics
$2.5K–$25K/mo
Typical AEO agency retainer range, depending on scope
AI Advantage Agency 2026
4.4×
Higher conversion for AI-recommended products vs traditional search
McKinsey via MetaRouter
2–4
AI platforms typically covered in mid-tier AEO citation tracking
AI Advantage Agency 2026
$97B
Financial services agentic AI investment forecast by 2027
WEF / Accenture

The strategic case in one sentence: Enterprises that wait until agentic commerce volumes are unmistakable will be retrofitting structured data, protocol integration and verified credentials under competitive pressure, at exactly the moment competitors who started in 2026 are compounding an early-mover advantage in agent recommendation rates that gets harder to close every quarter.

Enterprise structured data programme
$60K–$150K
Schema and structured data coverage across the full catalogue, multiple markets/brands, integrated into existing PIM/content workflows.
Protocol integration
$80K–$200K
Integration with one or more commerce/agent protocols (UCP/ACP) and payment-layer readiness (AP2-equivalent), including technical architecture work.
Verified credentials & trust infrastructure
$50K–$140K
Implementation aligned to sector-relevant KYA-equivalent frameworks (e.g. Experian Agent Trust, Mastercard Verifiable Intent for financial services).
Agentic measurement & reporting
$40K–$90K
Agentic conversion rate tracking, agent-originated traffic attribution, and integration into existing analytics and reporting infrastructure.
Governance & advisory
$60K–$120K
Independent advisory on protocol selection, governance alignment with EU AI Act and sector regulation, and cross-functional programme leadership.
Total investment range
$290K – $700K
Programme duration
6–9 months
Ongoing AEO retainer
$2.5K–$25K/mo
Months 1–2

Diagnostic & protocol strategy

Run the Agent-Readiness Diagnostic across the organisation. Assess current schema coverage at scale. Evaluate which commerce/agent protocols and trust frameworks are most relevant to the sector and markets.

Months 3–5

Structured data & protocol integration

Roll out structured data coverage across the catalogue, integrated into PIM workflows so it doesn't become a one-off project. Begin protocol integration in a priority market or product line.

Months 6–7

Verified credentials

Implement verified credential infrastructure aligned to sector frameworks. Coordinate with payments, legal and compliance. This is where banking/insurance organisations should lean on existing regulatory relationships.

Months 8–9

Measurement, scale & review

Agentic conversion rate and agent-originated traffic live in reporting. Expand structured data and protocol coverage to additional markets/product lines based on early results. Board-level review against business case.

High

Protocol landscape still consolidating

UCP, ACP and other protocols are evolving rapidly, and today's integration choice may need revision. Mitigation: prioritise structured data and trust signal work, which is protocol-agnostic and remains valuable regardless of which protocols win.

High

Regulatory requirements outpace internal readiness

EU AI Act and sector-specific regulation (especially financial services) carry significant penalties for non-compliance. Mitigation: legal and compliance embedded from Month 1, not bolted on after technical implementation.

Medium

Structured data programme becomes a one-off project, not embedded

Schema coverage decays if not built into ongoing content/PIM workflows. Mitigation: success criteria explicitly include workflow integration, not just initial coverage percentage.

Medium

Cross-functional ownership unclear

This work spans marketing, technology, data, legal and, in financial services, compliance, and without clear ownership it stalls. Mitigation: a single accountable owner with a cross-functional steering group, mirroring the Volume 1 governance recommendations.

Low

Limited early measurement signal

Agentic traffic volumes may still be small in absolute terms in 2026, making early ROI hard to demonstrate. Mitigation: frame early measurement as building the capability and baseline, with ROI expectations set for 2027-2028 as volumes scale.

Sources: AI Advantage Agency, AEO Agency Pricing 2026 · MetaRouter, Agentic Commerce Trends 2026 · WEF/Accenture financial services AI investment forecast · IMF Notes Vol. 2026 Issue 004 · Experian, Mastercard, Visa, FIS announcements 2026

Next Steps
Ready to find out where you stand on both ladders?
Satya's Martech Diagnostic now includes an Agent-Readiness assessment alongside the decisioning maturity diagnostic from Volume 1, giving you a complete picture of how ready your organisation is for both your AI agents and your customers' AI agents.
Connect with Satya → Run the agent-readiness diagnostic ↑
Closing Reflection: The Question After This Question

You're now legible to the agents searching, comparing and buying. The next question is what happens once they're inside your organisation.

This dossier has focused on the threshold: the moment before a transaction, when an agent is searching, comparing, verifying. Getting that right is necessary, but it's also just the door.

What happens after the agent walks through it? Once a customer's agent has bought from you, or a B2B procurement agent has placed a recurring order, or a financial agent has opened an account, your organisation now has an ongoing relationship with a non-human counterparty. Service requests, renewals, complaints, negotiations, account changes: all of these will increasingly come from agents too, not just from the humans they represent.

This raises questions Volumes 1 and 2 have only touched on: What does your operating model look like when a meaningful share of your "customers" are AI agents acting under delegated authority? What does your team need to know, and how does their role change, when the person on the other end of a support ticket might not be a person at all?

That's a conversation about organisational design, team capability, and the human work that remains essential precisely because the agentic work has been automated. It deserves its own dossier.

This dossier ends here. The next one begins with the team on the other side of the agent.

Sources: Stord State of AI in E-Commerce 2026 · commercetools 7 AI Trends 2026 · Tredence Agentic Commerce in Banking 2026 · Gartner via Digital Commerce 360 · IMF Notes Vol. 2026 Issue 004 · Neurons Lab 2026