AI Advertising and Marketing Automation: A Value-Chain Study

This report builds on the latest public information as of May 19, 2026, and focuses on one investment question: across advertising, marketing, customer acquisition, customer operations, the content supply chain, data collaboration, and sales conversion, which AI capabilities have already moved from "internal efficiency gains" into a genuine commercialization stage, one where customers are willing to pay, advertisers are willing to add budget, and platforms can amplify the profit pool. The conclusion: the largest profit pool still sits first with the media/e-commerce/advertising platforms that hold distribution rights, closed-loop conversion data, and automated delivery systems; second in line are the software platforms that own first-party data, CRM/CDP, the content supply chain, and approval/governance workflows; and the most exposed to being commoditized or eroded by price wars are standalone creative generation, low-barrier agentic execution, and pure outbound automation tools detached from data and distribution. In 2025, global advertising revenue surpassed $1 trillion, with digital advertising at 73.2% and user-generated platforms accounting for 54.7% of global advertising revenue. At the same time, marketers' adoption of generative AI has shifted from concept to mainstream: a Gartner survey shows 47% of marketers already use generative AI in their work.

Core Conclusions

• Within the AI value chain, AI advertising and marketing automation has moved from a "model-capability showcase layer" to a "revenue-execution layer." What commercializes first is not general-purpose generation, but products tied directly to ad budgets, customer data, attribution, and delivery automation: Google Performance Max / AI Max, Meta Advantage+, Amazon Ads, AppLovin Axon, The Trade Desk Kokai, along with software platforms such as Salesforce, HubSpot, Klaviyo, and Adobe that embed AI into marketing workflows.

• The scenarios that commercialize first and carry the highest revenue certainty are automated delivery and budget allocation. Google states that search ads with AI Max enabled typically deliver 14% more conversions or conversion value at similar CPA/ROAS; for ad groups still dominated by exact/phrase match, the typical lift can reach 27%. The Trade Desk disclosed that Kokai achieved 5x ROAS in a U.S. and Canada sample, with average CPA down 34% in the Kokai beta sample. Amazon DSP Performance+ directly automates the high-complexity DSP configuration process and optimizes it with first-party signals and a real-time bidding model.

• The second scenario already generating real revenue is "closed-loop retail media + AI optimization." Amazon's advertising services were disclosed across the four quarters of 2025 at $13.921 billion, $15.694 billion, $17.703 billion, and $21.317 billion respectively; on a per-quarter disclosure basis, that implies roughly $68.635 billion for the full year, and Q1 2026 still reached $17.243 billion. Amazon also connected Sponsored Products and Brand Prompts to Rufus, and officially disclosed that among shoppers who encounter a brand Prompt, close to 20% go on to engage in further conversation with that brand. Alibaba likewise stated explicitly that growth in Taobao-Tmall customer-management revenue benefits from rising penetration of the AI marketing tool "Quanzhantui."

• The third scenario already generating real revenue is CRM / lifecycle marketing / customer-operations Agents. Salesforce FY2026 revenue was $41.5 billion, Agentforce ARR reached $800 million, Agentforce + Data 360 ARR exceeded $2.9 billion, and 29,000 Agentforce deals have closed. HubSpot's Q1 2026 revenue was $881 million across 299,458 customers, and the company explicitly stated that Customer Agent, Prospecting Agent, and Data Agent have begun delivering results for customers. Klaviyo's Q1 2026 revenue was $358 million with more than 196,000 customers; its cohort of customers above $50,000 ARR grew 38% year over year, and NRR was 110%. These are not "trial" stages but have entered the phase of subscription expansion and customer upsell.

• Creative generation already has real demand, but a standalone profit pool may not exist on its own. Amazon's Video Generator is offered explicitly "at no additional cost"; Google, Meta, Pinterest, and Snap all fold creative generation, copy personalization, automatic asset assembly, and automated testing into their advertising platform product bundles. In other words, AI creative is the easiest thing for platforms to "make free / bundle." The defensible point for standalone creative companies is not "generation" itself, but brand governance, approvals, DAM, compliance, cross-region localization, asset reuse, and workflow integration. This is precisely the direction Adobe GenStudio is betting on.

• The profit pool for AI marketing Agents will not mainly remain in the "pure chat interface," but will settle with the platforms that hold the system of record and customer context. Salesforce emphasizes unified data, workflows, and Agents; HubSpot binds Smart CRM, 2,000+ integrations, and Agents together; Klaviyo states explicitly that the value of its Agent comes from 14 years of accumulated marketing intelligence, billions of consumer interactions, and its customer data platform. An Agent with no system of record, no event stream, and no approval chain easily degenerates into a demo layer.

• One of the most concrete "hard proof" samples worth watching in the China market is Baidu. In Q3 2025, Baidu disclosed AI-native marketing services revenue of ¥2.8 billion, up 262% year over year; in Q4, that revenue was ¥2.7 billion, up 110% year over year. This is a rare case among global internet platforms of directly breaking out AI-native marketing revenue as a public disclosure. Alibaba has tied "Quanzhantui" penetration directly to growth in Taobao-Tmall customer-management revenue; Tencent's public commentary also acknowledges that its advertising platform benefits from AI-driven enhancement.

• The first winner of the profit pool is still the platform, not the tool. In Q4 2025, Google Search and Other advertising revenue was $63.1 billion and YouTube advertising revenue was $11.4 billion; Meta's advertising revenue for the first three quarters of 2025 was $138 billion, and Reuters, citing Meta's Q4 figures, put single-quarter advertising revenue at $58.14 billion. On disclosed figures, Meta's 2025 advertising revenue was roughly $196 billion. By comparison, The Trade Desk's 2025 revenue was $2.9 billion, AppLovin's 2025 revenue was $5.48 billion, and Salesforce Agentforce ARR was $800 million. The scale gap shows that the delivery efficiency and creative efficiency that AI improves are first internalized by the traffic platforms.

• The second winner is platform-type marketing software, not point tools. Adobe's Q1 FY2026 revenue was $6.4 billion, with AI-first ARR up more than threefold year over year; its GenStudio already strings creative generation, brand checks, approvals, DAM, Workfront, and CDP into a single content supply chain. Publicis, meanwhile, uses Epsilon + Lotame + CoreAI to extend into data, identity, and retail media, with Lotame covering 1.6 billion IDs and 100+ data sources, and claims that combined with Epsilon it can reach more than 90% of consumers worldwide.

• Still at the pilot, concept, subsidy, or low-price-competition stage are the single-point AI marketing tools detached from data and channels. Typical examples include general-purpose copy/image/short-video ad generation, marketing Agents lacking a first-party data foundation, AI SDRs reliant on cold email/cold outreach, and the many tools that sell only "content generation" without governance or activation capabilities. Klaviyo's Composer is still in private beta; many AI creative tools can demonstrate efficiency gains, but publicly auditable evidence of ARR, retention, budget migration, and ROAS improvement remains limited.

• AI's impact on the advertising industry is reallocation, beyond mere "efficiency." It raises delivery efficiency and lowers creative costs while also rewriting the bargaining power of agencies, DSPs, SSPs, data platforms, and marketing SaaS. The weight rises for programmatic buyers, retail media, closed-loop advertising platforms, CRM/CDP, and the content supply chain; value compresses for pure manual optimization, low-value-add creative outsourcing, traditional DMPs, and tool stacks that sell reach but not data/attribution/governance. Google is no longer pursuing a separate third-party cookie prompt in Chrome; at the same time, rules such as CCPA/CPRA, DMA, DSA, PIPL, and ATT make "usable, compliant, attributable first-party data" more valuable.

• The most important catalyst over the next 12–24 months is not a new model release, but "budget migration + external monetization." Key items to track: whether Meta/Google/Amazon/AppLovin continue to expand the share of AI-automated advertising revenue; whether Salesforce/HubSpot/Klaviyo/Braze turn Agents from bundled features into upside drivers of ARPU, RPO, and NRR; and whether Adobe/Publicis prove that enterprises will pay separately for the content supply chain, brand governance, and cross-channel activation.

• The biggest risk is not technical failure, but platform commoditization, tightening privacy, ROI disproof, and budget cycles. The FTC has launched "Operation AI Comply" targeting false AI claims; the EU has fined Apple and Meta under the DMA; the DSA requires a searchable ad repository for advertising transparency; the UK's ICO has updated guidance on storage and access technologies; and China's PIPL stresses minimum necessity, clear purpose, and transparency. Any business model that builds AI advertising on "gray-area data acquisition" or "unverified improvement claims" risks being rejected by both regulators and customers.

Value-Chain Landscape and Profit Pools

The table below outlines the main links of AI advertising and marketing automation in the order of "budget → data → creative → delivery → measurement → conversion → customer operations." The scores are the author's subjective ratings based on commercialization maturity, data/channel barriers, platform bargaining power, and valuation elasticity, out of 10.

The underlying evidence for this table comes mainly from: advertising platform product documentation, the latest earnings reports/calls of major companies, public strategy documents from agency groups and data platforms, and 2025–2026 privacy and advertising regulatory updates. On the platform side, the strongest evidence comes from Google, Meta, Amazon, AppLovin, and The Trade Desk; on the software side, from Adobe, Salesforce, HubSpot, Klaviyo, and Braze; and on the agency/data side, from public materials from Publicis, WPP, Dentsu, LiveRamp, and others.

Profit-pool attribution:

The platform layer's advantage comes from three things: distribution rights, closed-loop data, and default workflows. Google couples AI Max directly with Search, saying it has unlocked billions of "previously uncovered" incremental searches; Meta's Advantage+ continues to embed in its core advertising system, and industry media cite management as saying Meta's AI ad infrastructure now runs at an annualized revenue rate above $60 billion; Amazon uses AI not only in DSP but also connects it to the shelf and the shopping assistant Rufus. Platforms have not treated AI as a separately sold "add-on," but use it to improve ad-auction efficiency, conversion rates, and budget-capture capability.

So the first profit pool of AI advertising will most likely continue to sit with budget-gateway platforms such as Google, Meta, Amazon, TikTok, and AppLovin; the second profit pool sits with platforms such as Adobe, Salesforce, HubSpot, Klaviyo, and Publicis/Epsilon that own the content supply chain, customer data, and cross-team workflows; and only the third profit pool may belong to a few AI-native tools, but only on the condition that they own unique data, a specific channel, or an irreplaceable approval/governance scenario, rather than merely being able to "generate an image / a video / an email."

Business Models and Scenario Assessment

Charging methods and pros/cons:

• Charging via ad-revenue take / media rates: best suited to platforms and DSPs; the upside is that it scales naturally with budgets, while the downside is strong cyclicality and heavy exposure to advertiser-sector rotation. Google, Meta, Amazon, AppLovin, and TTD fall into this category.

• Charging via SaaS subscription / seats: best suited to CRM, marketing automation, content supply chain, and data platforms; the upside is predictable revenue and high gross margin, while the downside is the need to prove AI can drive expansion, price increases, or retention improvement. Salesforce, HubSpot, Klaviyo, Braze, and Adobe fall into this category.

• Charging by generation volume/API calls: suited to creative tools with higher inference costs such as video, image, and voice, but the most easily squeezed by platforms' built-in capabilities and falling model prices. Adobe has publicly disclosed its credit-based billing logic for Firefly/video and the like, while Amazon chooses to bundle some generation tools for free.

• Charging via ROI / leads / conversion share: most attractive to advertisers, but it usually requires extremely strong attribution and control, so it suits platforms, agency-managed services, and a few highly closed-loop tools rather than most standalone AI creative companies. Amazon, AppLovin, and TTD are closer to this logic; the problem for many AI SDR companies is that "can reach at scale" does not equal "can convert reliably."

Core assessments:

• AI delivery optimization can directly lift ROAS and expand budgets. This is the segment with the strongest current evidence. Google, TTD, Snap, Amazon, and AppLovin have all provided clear effectiveness lifts or automation-penetration data.

• AI creative generation brings real productivity gains, but whether it becomes a standalone revenue pool depends on having brand governance and approval workflows. Adobe's Lumen case shows content output can rise 35x with 65% faster speed; but on the platform side, more and more creative functions are being bundled into the advertising products themselves.

• AI marketing Agents can raise the ARPU of CRM/marketing automation. The public commentary of Salesforce, HubSpot, and Klaviyo has all linked Agents to customer expansion, paid deals, or multi-product adoption, showing this is no longer simply an "internal Copilot."

• AI SDRs will disrupt traditional sales outsourcing and human SDR teams, but the industry has yet to complete ROI proof. The strongest current business model is not "auto-send lots of emails" but connecting AI to CRM, intent data, data enrichment, and RevOps workflows; otherwise it easily runs into deliverability, spam-folder, compliance, and duplicate-outreach problems. The FTC's enforcement against exaggerated AI business outcomes also reinforces this risk.

• Retail media AI is a new high-growth advertising channel. GroupM expects retail media to grow 15% in 2025 and reach 42.4% of digital advertising excluding China; disclosures from Amazon and Alibaba show that AI can directly drive shelf advertising and retail-media performance budgets.

• Data collaboration / clean rooms are "necessary infrastructure" for the privacy era, but not necessarily the largest standalone profit pool. They are most valuable when sold bundled with activation, identity resolution, retail-media measurement, CDP, and agency services, rather than in isolation. Publicis acquired Lotame precisely to feed identity and data assets into CoreAI and Epsilon.

Three-scenario assessment:

The above scenarios rest on several realities already validated by public information: the global advertising market keeps growing with a rising digital share; marketers have begun using generative AI at scale; platform-level automation metrics at Google, Meta, Amazon, TTD, and Snap are improving; and enterprise marketing software companies have begun binding Agents to incremental ARR, customer expansion, and workflows.

Value Sizing and Segment Breakdown

Large advertisers' AI budgets are spent first in four places: First is media-delivery automation, because it directly determines budget efficiency and incremental acquisition; second is the content supply chain, because brands need faster refresh, localization, and testing across more channels at lower cost; third is customer data and segmentation, because after privacy tightening there is less usable data but data quality matters more; and fourth is attribution and incrementality measurement, because the CFO will not pay long-term for "generation that looks faster," only for "higher conversion / lower CAC / higher LTV." The public commentary of Google, Amazon, TTD, Adobe, Salesforce, HubSpot, and Klaviyo each maps to one of these four budget buckets.

Agency companies' and marketing service providers' AI budgets go more toward replacing headcount and protecting bargaining power. Publicis binds AI deeply with Epsilon, Lotame, CoreAI, retail media, and its agency system rather than building only an internal assistant; WPP explicitly positions WPP Open as an agentic marketing platform, but its latest quarterly revenue remains under pressure, showing that "having an AI platform" does not equal "near-term financial payoff"; Dentsu emphasizes using AI to unlock platform algorithms, creative, and investment optimization, which looks more like systems engineering amid transformation. The conclusion: the success or failure of agency companies' AI hinges on whether they turn AI into data and process assets customers will pay for, not into an internal cost-cutting tool.

The cost items most easily reduced by AI are asset cropping, short-video derivatives, first-draft copy, ad variants, basic delivery operations, report compilation, lead routing, basic customer service, and part of SDR work. The items most likely to create incremental revenue are: automated budget expansion, closed-loop retail-media advertising, customer-operations upsell, suboptimal-audience expansion, cross-region scaling of the content supply chain, and scenarios that charge advertisers directly, such as AI-native marketing services. Baidu has already provided AI-native marketing services revenue; Salesforce and HubSpot have provided Agent-related deals and customer expansion; Meta/Google/Amazon/TTD have provided evidence of delivery-efficiency improvement.

The segments most worth continuing to track, ranked by investment attractiveness, draw the following judgments:

The high-confidence evidence behind this segment ranking is clear: platform delivery automation, retail media, and CRM/marketing Agents already show up as advertising revenue, ARR, customer expansion, or clear effectiveness lifts in the earnings or official product commentary of Google/Meta/Amazon/AppLovin/TTD and Salesforce/HubSpot/Klaviyo/Braze respectively; whereas the main evidence for pure creative-generation tools still leans toward "faster output," not "bigger budgets / fatter margins."

Investment Targets Master Table and Scoring

The table below sorts the key companies into five categories: Category A core direct beneficiaries, Category B clear beneficiaries with valuation or competitive risk, Category C more efficiency tools, Category D narrative stronger than evidence, and Category E potential casualties. The "valuation judgment" references mainly the latest market cap/PE and public revenue figures; where complete EV/EBITDA, PS, or FCF data is lacking, it states "needs further verification" explicitly.

My priority ranking:

• Category A: Meta, Alphabet, Amazon, AppLovin, The Trade Desk, Adobe, Salesforce, HubSpot, Publicis.

• Category B: Klaviyo, Braze, Pinterest, Snap, Reddit, Alibaba, Baidu.

• Category C: LiveRamp, Tencent, Dentsu, WPP, DoubleVerify, IAS.

• Category D: Most single-point AI creative tools lacking audit-grade revenue/ARR/budget-migration evidence, plus some A-share marketing service providers with insufficient financial disclosure.

• Category E: Traditional DMPs, low-value-add creative execution, pure manual delivery optimization, and some SSP/sell-side technology platforms.

In-Depth Analysis of Key Listed Companies

Below we screen the 15 listed companies most worth continued in-depth research. The format is compressed, but each company answers: segment, product, commercialization stage, direct revenue exposure, margin impact, moat, valuation, and follow-up tracking metrics.

The 10 listed companies most worth deeper model and channel verification: Meta, Alphabet, Amazon, AppLovin, The Trade Desk, Adobe, Salesforce, HubSpot, Publicis, Alibaba.

Private-Market Opportunities and Industry Restructuring

Of the ten private/primary-market directions and representative companies most worth tracking, I would prioritize: TikTok Ads/ByteDance, Moloco, Hightouch, Canva, Writer, Jasper, Typeface, Clay, Apollo.io, and Attentive. The reason is not that these companies have all proven enormous revenue, but that they respectively hold key positions in social traffic gateways, mobile-acquisition optimization, reverse ETL and data activation, lightweight creative workflows, brand-copy governance, AI SDR/sales automation, and SMS/private-domain customer operations. For these companies, what truly merits scrutiny is not the demo but four metrics: paid customer count, net retention, depth of integration with the system of record, and whether customers are willing to migrate existing budgets away from agencies/headcount/legacy SaaS. Many private companies still disclose insufficiently on funding, ARR, gross margin, and retention, and need further verification.

The main thread of industry restructuring is already clear:

• Advertising agencies will be forced to shift from "selling labor-days" to "selling data, models, and workflows." Publicis is already doing this; WPP is doing this; Dentsu still has execution in the Japanese market but is still transitioning globally. Agencies without data assets, identity resolution, or a unified media-and-creative workbench will see their bargaining power decline.

• The structure of ad-optimizer, asset-production, marketing-operations, and customer-service-operations roles will change. Basic bidding, frequency control, asset remaking, localization, report writing, and FAQ customer service will be clearly squeezed by automation; but the value of roles in experiment design, brand guidelines, growth strategy, data governance, attribution, and agent orchestration will rise. The product roadmaps of Adobe, Salesforce, HubSpot, and Klaviyo all reinforce the model of "humans own the guardrails, Agents own execution."

• AI creative will increasingly be made free by platforms such as Google/Meta/Amazon. This is the core business risk for standalone AI creative tools. Platforms have cheaper acquisition channels, stronger model-feedback loops, and more natural embedding points. Standalone tools can sustain pricing only by inserting themselves into brand governance and the content supply chain.

• Automated delivery will disrupt standalone DSPs and media agencies, but will not eliminate them entirely. On the contrary, the DSPs and agencies that truly survive will emphasize first-party data activation, open-internet curation, CTV, retail-media measurement, and customized customer strategy, rather than relying on pure manual tuning. TTD is the strongest "upgrade-type winner" in the open internet; many sell-side technology platforms lean more toward being "re-pricing targets."

• AI data collaboration and clean rooms will become privacy-era infrastructure, but leading companies will bundle them into larger platforms. So "standalone clean rooms" may not be the optimal investment category; better are platforms that bind clean rooms with activation, identity, retail media, and marketing SaaS. Publicis/Epsilon/Lotame, Adobe, and Salesforce/Informatica are closer to this model.

Risks, Open Questions, and Final Conclusions

Systemic risks:

• Platform-commoditization risk: Amazon has already set some creative-generation tools at no additional cost; Google, Meta, Pinterest, and Snap all build automation and generative capabilities into their platforms. Standalone creative/delivery tools without governance and workflows will find long-term monetization difficult.

• Privacy and compliance risk: Google maintains its current third-party cookie management approach, but the industry has not returned to the old world; CCPA/CPRA, DMA, DSA, PIPL, and ATT continue to raise data-acquisition costs and strengthen requirements for transparency, choice, and minimum necessity.

• ROI-disproof risk: the FTC has acted against companies exaggerating AI capabilities and business results. If marketing AI cannot prove incremental impact, it will retreat from a "growth budget" back to an "IT cost."

• Brand-safety and false-claims risk: AI-generated assets may bring false selling points, brand inconsistency, and copyright/likeness/voice-rights issues, so enterprise procurement will lean more toward governance-focused vendors such as Adobe, Salesforce, and Publicis.

• Budget-cycle risk: advertising budgets are inherently cyclical. Even if AI tools work, when the macro is soft the first to feel pressure are still small advertisers and high-elasticity marketing budgets. Platform-type companies can defend share through automation, while point tools are more easily cut.

Open questions and limitations:

The most fully documented today are the U.S. platforms, marketing SaaS, and large agency groups; China A-share/HK-listed marketing service companies and many private AI creative/AI SDR companies generally suffer from insufficient revenue breakouts, undisclosed ARR, customer cases skewed toward marketing materials, and unclear boundaries between pilot and scale. Therefore, for these companies this report emphasizes a "validation framework" rather than conclusive judgments. For many private companies, I prefer to place them on a "watch list" first rather than assign a high-certainty rating up front.

Final conclusions:

AI advertising and marketing automation is one of the few application layers in the AI value chain that can already map directly to advertising revenue, budget migration, subscription expansion, margin improvement, customer retention, and data barriers. But what truly holds long-term investment value is not "any company that generates content," but the following four categories:

• AI advertising platform companies: Google, Meta, Amazon, AppLovin, The Trade Desk.

• AI marketing platform software companies: Salesforce, HubSpot, Adobe, Klaviyo, Braze.

• Data/identity/retail-media infrastructure companies: Publicis/Epsilon, LiveRamp, Alibaba, Baidu.

• Service/workflow platforms that can turn AI from internal cost-cutting into paid customer products: Publicis, Adobe, and parts of the Salesforce/HubSpot ecosystem partners.

If we narrow the scope further, I believe the most worthwhile next direction for deeper work is not a generic "AI marketing," but:

"AI content supply chain and brand governance." The reason is simple: this is the battleground that best distinguishes platform-commoditization risk from standalone pricing power. Whoever can upgrade AI from "an asset in a few minutes" to a "cross-region, multi-channel, approvable, compliant, reusable, attributable" enterprise workflow is more likely to capture truly sustainable subscription revenue and a high-quality profit pool. Adobe has provided the clearest product form, Publicis is converging here on the agency side, and Salesforce/HubSpot/Klaviyo on the customer-data side.

This report is based on public information and does not constitute investment advice. Markets carry risk; invest with caution.