Ninety percent of CMOs are experimenting with AI, yet fewer than ten percent have captured value from it across end-to-end workflows.1 While pilot programs are common, widespread profitability remains rare.
This gap is not a technology problem. AI models are increasingly capable and cost-effective. The problem persists because enterprises attach AI to older operating models built for long cycles, annual plans, static segments, and slow budget reviews.
The current transition goes beyond generative AI. Customer expectations, media fragmentation, lower content costs, expanded first-party data, and new forms of discovery are coming together. Combined, these factors shift marketing from a series of isolated projects to a continuously learning system.
Across enterprise marketing teams, a clear pattern emerges around five connected operational shifts. Each shift makes the others more powerful.
Shift 01From campaign thinking to continuous engagement
A typical campaign calendar assumes organizations decide when conversations begin. Customers, however, research, compare, and ask for help on their own schedules. Google identifies this as the “messy middle”: an exploration and evaluation loop running until a decision is made.2
Expectations have moved faster than operations. Currently, 83% of customers expect to interact with someone immediately when they contact a company, and 85% expect consistent interactions across departments.3 Fewer than 10% of organizations have updated their marketing structures end to end to deliver this.1
Continuous engagement replaces the campaign as the primary unit of work. An always-on decision layer monitors meaningful changes, such as product views or competitor launches, and determines the best response, audience, and timing. The goal is not more frequent contact, but removing irrelevant decisions.
- Real-time signals
- Customer data
- Performance data
- Product data
- Market data
- WhatThe valid message, offer, or service response.
- WhenThe moment when expected value exceeds contact cost.
- WhereThe permitted channel with the strongest response probability.
From static segments to signal-driven segments of one
Traditional segmentation compresses people into averages, which helps with planning but fails in real-time decisions. Two individuals with identical demographics can be at entirely different stages of the buying journey.
A segment of one is a specific decision context combining consented history, live behavior, predicted intent, product details, and channel timing at the exact moment of decision. Consumer expectations support the ambition, up to a clearly drawn line. Although 71% of customers expect personalization and 76% express frustration when it is missing,4 only 39% are comfortable with brands using AI to predict emotional states.5
The financial case is clear. Companies that grow faster drive 40 percent more of their revenue from personalization than their slower-growing peers.4 However, 98% of marketing teams already using AI run into at least one data-related barrier, while 78% of marketers say they need more personalized content than they are able to produce.6 This dependency leads directly to the third shift.
Shift 03From asset production to an infinite test lab
Creative production used to be the scarcest resource in marketing, leading to long agency cycles and rigid approval chains that slowed down testing.
This limitation is gone. Modular content systems and generative models now produce hundreds of approved variants rapidly, deploy them in parallel, and iterate based on live performance data. BCG estimates that CMOs who embrace generative AI across every pillar of marketing can realize a three to six times higher return on total marketing spend.7
Consumer acceptance is moving quickly. Over the past year, 57% of online users aged 14 to 44 watched a virtual creator, a sample of just 300 virtual creators on YouTube earned more than 15 billion views in 2024, and major brands are actively testing virtual talent.8 The main question is no longer whether AI-generated creative performs, but whether an organization’s operating model can learn faster than its competitors.
Volume without careful experiment design creates little value. Leading teams treat every creative asset as a hypothesis, control variables strictly, and carefully store the resulting lessons.
Atomic content
Approved claims, proof points, imagery, formats, and calls to action.
Dynamic content
Audience-specific combinations in which each tested variable remains identifiable.
From performance focus to full funnel
Performance marketing dominated recent years because it was easy to measure, but it now faces three structural challenges.
First, search has lost its intent monopoly. 46% of Gen Z already prefer social platforms over search engines for discovery.9 Buying journeys start in feeds and communities where brand recognition matters more than bidding strategies.
Second, content volume is expanding rapidly. As generative tools spread through every marketing team, the cost of producing content collapses and feeds fill with machine-made variants. In a world of infinite content, distinctiveness becomes the primary scarce asset.
Third, purchasing behavior is shifting to agents. Among consumers who already search with AI, 44% call it their primary and preferred source for discovery,10 and 34% of US consumers say they are comfortable letting AI handle even larger purchases for them.11 As software begins shopping for customers, brand equity changes from a soft metric to a vital trust anchor and input signal for algorithmic selection. As a result, full-funnel marketing is essential, recognizing that brand equity creates the demand that performance marketing captures.
Shift 05From siloed tools to hybrid teams
Many enterprises deploy AI as isolated point solutions for drafting copy or summarizing reports. While productivity increases, the core operating model stays the same.
Advanced models integrate AI agents as actual team members with defined roles, permissions, and shared context. Specialized agents can monitor signals, generate approved variants, check regulatory compliance, and suggest new tests. Human personnel set goals, constraints, and guardrails, keeping ownership of the critical decisions.
The proven value is significant: McKinsey estimates agentic AI will come to power as much as two-thirds of current marketing activities.1 Among organizations that use or plan to use AI agents, 82% expect major or moderate ROI improvements, yet only 13% of marketers work with agentic AI today.6 Moving from basic assistance to autonomous operation is a fundamental operating-model decision rather than a simple software purchase.
AI drafts options. People select, edit, and approve every output.
Agents execute bounded tasks. People approve the completed work.
Agents manage approved workflows. People review exceptions and material decisions.
Agents optimize within policy. People set objectives, limits, and brand guardrails.
The five shifts are one shift
These five shifts depend on each other. Continuous engagement requires highly specific segmentation, which then demands infinite creative capacity. This resulting flood of content makes brand distinctiveness more important. None of this works without a hybrid team structure combining human oversight with agent execution.
Treating these shifts as isolated projects leads to pilot fatigue. Enterprises that combine them into a unified system report 4 to 7% growth in revenue and conversion value, double to triple productivity gains, and 10 to 30% savings in production and execution costs.12
Implementation requires starting with a single brand, product, or measurable journey. Key design principles include:
Combining behavioral and commercial data into a single decision context.
Specifying when systems should act, wait, or escalate.
Ensuring variations tie back to specific strategic questions.
Carefully storing lessons to inform future cycles.
The ultimate goal is not complete automation, but removing the wait time between signals, decisions, and outcomes, focusing human judgment where it delivers the most strategic value.
Footnotes and sources
- McKinsey & Company (2026), “Reinventing marketing workflows with agentic AI”.↩
- Google, Decoding Decisions: Marketing in the Messy Middle, replicated across more than 15 global markets.↩
- Salesforce, “State of the Connected Customer”, fifth edition.↩
- McKinsey & Company (2021), “Next in Personalization: The value of getting personalization right or wrong is multiplying”.↩
- Adobe (2026), “AI and Digital Trends”, consumer research.↩
- Salesforce (2026), “State of Marketing, 10th Edition”, based on a survey of about 4,500 marketers.↩
- BCG (2025), “Executive Perspectives: The Future of Marketing with GenAI”.↩
- YouTube Culture & Trends (2025), virtual creators research.↩
- eMarketer (2025), social search research, via Search Engine Land.↩
- McKinsey & Company (2025), “New front door to the internet: Winning in the age of AI search”, AI Discovery Survey, August 2025.↩
- Quad and The Harris Poll (2026), AI shopping survey of 2,180 US adults, February 2026.↩
- McKinsey & Company (2026), “From campaigns to continuous growth: AI capabilities shaping marketing”.↩
