Salesforce's State of Marketing: AI dominates the plan, but data still slows the follow-through
Salesforce's 10th State of Marketing is blunt: marketers are racing to adopt AI while still tripping over integration, data fragmentation, and skills gaps. The study is based on a double-blind survey of 4,450 marketers across 26 countries and confirms a dual mandate-innovate with AI and deliver better customer experiences. Those two goals are connected. AI only pays off when it's wired into your data and actual workflows.
For a deeper read, see the Salesforce State of Marketing (10th edition).
What marketers are prioritizing now
The top two priorities are clear: implement and operationalize AI, then adapt to AI's impact on marketing. Efficiency is the first target-automating tasks, speeding content, and improving decisions.
- 76% of respondents use at least one form of AI (predictive, generative, or agentic).
- Only 13% use agentic AI (autonomous agents that execute marketing tasks with minimal human input).
- 82% who use or plan to use agents expect moderate or major ROI gains.
- High performers reclaimed 8 hours per week with AI agents and improved ROI.
Reported outcomes when AI is deployed at scale:
- +20% ROI
- +20% customer satisfaction
- +19% conversion rates
- -19% costs
Adoption is high, integration is not
61% say AI adoption is strong, yet full integration is still "in progress." The blockers are predictable: privacy and security concerns, accuracy questions, lack of in-house expertise, immature tooling, and fuzzy use cases.
Translation: many teams bought AI, fewer teams changed how work gets done. Without clear operating models and data access, AI stalls in pilot mode.
Re-thinking customer engagement
Marketer reality check:
- 64% struggle to keep up with changing customer behavior.
- 69% say new customer acquisition is getting harder.
- 51% say campaigns still feel generic; 37% report inconsistent messaging.
The top AI use cases are practical and customer-facing: content personalization, predicting campaign performance/ROI, generating visuals and copy, and predicting customer behavior. The lesson: stop thinking about "AI tools" and start defining AI-driven use cases that fix specific breaks in the journey.
SEO and AEO are being rewritten
85% are reshaping SEO, and 88% are optimizing for AI-driven search experiences (e.g., ChatGPT responses and Google's AI Overviews). This makes sense, but the AEO playbook is still fluid. Focus on durable moves: clear information architecture, structured data, expert content, and fast feedback loops from query to conversion.
Data and personalization: the sticking point
The average marketing team is wrangling seven data sources to support agentic marketing. Only a little over half have access to the Service, Sales, and Commerce data they actually need. Despite that, 71% report they're satisfied with connectivity-largely because teams already running agents did the plumbing work.
Personalization exposes the cracks. 46% lack preference data; 98% of AI-using teams report at least one data barrier (silos, too much data, or poor quality). If data is "connected" yet personalization still underperforms, the issue is usually coverage (missing key signals), quality (stale or inconsistent), or activation (no clean path from data to decision to content).
A practical data checklist for AI
- Inventory sources by use case, not system. Map which fields power each decision (e.g., offer, timing, channel).
- Close the "service-sales-commerce" gap first-those signals lift relevance fast.
- Capture declared preferences (zero-party data) and unify with behavioral data.
- Define data contracts: freshness, quality rules, and access controls per use case.
- Operationalize feedback loops so models learn from outcomes (opens, clicks, purchases, churn, CSAT).
What CMOs should do next
The CMO scope now spans brand, product marketing, analytics, data strategy, and revenue operations. Expectations have risen on both data connectivity and qualified pipeline. That needs an AI strategy grounded in measurable business outcomes, not tool experiments.
- Set 3-5 AI use cases tied to revenue or cost (e.g., next-best-offer, lead scoring, creative versioning, win-back).
- Fund the data foundation those use cases require-before you scale agents.
- Ship in 90-day cycles: pilot, measure, standardize, then automate.
- Keep humans in the loop for brand, compliance, and edge cases.
- Report on ROI, CX impact, and time saved to earn more runway.
For structured upskilling and org design support, see the AI Learning Path for CMOs.
Skills to build in 2026
- Data analysis and interpretation (from dashboards to causal insight)
- AI tool management (prompting, evaluation, agent orchestration)
- Strategic and creative thinking (test design, concept development)
- Content strategy and curation (brand-safe, channel-specific)
- Privacy and compliance management (policy, consent, auditability)
If you're responsible for deployment and day-to-day execution, the AI Learning Path for Marketing Managers will shorten the learning curve.
Quick-start playbook
- Pick one journey break to fix (e.g., cart abandonment, lead handoff).
- List the minimum data needed; fill the gaps with first-party signals.
- Deploy a narrow AI use case (prediction or generation) with guardrails.
- Measure uplift against a clean control. If it works, automate and expand.
- Document the workflow so it survives team churn and tool changes.
Bottom line: AI can carry the operational load so your team can get back to creative and strategic work. But it only works if your data is trustworthy and your use cases are specific. Build that backbone, then scale.
Your membership also unlocks:
