How can artificial intelligence be used in marketing?

The Real Problem: Adoption Is Easy, Scaling Is Hard

The gap between using AI and scaling it is widening, creating a critical bottleneck for marketing teams. While adoption rates soar, the jump from pilot to production remains elusive for most organizations.

Approximately 75% of marketers have implemented AI tools, with 85% reporting faster campaign execution according to enterprise data. Yet, this widespread usage masks a deeper operational reality. Only 31% of studied AI use cases reach full production, and just one in four initiatives hits expected growth ROI as reported by ISG.

This disconnect stems from treating AI as a standalone tool rather than an integrated workflow component. Organizations often buy solutions without establishing the necessary governance or process structures to support them. Consequently, only 1% of C-suite leaders call their generative AI initiatives mature per adoption statistics.

The core challenge is no longer tool acquisition; it is workflow integration. To bridge the gap between experimentation and scalable results, marketing teams must focus on:

• Codifying lessons from rapid experiments into repeatable, compliant processes
• Vetting third-party AI components for security and quality consistency
• Aligning individual tool usage with broader enterprise governance frameworks
• Measuring ROI through workflow efficiency rather than isolated output speed

Without this structural shift, productivity gains remain fragmented and difficult to scale across departments. The data suggests that speed alone does not equate to strategic advantage. Teams that harden their AI workflows into robust systems will outperform those relying solely on ad-hoc tool usage.

Workflow integration requires verifying the reliability of every AI component plugged into the stack. Published research indicates that ~26% of agent skills carry at least one vulnerability, highlighting the risk of unchecked automation. SkillzBench is building a free directory and MCP tool to address this trust gap. It will allow users to discover, verify, and benchmark AI agent skills before deployment, ensuring that scaling efforts are built on secure, high-quality foundations.

By focusing on the quality of integration rather than the quantity of tools, marketers can transform AI from a novelty into a durable competitive asset. This approach mitigates the risks of "boiling the ocean" while avoiding siloed, temporary fixes. The goal is to create scalable, compliant processes that deliver consistent value.

As the market matures, the ability to manage complexity will define success. Teams that prioritize robust workflow integration over rapid tool adoption will be better positioned to capture long-term ROI. The next phase of AI in marketing is not about doing more; it is about doing it reliably at scale.

Where AI Actually Works in Marketing Today

AI in marketing has crossed the line from experiment to production, and the numbers show exactly where the value is landing. Marketing platforms captured $660 million of the $7.3 billion departmental AI market in 2025, driven almost entirely by content generation and campaign optimization, according to Menlo Ventures' enterprise research.

Content creation leads the pack. Roughly 75% of marketers have implemented AI tools, and 62% cite content creation as their top use case. Among content creators specifically, adoption is nearly universal — 92% have used generative AI, and 52% say it makes them more creative while saving production time.

Campaign optimization is the second proven workhorse. OpenAI's State of Enterprise AI report found that 85% of marketing and product users report faster campaign execution, with users attributing 40–60 minutes of savings per active day (OpenAI, 2025). Note this is self-reported vendor data, but it aligns with what enterprises are actually paying for.

The outcomes are concrete, not hypothetical:

• Nestlé cut product ideation cycles from six months to six weeks using generative AI
• Welspun Living achieved an up-to-10% top-line revenue lift from AI-driven marketing
• Alibaba saw a 40% increase in purchase intent from genAI-powered search experiences

Beyond content and campaigns, two analytical use cases round out the proven list. In retail and CPG, ROI-focused adopters rank marketing content generation first (23%), followed by customer analysis and segmentation (19%) and hyper-personalized recommendations (18%). These analytical applications turn customer data into targeting precision that manual teams simply cannot match at scale.

The common thread across all four use cases is workflow integration — AI works when it is embedded into repeatable processes, not used for one-off queries. OpenAI's Chief Economist describes the shift as moving "from asking models for outputs to delegating complex, multi-step workflows." That is where agent skills come in: instruction packages that extend coding agents and chat assistants can automate content drafting, campaign reporting, and segmentation workflows end to end. SkillzBench is building a free directory of these skills — each security-scanned, version-pinned, and benchmarked — so marketing teams can extend these proven use cases without inheriting hidden risk.

The Shift You Can't Afford to Miss: SEO to GEO

For years, SEO was the front door of the internet. That door has moved. According to McKinsey's analysis of AI-powered search, half of consumers already use tools like ChatGPT, Gemini, and Perplexity to research purchases, and $750 billion in US revenue is projected to flow through AI search by 2028.

The behavior shift is deeper than channel migration. Among AI search users, 44% call it their primary insight source — ahead of traditional search (31%), brand websites (9%), and review sites (6%). Most ask top-of-funnel questions, meaning brands are being evaluated before they ever appear on a results page.

Here's the uncomfortable part: traditional brand strength doesn't carry over. Brand-owned sites make up just 5–10% of the sources AI search references — in categories like CPG and financial services, more than 65% of cited sources are publishers, user-generated content, and affiliate sites. Your visibility now depends on what the rest of the web says about you.

And almost nobody is watching. Only 16% of brands systematically track their AI search performance, which McKinsey bluntly frames as: if you're wondering when AI will change search behavior, you're too late.

McKinsey recommends four moves for marketers:

• Run a diagnostic assessing your visibility and sentiment across AI platforms.
• Rebalance content investment across owned, third-party, and community sources.
• Optimize content for AI engines — credibility, structure, clear headings, precise language.
• Build GEO as a core capability with its own KPIs and cross-functional ownership.

That last move is where most teams stall. GEO monitoring, LLM-optimized content production, and KPI tracking are repetitive, multi-step workflows — exactly the kind of work agents can handle when equipped with the right skills. Since roughly a quarter of published agent skills carry at least one vulnerability, vetting matters; SkillzBench is building a free directory that security-scans and benchmarks every skill, plus a free MCP connector to equip your agents with only the ones you choose.

The window is open now. With 20–50% of traditional search traffic at risk for unprepared brands, GEO isn't a future problem — it's this quarter's.

Build the Workflow: Experiment, Codify, Harden

Most marketing teams don't fail with AI because they lack tools — they fail because they never turn scattered experiments into repeatable processes. ISG's adoption research identifies two failure modes: "boiling the ocean" (waiting on a multi-year data transformation before starting) and "bypassing the mess" (building siloed pipelines for immediate needs). The winning path, ISG argues, is to experiment rapidly, codify lessons, and harden them into scalable, compliant processes.

The pressure to get this right is real. Only 31% of AI use cases reached full production in 2025, and ROI expectations are broadly underdelivering — just one in four initiatives hits expected growth ROI, and only half achieve expected efficiency gains. Meanwhile, according to Menlo Ventures, 76% of AI use cases are now purchased rather than built internally, up from 53% in 2024. Your marketing AI stack is increasingly assembled, not engineered — which makes disciplined workflow integration matter more than tool count.

Adoption is also arriving bottom-up. Product-led growth drives AI adoption at roughly four times traditional software rates, and "shadow AI" — employees using personal subscriptions — may account for close to 40% of application AI spend. Your marketers are already building custom GPTs and agent workflows; 75% of workers now complete tasks outside their original role boundaries, per OpenAI's enterprise report. The question is whether that individual initiative becomes governed capability or unmanaged risk.

A practical experiment-to-production loop for marketing teams looks like this:

• Pick one high-value workflow — content generation (cited by 62% of marketers) or campaign optimization — and run rapid two-week experiments rather than pilots.
• Codify what worked into written prompts, templates, and evaluation criteria before scaling to the wider team.
• Harden the winners: version-pin every component, security-scan anything third-party, and attach clear KPIs so production workflows stay measurable.
• Bring shadow AI into the light by giving teams sanctioned, better-equipped alternatives to their personal tools.

That last step is where vetting matters most. Published research has found that roughly 26% of agent skills carry at least one vulnerability, and only about 20% of organizations report mature governance for AI agents. SkillzBench is building a free answer to this: a directory where every skill is version-pinned to a commit SHA, security-scanned, and benchmarked, plus a free MCP connector so any agent loads only the skills you choose. Skills, proven — that's the standard hardened workflows deserve.

Vet What You Plug In: Closing the Trust and Governance Gap

The fastest-moving corner of marketing AI right now is agentic — autonomous assistants that draft campaigns, optimize content, and run multi-step workflows. But the guardrails haven't kept pace with the enthusiasm, and that gap deserves your attention before you plug anything into your stack.

The numbers tell the story. According to adoption research, 61% of CEOs say their organization is actively adopting AI agents, and nearly three in four companies plan agentic deployments within two years — up from 23% today. Yet only about 20% of organizations report mature governance frameworks for those agents. Meanwhile, ISG's adoption research shows most enterprises are hardening AI into production processes without the governance layer to match.

The risk is concrete, not theoretical. Published research found that roughly 26% of agent skills — the instruction packages that teach agents like Claude Code, Codex, and Cursor how to perform tasks — carry at least one vulnerability. And since 76% of AI use cases are purchased rather than built, per Menlo Ventures' 2025 enterprise analysis, most marketing teams are assembling workflows from components they didn't write and can't fully inspect.

That's why vetting what you plug in matters as much as choosing what to automate. A disciplined approach to equipping marketing agents looks like this:

• Pin every skill to a specific version (a commit SHA), so you know exactly what's running in production
• Run security scans and license checks before a skill touches customer or campaign data
• Benchmark skills in a sandbox — same model, with and without the skill — to see whether they actually improve output
• Give each agent only the skills and tools it needs, rather than loading everything by default

This is the gap SkillzBench is being built to close. Its Verify pillar will pin every skill version, scan for vulnerabilities, and check licenses, while its Benchmark pillar will test each skill in a controlled sandbox and publish the measured Lift in performance. Every skill will carry a 0–100 trust score built from security, quality, license, maintenance, and community signals — and money never touches scores, because everything on the platform, including the MCP connector and per-agent loadouts, will be free.

For marketers, the payoff is simple: you can adopt agentic workflows at the speed the market demands without inheriting someone else's security debt. Vet deliberately, equip selectively, and let measured results — not hype — decide what earns a place in your stack.