More than 90% of marketers say they use AI, yet most still cannot get AI out of their LLM chat windows and into their actual workflows. That is the central finding of MarTech's Q4 2026 assessment of AI in marketing, which draws on industry reports and the author's on- and off-the-record conversations with marketers and martech leaders, many of them featured on the Conversations with MarTech podcast.

The reports MarTech reviews daily largely agree on the basics: nearly universal adoption, content creation as the most common use case, and growing use in app building, personalization and workflow enhancement. They diverge, however, on the ROI marketers are actually seeing and on AI's impact on workforce and hiring plans.

Why can't marketers cross the workflow chasm?

Almost everyone now has an LLM of choice, or even two, according to MarTech. Adoption of basic AI tools for basic tasks "isn't a problem and hasn't been for quite some time." The problem is integration.

• Connecting LLMs to other apps is technically challenging and requires jumping through hoops.
• IT permissions and available technical assistance vary by company.
• Teammates may lack the fluency to use what has been built.

The result: "so many people are (literally) cutting and pasting from LLMs," MarTech notes. "Cutting and pasting is easy. Integrations are hard. But cutting and pasting is not a workflow."

Is AI in every app creating smarter silos?

Nearly every tool marketing teams use now has AI embedded in it, from elegant assistants to "plain-old bolted-on AI that looks awkward at best." The same applies to productivity tools like Excel, Google Sheets and PowerPoint. But AI helps within each app — not across the next one.

Instead of using AI to connect the CRM, the marketing automation platform and productivity tools, teams are building AI-powered silos. MarTech warns vendors face real risk here: embedding AI costs real money, and marketers have plenty of alternatives to choose from. Agents may be part of the answer, but almost every vendor now offers those too.

Is bad data undermining trust in AI output?

Many marketers lack confidence in their data quality and know AI will amplify bad data. Poor data delivers wrong information to customers and teammates.

MarTech points to Salesforce as the example worth heeding. Despite the criticism its ecosystem receives, with Data 360 sitting beneath Agentforce and the CRM, "Salesforce is right," MarTech argues. Companies may feel locked into expensive contracts, but data management and governance are hugely important — and Salesforce, having bet the company on AI, cannot let bad data sink its AI initiatives. The rule remains: garbage in, garbage out.

Is AI just feeding marketers' fixation on more?

More content, more designs, more drafts, more emails. Several marketing leaders MarTech spoke with believe marketers are overfocused on volume. AI creates more easily but "doesn't think about resonance and the power of what's being created."

The checking burden compounds the problem. Many marketers report that reviewing AI's prolific output creates more work and fails to deliver the promised efficiency. Prompts help with ideation and an initial draft, but there is more work to be done. As MarTech puts it: "AI-generated work isn't necessarily efficient work."

What do customers actually want?

Customers do not necessarily want more content, and they do not always find AI-assisted chat and help functions helpful. AI is great when people want a quick answer; less so when they need empathy and human understanding. The question marketing teams must ask: are you creating more because customers want it, or because AI makes it easy?

Can marketers prove AI ROI?

Fewer marketers are confident they can prove ROI in 2026 than in 2025, according to MarTech's earlier reporting. As with advertising, teams that want to hold on to the budget must now prove ROI for AI too. A significant measurement problem is looking beyond outputs like content pieces created — teams must account for rework, integration costs, customer outcomes and business impact.

What happens when models change?

AI models are changing rapidly, and those changes introduce unpredictability into workflows. Marketers are left puzzling over why a tool that worked fine one day stops working the next. The chaos this causes depends on how deeply the AI was embedded in the first place.

So what?

For MarTech, the eight observations point to one conclusion: "AI adoption in marketing is widespread, but meaningful AI transformation remains elusive for many organizations." This is not a marketing-specific problem — AI adoption has outpaced organizational strategy, and integrating AI platforms, the existing stack and the data must happen at the organizational level, not by the marketing team alone. Technology and business leaders need to collaborate on the next step. And on the familiar issues — quality versus quantity, workflow efficiency, customer experience, business value — marketers shouldn't expect AI to solve them.