AI for marketing used to be fairly easy to explain. Most products generated copy, while predictive analytics and personalization carried on in older corners of the MarTech stack. The label has survived. The market underneath it has changed considerably.

In our new AI in Marketing 2026 evaluations, Real Story Group reviewed 14 vendors against the work enterprise marketers are actually trying to get done. We found useful advances in content production, research, workflow assistance, decisioning, and custom development. We also found that putting all these products on one undifferentiated shortlist makes very little sense.

Compare Adobe, Jasper, and OpenAI, for example. All three can help a marketing team create content. They enter the enterprise through different buying centers, depend on different data and governance arrangements, and play very different roles once the pilot ends.

We therefore organized the evaluations into three categories:

• Suite-Native AI: AI embedded inside broader marketing, CRM, customer engagement, or DXP suites.
• Specialist Marketing AI: purpose-built AI products for marketing content, GTM workflows, creative production, decisioning-adjacent work, or enrichment.
• Enterprise AI Platforms: horizontal AI workbenches, model platforms, and developer services that marketing teams configure for their own needs.

What’s missing here? Nearly every other platform that calls itself “AI,” but is really a CDP, ESP, or WCM with AI functionality bolted on.

What Changed in RSG’s Research

The old evaluation model gave too much space to generic writing and editing capabilities. Those features still matter, but they no longer tell buyers very much about enterprise fit. This year, we shifted to a 10-scenario taxonomy:

Content & Creative AI

• Brand-Safe Content Supply Chain
• Content Localization and Market Adaptation

Intelligence & Insights AI

• Audience Intelligence and Segmentation
• Market and Customer Insight Synthesis
• Planning and Brief Generation
• Predictive Customer Analytics

Decisioning and Personalization AI

• Journey Orchestration and Real-Time Personalization
• Lifecycle Marketing and CRM Automation
• Paid Media Optimization

Agentic Marketing AI

• Agentic Campaign Orchestration

We also tightened our strategic assessments around governance, responsibility, value for money, and composable-stack friendliness. An otherwise useful stand-alone “platform assistant” can become a very different proposition once it touches customer data, approval workflows, channel execution, or paid-media budgets.

The 2026 set combines updated evaluations with newer entrants and products that have changed direction. The 14 vendors are Adobe, Amazon, Anthropic, Clay, Copy.ai, HubSpot, IBM, Jasper, Microsoft Copilot, OpenAI, Persado, Salesforce, Typeface, and Writer.

What We Found

The marketplace now contains three different architectural options.

Suite-native products have the closest access to customer data and campaign systems. That access can also deepen suite dependence and make packaging harder to decipher. RSG remains very skeptical of this class of solutions. Specialists tend to be clearer about the job they perform, although few can replace core MarTech systems. Enterprise AI platforms offer far more freedom to build, yet the customer inherits most of the integration, operating-model, and governance work.

Content generation has become the entry ticket. Nearly every vendor can produce a usable first draft. The more revealing tests concern brand rules, rights, legal review, approvals, provenance, DAM or CMS integration, and exception handling. Most product demos spend remarkably little time on the exceptions.

Agentic marketing was the least convincing part of many vendor stories. Products can automate individual tasks around briefs, copy, enrichment, and routing. Campaign authority raises harder questions. Who can approve an action? Which connectors can the agent use? What gets logged? Can a team undo the action, and can it cap the system's spending? Buyers should insist on seeing those controls rather than accepting an orchestration diagram.

Decisioning has also returned to the center of the conversation. Personalization, next-best action, offer selection, and send-time optimization can create value. However, they depend on identity, consent, channel ownership, reward signals, and credible measurement. Weak foundations turn AI decisioning into expensive guesswork.

Then there is pricing. Vendors now mix seats with credits, actions, tokens, model calls, enterprise minimums, add-ons, services, and usage tiers. An inexpensive pilot can reveal almost nothing about production cost. Model the volume and failure paths of a real scenario before treating an entry price as meaningful.

How Buyers Should Use the Evaluations

Start with the scenarios: which ones are truly high priority? A general ambition to “use more AI in marketing” is not a useful requirement and will produce an equally vague shortlist.

Make the vendor work with representative data and content. Test your approvals, channels, governance rules, and expected production volume. Pay attention to who will operate the system after the pilot team moves on.

If a vendor cannot explain who approves an action, what gets logged, how a bad action is undone, and what the system costs at production volume, you have learned more than the scripted demo was designed to tell you.

The updated evaluations are available to RSG Corporate Members. If you’d like help evaluating AI technologies for your enterprise, contact us directly.