How AI in B2B Marketing Delivers Results

Nearly nine in ten companies now use AI regularly in at least one business function, B2B marketing included. Few see it in their bottom line. In McKinsey’s State of AI 2026, 37% of respondents attribute at least some of their operating profit to AI, practically unchanged from the previous year. Only about 6% are high performers that attribute 5% or more of their operating profit to AI. What sets them apart is mainly how they work. Nearly 75% of them have fundamentally redesigned their workflows because of AI. Among all other companies, the figure is about 25%.

For B2B marketing, this means AI needs an operating model. It defines what AI is used for, who steers it, and by which rules. In marketing, AI takes on three kinds of work. It analyzes data, calculates forecasts, and answers questions about a company’s own data. It creates content, campaigns, and variants. And it executes. As an agent, it completes multistep tasks on its own and uses other systems to do so. With each of these steps, AI works more independently, and the need for control grows. Governance sets the framework for this, and clear roles define who does which work.

Area	What AI does	Which AI is behind it	What people need to do
Analyze	Analyze data, calculate forecasts and scores, answer questions about your own data	Analytical AI, with generative AI as a language interface	Review results, ensure data quality
Create	Generate content, campaigns, and variants	Generative AI	Set direction, protect brand and facts
Execute	Complete multistep tasks independently across several systems	AI agents: a language model plans the steps and uses other tools, from analytics and generative AI to CRM and marketing automation	Set limits, build in approvals, monitor results

Analytical AI for Forecasting and Scoring

Analytical AI, also called predictive AI, works with structured data such as numbers from spreadsheets, CRM fields, and transactions. It uses past data to calculate what is likely to happen next: which accounts are ready to buy, which customers might churn, and how the pipeline will develop. Companies have used this kind of AI for decades, long before generative AI arrived. In many companies, it is already built into the CRM, marketing automation, and ABM platforms, and often only a fraction of it is used.

How AI Lets You Talk to Your Data

What is new is the combination with generative AI, which makes this data accessible in conversation. With “talk to your data,” a marketing leader asks in plain language which accounts in a given industry churned last quarter, and generative AI turns the question into a CRM query. The numbers in the answer come from the system and can be checked there. Errors mainly occur when the AI misunderstands the question. In a Gartner survey of 403 analytics and AI leaders, more than half said their organization already uses AI for automated insights and natural language queries. Many CRM and marketing platform vendors now build such language interfaces directly into their products, along with agents.

Why Good AI Answers Need Clean Data

An AI that draws on a poorly maintained CRM delivers poor answers in persuasive language. That is why data quality is one of the core responsibilities of the B2B CMO, because without it, neither forecasts nor measurements are reliable. As third-party data loses value under data protection rules, first-party data becomes the most important foundation for every further AI application in marketing.

How Generative AI Creates Content

Generative AI creates something new: text, images, variants, and drafts of entire campaigns. Its tasks are open-ended, and its results are harder to predict than the forecasts of analytical AI. That is why it needs more guidance.

A person sets the direction. A blueprint defines the boundaries: target customers, messages, tone of voice, and a clear definition of when a result is finished. A cycle of building, reviewing, and correcting secures quality until the result meets your own standard. This approach works regardless of the tool, with general-purpose language models as well as specialized marketing tools or custom AI assistants.

The Blueprint as the Foundation for Coherent Content

The blueprint consists of a library of approved building blocks such as value propositions, proof points, use cases, and answers to objections, each tagged by persona, industry, and stage of the buyer journey. It also includes templates, instructions for the AI, and verified facts: the figures, product data, and sources that all content relies on. This keeps the landing page, the email, and the sales deck aligned on substance. It also helps with AI search, because AI can place a company more precisely when it is described consistently.

For how things are said, the blueprint only sets guardrails, such as a tone of voice that fits the brand. Within these guardrails, tone and emphasis adapt to the channel and the audience. And what makes a company distinctive comes from its people: customer stories, project experience, and a clear point of view. When many competitors work with the same language models, that is exactly where the difference lies.

Why People Need to Review the Results

Even when AI writes only from verified sources, it makes mistakes. A Stanford University study examined AI tools for legal research that build their answers solely from a defined collection of court decisions and statutes. They made far fewer errors than general-purpose chatbots. Even so, depending on the tool, 17% to 34% of their answers were wrong.

Where generative AI reaches its limits, the experience of the people reviewing its output decides. Their judgment is an investment that CMOs should account for in capital allocation from the start.

How AI Agents Execute Tasks in B2B Marketing

An AI agent works toward a goal. A language model plans the necessary steps and uses other tools through interfaces. It queries data in the CRM, draws on the forecasts of analytical AI, writes text with generative AI, and operates conventional software such as marketing automation or email systems. Marketing automation keeps running as before. The agent takes over work that a person used to do, such as selecting contacts, preparing copy, and launching a campaign.

An Example from B2B Marketing

After a trade show, marketing needs to follow up on every conversation. An agent pulls the contact list from the CRM, uses scores to check which accounts are ready to buy, and writes a draft for each contact from the approved building blocks. It sets up the emails in the marketing automation platform and flags the accounts with the greatest potential to sales. Before any email goes out, a person approves it.

Why Agents Need Clear Limits

An agent acts in real systems. Its mistakes do not stay in a draft. They reach customers, data, and budgets. Gartner expects more than 40% of agentic AI projects to be canceled by the end of 2027 due to rising costs, unclear business value, or inadequate risk controls. Teams that deploy agents therefore define what an agent may decide on its own and where a person must approve. They also monitor what the agent does so that errors surface early. These rules are part of governance.

Governance as the Framework for AI in B2B Marketing

The more independently AI works, the more important clear rules become. That is why governance applies to all three kinds of AI work: analyzing, creating, and executing. It defines who decides, which data the AI may use, and how its results are reviewed. In practice, this comes down to a few clear rules. Which tools are approved? Which data may go into which system? Where must a person review and approve? And who is accountable when a result is wrong?

Without such rules, teams find their own way. Marketers may end up working with customer data in personal AI accounts because the official tools are missing or too cumbersome. This is usually a sign that teams want to work faster than their tools allow. Good governance makes the official path the easiest one.

What the EU AI Act Means for Marketing Teams

For companies in the EU, the EU AI Act sets the framework. Two obligations affect marketing teams directly. Since February 2025, companies that use AI must ensure their staff have sufficient AI literacy. National authorities have supervised this obligation since August 2026, and according to the European Commission’s Q&A, internal training records are sufficient proof. Since August 2, 2026, transparency obligations have also applied: people must be able to tell when they are interacting with AI, and realistic images or videos generated by AI must be labeled as such. The GDPR applies alongside, including when teams use generative AI with customer data.

Governance in the Workflow

Governance only takes effect when it is built into workflows: as an approval step before anything is published, as clearly limited AI access to data, and as a record of who reviewed what and when. For agents, it also defines which actions they may take on their own.

The Roles in an AI Operating Model

So far, we have looked at the work AI takes on in marketing. Now we turn to the people and roles that steer this work. McKinsey estimates that 75% of all roles need fundamental reshaping. Most of them will not disappear. They will change.

For working with AI, McKinsey distinguishes between two positions. People who work “in the loop” are part of the workflow itself and review and edit what the AI delivers. People who work “above the loop” stand over the workflow and define how it runs. Each of the following roles takes one of these two positions.

The Architect

The Architect works above the loop. This is usually the CMO or a senior marketing leader. The more work AI takes on, the less the Architect decides about individual pieces of content and the more about the system in which they are created. The Architect sets the strategy: which target customers marketing addresses, which messages apply, and which campaigns take priority. The Architect defines the limits for AI, meaning which decisions agents may make on their own and which stay with people. The Architect decides where AI is used and where the budget goes, and remains accountable for the result, even when AI did the work. And the Architect redesigns the team. Gartner ranks building hybrid teams of people and agents among the top priorities for CMOs.

The Orchestrator

The Orchestrator also works above the loop and turns the Architect’s guidelines into workflows in which people, AI tools, and agents work together. The closest role today is the marketing automation manager, who also builds workflows and connects systems. What is new is that AI does not follow fixed rules and produces a different result with every run. For each workflow, the Orchestrator therefore defines which data the AI may access, which instructions it follows, what agents may do on their own, and at which point a person reviews the result before it moves on.

The Content Engineer

The Content Engineer takes on this task for content, building and maintaining the blueprint described in the section on how generative AI creates content: building blocks, templates, instructions for the AI, and verified facts. In midsize companies, this is often the same person as the Orchestrator. In larger organizations, specialists fill these roles.

The Production Team

The Production Team works in the loop. It creates campaigns and content, increasingly together with AI agents. It reviews the results, adapts them to markets and customers, handles exceptions, and gives final approval. In the trade show example, this is the team that reviews every email before it goes out.

Role	Position	Task	Result
Architect	Above the loop	Sets strategy, limits for AI, budget, and team structure	A framework that all workflows align with
Orchestrator	Above the loop	Designs workflows of people, AI, and agents and defines where review happens	Workflows that deliver reliable results
Content Engineer	Above the loop	Builds and maintains building blocks, templates, instructions, and verified facts	A content system that produces coherent content
Production Team	In the loop	Works with AI and agents, reviews, adapts, approves	Campaigns and content at speed, within the brand

In this model, marketing work shifts. Less time goes into producing individual pieces of content, and more goes into strategy, designing workflows, and reviewing results. This raises familiar questions of organization in a new way. Does it still make sense to split in-depth content and the short-form copy derived from it, such as a white paper and its social media posts, between different teams or agencies, when AI can derive the posts directly from the white paper? And which tasks stay in the regions and which move to headquarters when AI can translate and adapt content, but review still requires people who know the market?

Content as a Source for AI Answers

So far, the focus has been on AI that marketing uses itself. AI also plays a growing role on the buyer side. According to Forrester, 94% of B2B buyers use AI in their buying process, and twice as many as the year before rate generative AI or conversational search as more important than any other source of information, including vendor websites. Which companies appear in the answers depends on what these systems find about them and how coherent the picture is that the website, sales, and partners paint together.

Google sets no special rules for this. Its AI features follow the same fundamentals as classic search: important content in text form, good internal linking, and structured data that matches the visible text. Content that answers a question directly, cites its sources, and is clearly structured is therefore easier for any system to evaluate and to cite in AI answers.

New Content from the Content System

This works best with content designed this way from the start. The blueprint provides the foundation: building blocks with clear answers, verified facts with sources, and templates that already include a clear structure and frequently asked questions with answers. New content is then created in a form that both people and AI systems can read easily.

Retrofitting Existing Content

Most companies, however, have a library of articles, white papers, and product pages written before AI search existed. Rewriting them costs too much. An example from our own work: our assistant for generative engine optimization (GEO) takes a finished article and creates the elements that search engines and AI systems evaluate. These include the SEO title and metadata, an introduction that answers the central question directly and cites its sources, and frequently asked questions with answers. The result is a first draft. Subject matter experts review and finalize it, because they can judge whether a statement is accurate and a source holds up.

Both approaches complement each other. Retrofitting quickly makes existing content visible in AI search. The content system ensures new content is created this way from the start.

How to Measure the Value of AI in B2B Marketing

That leaves the question every board asks: what does AI deliver? A company’s revenue depends on many factors, including product, price, sales, and market. The value of AI on its own is therefore best measured where it changes the work. How long does a result take? What does it cost? And how much rework does it need before it can be approved?

Which Metrics Show the Value of AI

The basis is a comparison with how things work without AI. This does not have to be complicated. For two or three recurring tasks, it is enough to record how long they take today, what they cost, and how often they need rework. The same values are then measured after AI is introduced. Costs now include licenses and compute, as well as the time people spend reviewing and correcting.

When an AI Agent Pays Off

Whether an AI agent pays off depends on the ratio between work time and review time, as David Tepper, CEO of Pay-i, calculates in a McKinsey Quarterly interview using a simple example. His example, applied to 100 tasks: a person needs two hours for each, 200 hours in total. If an agent takes over the tasks, a person reviews each result in six minutes, 10 hours in total. Whatever the agent fails to complete properly, the person does as before. If the agent succeeds at 5% of the tasks, the hours saved exactly offset the review time. Each additional percentage point saves two hours. At 50%, the effort drops from 200 to 110 hours. The calculation assumes, however, that a poor result can simply be discarded, such as a failed draft. That is why every step with external impact needs approval by a person, as in the trade show example.

Time saved and costs cut are not yet revenue. They create room for the work that drives results. Whether that work ultimately pays off shows up in the metrics marketing uses to prove its contribution to business results.

Where People Make the Difference

AI analyzes data, writes drafts, and completes tasks across multiple systems faster than any team. It does not set the direction. Which customers a company wants to win, which story it tells, and which risks it takes are decisions people make. The more work AI takes on, the more weight these questions carry.

This is also why AI delivers results in some companies and not in others. The high performers in McKinsey’s study redesigned their workflows around AI. They defined what AI takes on, where people review, and who is accountable. The technology is available to every company. The difference lies with teams that know where their marketing is headed and organize their work accordingly.