Your AI Knows Marketing. Does It Know Your Business?

This is an illustrative example from our recent presentation on unifying AI, but the underlying frustration is familiar: the knowledge needed to make a good decision already exists somewhere in the organization. The people making the decision just don't have it.

Now put AI into that meeting. Give it the traffic dashboard and ask for recommendations. Does it know why the campaign was stopped? Can it connect that decision to what analytics is seeing? Or will it help the team build a very convincing plan to undo something that was working?

That's the question I think marketing leaders should be asking as they invest in AI. An enterprise marketing leader described the problem to us this way: “We don't have coordination between departments.” Teams were adopting AI independently, without a shared understanding of the business behind their work. Making each of those teams faster doesn't resolve what gets lost between them.

What is AI context in marketing?

At Traction, we define context as data and institutional knowledge, applied to the decision in front of you. It brings together what happened, what your organization knows and believes, and what you're trying to accomplish now.

Institutional knowledge is the part I don't think gets enough attention. It tells your AI what your company knows and believes. It gives your AI soul. That includes the reasons behind your priorities, the lessons from work that didn't succeed, and the judgment your team has developed about what customers value. Those things shape a useful recommendation just as much as the numbers do.

Consider another example: a competitor cuts its price by 15%. A company built around value, with a goal of gaining market share, might match the price and promote aggressively. A premium brand focused on protecting margin might hold its price and invest in loyalty. Either response could make sense for the business making it.

The price change doesn't tell you which decision to make. You need to understand the company, its customers and what it is trying to achieve. We expect an experienced marketer to bring that judgment to the conversation. When we ask AI for strategic recommendations, we need to give it access to the knowledge that informs that judgment.

The difficulty is that this knowledge rarely lives in one place. Some of it is documented in research and strategy decks. Some belongs to an agency that remembers why a previous approach failed. Some exists in conversations and decisions that never made it into the official record. Having the information somewhere is different from making it available when someone needs it.

You can see the difference in a campaign brief.

In the presentation, we used an illustrative brief for a Q4 campaign aimed at enterprise CMOs evaluating AI. Give the system an audience, a pipeline objective and a general message about efficiency, and you get a familiar direction:

Unlock efficiency at scale — AI that helps your team do more, faster.

There's nothing obviously wrong with it. It identifies a benefit, sounds professional and could sit comfortably on a website. But it gives a CMO very little reason to choose this company over another one.

Now give the system the context behind the assignment. The immediate business priority is securing three enterprise pilots this quarter. Prospective customers already have too many disconnected AI tools. Security and control over models are stopping evaluations from progressing. The company's position is to work with the customer's existing stack, rather than replace it.

The direction becomes:

Make the AI you already have work together — governed, coordinated, proven on one operation.

That second direction responds to the buyer's situation and the obstacles to a purchase. It also supports the company's immediate commercial objective. In the example, the model hasn't changed; the understanding behind the work has.

This is the standard I'd use to evaluate AI-generated marketing. Does it reflect the things we know about our business that should materially change the answer? Could someone explain why this recommendation makes sense for us, rather than simply recognizing it as competent marketing?

A polished output can pass a quick review. A useful one needs to survive those questions.

What does a marketing context engine do?

A marketing context engine is a shared, governed layer that makes company knowledge available to connected AI tools and workflows. It brings together relevant data, institutional knowledge, objectives and rules, with a way to keep that context current. This is the foundation of the AI marketing platform we're building at Traction.

Go back to the traffic example. Storing the media plan would help, but the workflow also needs the reason the campaign was stopped and the performance information that followed. Otherwise, the system may retrieve a document without understanding the decision it is supposed to inform.

The same principle applies to the work marketing teams produce every day. A strategic brief should account for what failed last time. A customer journey map should draw on the research the company has already conducted. A media plan should reflect what the organization has learned from its spending. Those are the starting points we want to improve, rather than asking every person and every workflow to reconstruct them.

Our approach to unifying AI does not require every team to use the same tool. It calls for a shared foundation of context behind the tools and workflows that connect to it. The important question is whether the work starts from an agreed understanding of the business.

There is an operating responsibility here, too. Someone has to determine which information is authoritative, who can use it, and when it needs to be updated. New learning needs a route back into the system. Without that discipline, you risk making old assumptions more readily available instead of making decisions better.

Why we start with marketing system design.

Once you look at context this way, the work extends well beyond software. You have to understand how information moves, who owns a decision, where teams depend on one another, and what happens when their priorities conflict.

That is why our Marketing System Design practice looks across people, platforms, partners and process. We combine marketing operations, organization design and technology strategy, beginning with a data-driven diagnostic and a blueprint for what needs to change.

For enterprise clients, we lead with consulting and a Marketing System Diagnostic. Before recommending a platform or building a workflow, we need to understand where the current system is falling short. Which decisions are being made without the right information? Where are teams duplicating work? What requires too much effort to assemble, reconcile or approve?

The answers should shape the technology, along with the changes in ownership and ways of working that will make it useful. A briefing problem caused by conflicting objectives needs a different response from one caused by difficulty finding approved customer research. Both may benefit from AI, but they aren't the same problem.

The platform we're building is intended to put shared context to work across marketing workflows. For mid-market teams ready to adopt a common platform, it may offer a more direct starting point. For a complex enterprise, the first priority may be improving one operation within the environment it already has. We should choose the entry point based on the work and the organization's readiness, rather than assume everybody needs the same implementation.

Start with one operation you can improve.

The practical starting point in our presentation was deliberately modest: diagnose the gap, capture the context, co-build with the team, pilot one output and measure what changes. The ambition can be broad, but the first test needs to be specific enough to learn from.

Campaign briefing is one possible place to start. Map what it takes to produce an approved brief today, including the time spent gathering background, resolving disagreements and revising work. Then identify the context that should be available at the start: the commercial objective, customer research, positioning, past performance and relevant constraints.

Build the pilot with the people who write, review and use the brief. Decide in advance how you'll judge it. I'd want to know whether the team spent less time reconstructing background, whether reviewers had fewer fundamental corrections, and whether the final brief gave the people executing it clearer direction. Producing a first draft faster would be useful, but it wouldn't tell the whole story.

The presentation identifies less rework, faster time to market, greater relevance and better use of spend as the outcomes to pursue. A pilot should test which of those outcomes actually improves, rather than assume that connecting AI to more information guarantees a result.

What the team learns can then inform the next workflow, provided that learning is reviewed and captured. That is the longer-term value of building a shared context layer: useful knowledge doesn't have to disappear when the project ends.

A Marketing System Diagnostic is where we begin that work with enterprise teams. It gives us a way to identify a meaningful problem, understand the context it requires and decide what is worth building.

The next time website traffic drops, your team should be able to spend the meeting deciding what to do next, rather than rediscovering a decision it already made. That's a practical standard for the AI marketing system we're building.

FAQ

What is a marketing context engine? A marketing context engine is a shared, governed layer that makes company knowledge available to connected AI tools and workflows. It brings together relevant data, institutional knowledge, objectives, and rules, with a way to keep that context current. It is the foundation of the AI marketing platform Traction is building.

What does context mean in AI marketing? At Traction, context is data and institutional knowledge applied to the decision in front of you. It combines what happened, what your organization knows and believes, and what you are trying to accomplish now. Institutional knowledge is the part that gets the least attention, because it carries the reasons behind priorities and the lessons from work that did not succeed.

Why does AI give generic marketing recommendations? Because it usually only sees the dashboard. If traffic fell 30% last month, AI can recommend spending more on paid media without knowing the media team deliberately turned off a campaign that was buying clicks instead of customers, and that cost per transaction fell 40% while conversion held steady. Without that knowledge, it can build a convincing plan to undo something that was working.

How do I judge whether AI-generated marketing is any good? Ask whether the work reflects the things you know about your business that should materially change the answer. Then ask whether someone could explain why this recommendation makes sense for your company, rather than simply recognizing it as competent marketing. A polished output can pass a quick review, but a useful one has to survive those questions.

How do I start giving AI more business context? Start with one operation you can improve: diagnose the gap, capture the context, co-build with the team, pilot one output, and measure what changes. Campaign briefing is a practical place to begin, since you can map what it takes to produce an approved brief today and then supply the objective, customer research, positioning, past performance, and constraints up front. For enterprise teams, Traction starts with a Marketing System Diagnostic.