Go into almost any office or jump onto a team Slack channel today, and you’ll see the same thing happening. Business owners, founders, or marketers go to ChatGPT or Claude, type “Build a marketing strategy for my business,” then wait a few seconds. They receive what it looks like a thought-out and impressive marketing strategy. It may include practical tactics such as SEO, paid search and social, automated email nurtures, webinars, and ABM campaigns.
To a busy professional or marketer, this feels like an incredible time-saver. But then your team starts implementing it and you do not end up with the results you want. You have just wasted time and resources, and in some cases, you can lose business as well.
Then your CMO or a top executive will say, “no more AI for marketing.” What the marketing team really needs is proper protocols, and AI governance and policies.
THE STRATEGY LACKS CONTEXT
It all starts with proper AI education and understanding how large language models (LLMs) work. At their core, LLMs generate responses based on probability.
Think of it this way: If you say “hello” to a complete stranger, you can expect a polite, predictable reply like “hello” or “how do you do?” But if you say that exact same “hello” to a close friend, they might respond with “what’s up, buddy?”
Your prompt didn’t change at all. The context did. Without deep company context, an AI tool defaults to the polite, generic stranger response every time.
Same goes with your marketing. If you don’t give it custom context, desired outcomes, available resources, budget limits, and existing data, the model has no choice but to default to the most generic answer. Sure, it may look like a well-planned marketing strategy, but it very likely misses the mark for your business realities.
STRATEGIC COPILOT
This is why your prompts are important—hence the art of prompt engineering. But also important is context engineering, the art of providing proper data and context so AI can avoid generic answers. A combination of prompt engineering and context engineering is sometimes referred to as decision engineering by fellow AI nerds, who like to give names and labels to everything. Essentially, it means structuring your internal data and operational boundaries before you ask the model for a more customized recommendation.
In other words, don’t ask AI to be your pilot. Leverage it to be a seasoned and sharp copilot. Only then can it synthesize market research, draft campaign concepts, analyze customer feedback, and surface anomalies in your analytics.
The good news is that popular LLMs like ChatGPT and Claude now connect directly to internal business systems’ standard frameworks like Model Context Protocol (MCP) and via APIs. MCP, in particular, is very interesting. It essentially acts as a bridge that allows AI models to safely access your real-time data, like CRM data, marketing analytics dashboards, campaign performance, and project management tools, without requiring you to manually paste sensitive information into every prompt. This not only saves you a ton of time and headaches, but also allows AI to read your business inputs directly from the source.
YOU SET THE DIRECTION
However, technology alone still will not make the executive calls for you. It does not take into account the “non-digital data” and your knowledge that is not available as digestible data. There’s no gut feeling, experience, or intuition here, regardless of how much real-time data the pipelines feed your model.
The AI remains an advisor, a copilot, or a very smart intern who graduated from MIT, not an executive. It can process the context quickly, but you still have to set the direction. You have to define the boundaries and decide which risks are worth taking. It is ultimately your responsibility to make sure the marketing strategy is the best fit for your business. You can certainly use AI to draft your marketing strategy, as long as you understand the fundamentals discussed earlier.
Also, ask yourself and your team the questions below:
- What specific business outcome does this initiative directly support?
- What internal data or verified customer evidence underpins this recommendation?
- What lower-priority initiatives are we explicitly stopping to fund and staff this?
- Who owns execution, and do they have the bandwidth to succeed?
- What 30-, 60-, or 90-day metric will tell us this strategy is working or signal that we need to pivot?
If your AI-generated plan cannot answer these questions with real data from your business, you do not have a strategy yet. You have a plausible draft. Treat it as a starting point, apply your executive judgment, and make the tough tradeoffs that drive genuine growth.
Al Sefati is CEO of Clarity Digital.