STRATEGY · Chapter 2
Build the strategy before you buy a single tool.
The most expensive mistake I see in AI marketing is not picking the wrong tool. It is picking the right tool with no positioning behind it. You end up paying $200 a month for a subscription producing content nobody wants to read. This chapter is the order of operations that keeps that from happening.
Four questions. In this order.
Skip one and the rest of the playbook will not compound. I have watched smart founders skip question one, spend six months on questions three and four, then quietly start over. Do them in order.
- Who is the customer, really? Not a persona deck with a stock photo named "Marketing Mary." A specific company size, a specific role, and the specific event that makes them go looking for help. If you cannot point at three real customers in your CRM who match the description, the ICP is still too vague. Tighten it.
- What position do you actually hold? The one thing your ICP gets from you that they do not get from the obvious competitor. Here is the test I use now: ask ChatGPT a question your buyer would ask in your category. If the answer does not naturally use phrasing that sounds like yours, your position is not sharp enough yet.
- Which content pillars carry the position? Three to five topics deep enough that you can credibly own them. Each pillar is a hub page, a set of sub-pages, and a blog cluster. Not one article. Not a category page with three blog posts under it. A body of work.
- How does the work actually get done? Who writes the first draft. (Usually AI.) Who edits. (Always a human, ideally one who knows the customer.) Where the draft lives, how it gets QA'd, who hits publish, who checks it got indexed. The workflow is the strategy made operational. No workflow, no strategy.
- Founder-led SMBs that have never written down their ICP.
- Teams that bought AI tools, used them for a month, then drifted back to old workflows.
- Businesses with a real product and unclear category language.
- Nobody can book five customer calls in the next three weeks.
- The founder wants a strategy but will not commit to one position.
- Leadership thinks AI will replace the strategic conversation.
- You can name your ICP in one sentence with a role, a size, and a trigger event.
- Your position has a specific competitor it stands against by name.
- Each content pillar has at least 20 question-shaped sub-topics drafted.
What actually drives the cost
I do not publish dollar figures here because the inputs vary too much for any number to be honest. What I can publish is what makes the project cheap or expensive, so you can read your own situation and predict whether you are looking at two weeks or two quarters.
| Driver | What it is | What pushes it up |
|---|---|---|
| Foundation work | ICP definition, positioning, content pillars, schema and entity graph, AI-crawler-friendly robots and llms.txt. | Existing site debt, multiple service lines, multi-language, or a complicated org chart with no current marketing owner. |
| Content velocity | Pages, posts, and updates produced per month. | Wide topical surface, low existing topical authority, multiple personas, and weekly publishing cadence. |
| Tool stack | AI models (Claude, GPT, Gemini), CMS, analytics, CRM, automation, monitoring. | Duplicated tools, enterprise CRM seats not tied to revenue, model usage at scale without prompt caching. |
| Human review | Editorial sign-off, SME interviews, brand-voice QA, factual review for YMYL or regulated industries. | Regulated sectors (legal, health, finance), brand-sensitive topics, multilingual review. |
| Distribution and outreach | Email, social, paid amplification, PR, podcast and event placements. | Cold audience launch, no existing list, dependence on paid until organic and AI surfaces compound. |
The two-week strategy sprint
For a 5 to 50 person company with a working product, this is the schedule that consistently lands. I ran a Vancouver B2B services firm through this exact sprint last quarter. By the end of week two their sales team was using the new positioning on calls without being told to.
- Week 1. Book and run four to six 30-minute customer calls. Drop the transcripts into an LLM for thematic synthesis. Founder writes a one-page positioning draft. Do not skip the draft. Writing it is where the position gets sharp.
- Week 2. Lock the ICP, the position, and three to five content pillars. Write the workflow doc. Queue the editorial calendar for the first 90 days. Now, and only now, pick the tools.
Tool selection at the end of week two is fast because the requirements list now exists. Without that list, tool selection takes months and ends in the same shelfware regret.
Frequently asked
What goes into an AI marketing strategy for a small business?
+
Four answers, in this order. Who is your ideal customer. What position do you hold that the obvious alternative does not. Which three to five content pillars prove that position. How do AI and humans split the work of producing and measuring it. Pick your tools after those four are written down, not before. Picking tools first is how a business ends up paying for $400 a month in subscriptions nobody remembers to cancel, chosen for features that may not even match the strategy once it finally gets written. Write the four answers down on paper first, then go shopping for software.
Should I build the AI marketing strategy before hiring an agency?
+
Building the strategy first is almost always the right order. Hire the agency before that strategy exists, and the business ends up paying that agency's hourly rate just to figure out what the business even does and who it actually sells to. That is the most expensive research project a business will ever fund, and worse, it locks the whole strategy to whatever specific tactics that particular agency happens to sell, regardless of whether those tactics fit the business's real customers. Writing the strategy first means the agency gets hired to execute a plan, not to invent one on the business's budget.
How long does it take to define ICP and positioning?
+
Two to three weeks if the founder is actually in the room for the work, plus four to six customer calls of about 30 minutes each. The analysis itself is not the hard part. The hard part is getting those four to six calls actually booked on real customers' calendars. Every stalled strategy project worth remembering stalled at the calendar, not at the analysis step that came after it, so protecting that calendar time before anything else is the real first move.
Can AI help build the strategy itself?
+
AI works well as a fast research analyst for this kind of work. It can synthesize interview transcripts, draft a few positioning options to react to, and surface patterns a founder might miss while reading five customer call transcripts back to back in one sitting. What AI does not do is decide which customer segment to walk away from or which trade-off the business should make when two good options conflict. That decision belongs to the founder, and no model should be allowed to make it on their behalf.
What does a content pillar look like in an AI marketing context?
+
A pillar is a topic a business owns end to end, usually built from 30 to 80 specific questions a real buyer actually asks, mapped out across one hub page, several sub-pages beneath it, and a supporting blog cluster, all cross-linked back to each other. For most SMBs, three to five pillars is enough to cover the ground that matters. Going past five pillars tends to spread a small team thin across more topics without actually adding meaningful coverage for the buyer reading any single one of them.
How do I know if my positioning is actually differentiated?
+
Say the positioning sentence out loud, then ask honestly whether a direct competitor could say that exact same sentence about their own business without changing a word. If the answer is yes, what was written down is not real positioning, it is just a description of the category the business happens to compete in. Genuine positioning names one specific customer, one specific problem that customer has, and one specific reason this business solves that problem better than the obvious alternative a buyer would otherwise default to.
What is the biggest mistake SMBs make when writing their strategy?
+
The biggest mistake is writing the strategy once and then never revisiting it again. A strategy document written 18 months ago describes a business that has since changed its pricing, its ideal customer, or its main competitor, often without anyone noticing the strategy document never caught up. Reviewing the four core answers, meaning ideal customer, positioning, content pillars, and the AI-human work split, every two quarters rather than once a year keeps whatever content gets built on top of that strategy actually aimed at the right target instead of an outdated one.
Should the strategy document be shared with the whole team or kept with leadership?
+
The strategy should be shared broadly, but in a simplified form. The full strategy document, including raw customer-interview detail, can reasonably stay with leadership alone, but every person who touches content, sales conversations, or direct customer interactions needs the plain version of it: who the business sells to, what it says about itself, and what it deliberately does not do. A strategy that nobody outside leadership has actually read tends to get reinvented informally by each team member on their own, usually inconsistently and often in conflicting directions.