How to Build a 90-Day Marketing Plan With AI (In One Afternoon)

Most small businesses don't have a marketing plan. They have a vibe — post when inspired, run an ad when things are slow, panic in the quiet months. Not because owners don't know a plan would help, but because planning feels like a two-week project for a consultant they can't justify hiring.

It's an afternoon now. A real 90-day marketing plan — who you're targeting, what you're doing each week, what you're spending, how you'll know it worked — is exactly the kind of structured document AI drafts well, provided you feed it real inputs and then argue with the output.

That second part matters. This post walks through the actual working session: what to gather, the prompt to run, how to pressure-test the draft, and how to turn it into a Monday-morning checklist. Block off three hours.

Hour one: gather your inputs

The plan will be exactly as good as what you feed it, so don't skip to the prompt. You need four things, roughly a page total.

• Who you sell to. Not "homeowners" — your best customers specifically. Who they are, what they're buying from you really, what almost stopped them. If this is fuzzy, spend twenty minutes with the customer persona builder first; a plan aimed at everyone reaches no one.
• What worked and didn't last year. Where did your best customers actually come from? Referrals, Google, the sign on your truck? Write down your best guess per channel, even if it's rough. Also write what you tried that flopped.
• Your numbers. Monthly marketing budget you can sustain for three months, hours per week you'll honestly give this (be pessimistic), and one revenue goal for the quarter.
• Your one priority. More leads, higher-value jobs, reactivating past customers — pick one. A 90-day window is too short for three priorities, and asking AI for everything gets you a plan that's a to-do list in a trench coat.

If you've already built a brand voice card, have it handy too — the plan's content weeks will reference it.

Generate the 90-day marketing plan skeleton

Now the prompt. Paste this in with your inputs filled, and don't trim the rules at the bottom — they're what separates a plan from a listicle.

You are a marketing strategist who builds realistic 90-day plans
for small businesses with no marketing team.

MY BUSINESS: [WHAT YOU SELL, TO WHOM, WHERE]
BEST CUSTOMERS: [PASTE YOUR PERSONA NOTES]
WHAT WORKED LAST YEAR: [CHANNELS THAT BROUGHT REAL CUSTOMERS]
WHAT FLOPPED: [WHAT YOU TRIED THAT DIDN'T WORK]
MONTHLY BUDGET: [$X] | HOURS PER WEEK: [X] | QUARTER GOAL: [ONE GOAL]
MY ONE PRIORITY THIS QUARTER: [E.G. MORE QUALIFIED LEADS]

BUILD A 90-DAY PLAN WITH:
1. A one-paragraph strategy statement tied to my priority
2. Two channels maximum, chosen from what already worked —
   justify each choice against my inputs, not in general
3. A month-by-month arc: Month 1 = foundations, Month 2 = volume,
   Month 3 = double down on what's working
4. For each month: specific activities, hours each takes,
   and what it costs
5. Three numbers I check every Friday to know if it's working,
   each measurable for free in under 10 minutes
6. The assumptions you made about my market — list every one
   explicitly so I can correct them

RULES: Stay inside my budget and hours — if the plan needs more,
say so instead of pretending. No channel I haven't mentioned
unless you flag it as an experiment with a small capped budget.
No jargon. Every activity must name what I actually do that week.

You'll get back something 70% right and specific enough to disagree with. That's the point — you can't argue with "post consistently," but you can argue with "publish two posts a week targeting spring renovation searches," and arguing is where the plan gets good.

Pressure-test the draft

Don't accept the first version. Spend the next 45 minutes attacking it, in the same chat.

Start with the assumptions list you asked for in the prompt — this is the step most people skip and the one that matters most. The AI assumed things about your market, your margins, your customers' habits. Some are wrong. Correct them in plain language: "Assumption 3 is wrong, my customers are on Facebook, not Instagram" — and ask it to revise the plan accordingly.

Then run three challenges, one message each: "Which part of this plan is most likely to fail, and why?", "Cut the budget by 40% — what survives?", and "I only did the Month 1 activities and nothing else. How much of the value remains?" The answers tell you where the plan is load-bearing and where it's padding. Cut the padding.

We see this constantly in client work: the first draft of any AI plan is plausible, and plausible is the trap. The pressure-test round is what turns plausible into yours.

Break it into weekly actions

A 90-day plan you look at monthly is a decoration. The last step is converting it into thirteen weekly checklists: ask the AI to output Week 1 through Week 13, each with three to five concrete actions, the hours they take, and the Friday numbers check.

Read the weekly hours honestly before you accept them. If Week 4 assumes six hours and you gave it four in the inputs, push back in the chat — a plan that quietly overruns your time budget dies in Week 5, and you'll blame yourself instead of the math.

Put Week 1 somewhere you'll see it Monday morning. Not in a folder — taped to the wall or pinned in whatever app you actually open. The 90-day marketing plan tool in our toolbox packages this whole session — inputs checklist, master prompt, pressure tests, and the weekly breakdown — if you'd rather follow it as one guided workflow.

The honest limit: AI can't know your market

Here's what this afternoon doesn't buy you. The AI has never met your customers, doesn't know that your town's spring market starts three weeks earlier than the almanac says, and can't tell that your biggest competitor just hired two salespeople. It generates a structurally sound plan around whatever you told it — and fills every gap with generic-market assumptions.

Which means the editing step isn't proofreading, it's the actual work. If you skim the assumptions and hit go, you'll execute a fiction with confidence. Every hour you spend correcting what the AI got wrong about your market is an hour the plan gets closer to reality.

Once the plan exists, the doing gets easier too — the seven marketing workflows you can copy cover executing the content weeks, and the marketing toolbox has the tools for each. One afternoon for the plan. Thirteen Mondays to run it.