The AI marketing stack we actually use to run marketing for small businesses
Ask AI to Summarize
Seven tools, what they cost, and the way we use each one for roofers, med-spas, realtors and property managers. No hype, no affiliate-padded list — just what’s open on our screens every day at Kihan.
If you run a small business, you’ve probably been told to “use AI for marketing” by someone who can’t tell you which AI, for what, or why. We can. At Kihan we run marketing for clients across seven industries — roofing, med-spa, real estate, property management and a few others — and we’ve spent the last while finding out which AI tools actually move the needle and which are just expensive distractions. This is the stack we pay for in 2026.
Who this is for: small business owners, in-house marketers, and agency operators who want a real-world picture of what AI marketing looks like in practice. Not a tool-review listicle. We get nothing if you sign up for these — we just use them.
TL;DR — the stack at a glance
| Tool | What we use it for | Category |
|---|---|---|
| Claude | Strategy, long-form drafts, complex analysis | AI assistant |
| ChatGPT | Quick drafts, brainstorm, image briefs | AI assistant |
| Firecrawl | Competitor scraping, local SEO research at scale | Web research |
| Perplexity | Cited research, market sizing, fact-checking | Research |
| Midjourney / Imagen | Ad creative, social photography, hero images | Visuals |
| SocialBu | AI-assisted social scheduling across platforms | Social |
| Custom AI workflows | Our own repeatable client workflows (Antigravity Kit) | Internal |
Why a small business should care about AI marketing in 2026
Three things changed in the last eighteen months that make AI genuinely worth the time for a small business — not just a side project for the marketing nerds in the room.
First, the cost of producing good content collapsed. A blog post that used to take a freelancer four hours and three hundred dollars can be drafted in fifteen minutes for under a dollar of AI usage. The skill shifted from writing to editing, and from production to judgement — what to publish, for whom, and how to make it actually useful.
Second, competitor research stopped being a moat. Anyone with the right tools can scrape every page of every competitor in their city in an afternoon and find the gaps in fifteen minutes. Local businesses that haven’t done this are losing leads to the ones that have.
Third, ad creative testing got cheap. You can generate twenty headline variants and ten image variants for a Meta campaign before lunch. The bottleneck moved from “how do we make more creative” to “which creative actually converts” — and that’s a much better problem to have.
TOOL 01
Our default brain for anything that needs to be thought about, not just produced.
- Strategy & writing
- Free tier + paid plan
- Used daily at Kihan
Claude is what we open first. It handles client strategy briefs, long-form blog drafts, audit reports, and the awkward conversations where we need to explain a recommendation in three different ways for three different stakeholders. The output reads like a thoughtful colleague rather than a template, which matters when you’re writing for a homeowner researching a roof or a patient researching laser hair removal.
How we use it
Drafting long-form FAQ and education pages for med-spa clients — sourcing real questions from review sites, asking Claude for clinically considered explanations, then editing for the brand voice. The result reads like a thoughtful clinician, which is exactly what a homeowner or patient is looking for.
| Best for | Long-form content, strategy memos, client-facing writing |
| Skip if | You only need quick one-liners or image generation |
| Get it | claude.ai |
What we love
- Reads like a senior writer, not an intern
- Holds long context for full briefs
- Honest when it doesn't know
Trade-offs
- No native image generation
- Slower than ChatGPT for short tasks
TOOL 02
The Swiss Army knife — fast, ubiquitous, and good at the small stuff.
- Quick drafts & ideation
- Free tier + paid plan
- Image gen included
We use ChatGPT for the high-velocity work — meta descriptions, email subject lines, social captions, image briefs, ad headlines. The native image generation is good enough for thumbnails and first-pass concepts. For local SEO work, the ability to ask “rewrite this in plainer English for a homeowner in Austin” without breaking flow is what keeps it in the stack.
How we use it
Generating ad headline and primary-text variants for Meta campaigns — for example, a roofing client’s storm season push. Brief the angle once, pull a batch of variants, ship the strongest ones to test. The point is volume of creative, not a single perfect line.
| Best for | High-volume short copy, image briefs, quick reformatting |
| Skip if | You need consistent long-form voice or careful reasoning |
| Get it | chatgpt.com |
What we love
- Built-in image generation
- Fastest mainstream model on short tasks
- Massive plugin ecosystem
Trade-offs
- Voice can feel generic without heavy prompting
- Quality drift across model versions
TOOL 03
Turns any competitor’s website into a spreadsheet you can actually use.
- Research & scraping
- Free tier + paid plans
- Underrated
If you’ve ever needed to know every service page, every location page, every blog post across a handful of competitor websites — Firecrawl is the unfair advantage. It crawls a domain, returns clean markdown, and feeds straight into Claude or ChatGPT for analysis. The kind of audit work we used to outsource, we now run internally.
How we use it
Building competitor content gap reports for real estate clients — scraping every neighbourhood and listing page a set of competitors ranks for, then handing the clean markdown to Claude to spot which high-intent searches no one has covered well. The output becomes the next quarter’s content plan.
| Best for | Competitor audits, local SEO gap analysis, lead enrichment |
| Skip if | You only manage one site and have no competitors worth studying |
| Get it | firecrawl.dev |
What we love
- Clean markdown — works with any AI tool
- Handles JavaScript-rendered pages
- Generous free tier to start
Trade-offs
- Higher tiers needed for full-site crawls
- Steeper learning curve than no-code tools
TOOL 04
Research with sources you can actually verify and link to.
- Research
- Free tier + paid plan
- Trust-worthy
Claude and ChatGPT are great at writing but bad at telling you where they got their facts. Perplexity flips that — every answer comes with citations you can click. We use it for market sizing, regulatory questions (especially med-spa work where the rules differ by state), and any time a client asks “is this still true in 2026?” before we publish.
How we use it
Checking regulatory questions that come up in client work — for example, which marketing claims a med-spa can legally make under their state’s cosmetology board rules. Perplexity returns the relevant guidance with sources we can verify before anything gets published.
| Best for | Competitor audits, local SEO gap analysis, lead enrichment |
| Skip if | You only manage one site and have no competitors worth studying |
| Get it | firecrawl.dev |
What we love
- Real citations, every answer
- Fast for live news and recent data
- Good free tier
Trade-offs
- Writing quality is utilitarian
- Some answers still need a human sanity check
TOOL 05
Ad creative and hero imagery that doesn’t look like every other AI-generated image.
- Visual / creative
- Paid plans
- Two-tool combo
We pair both because they’re good at different things. Midjourney gives us cinematic, editorial-feel imagery for hero sections and brand work. Google Imagen is better at clean product shots, infographics, and the kind of “trustworthy local business” photography you need for service pages. The mistake most small businesses make is using a single tool for everything — the outputs all look the same and Google starts to notice.
How we use it
Producing visual identity assets for clients — like a warm bronze palette and sunlit terrace photography for a Caribbean real estate brand. Lifestyle shots that read aspirational without crossing into stock-photo territory, generated against a brand mood board so the output is consistent across the campaign.
| Best for | Ad creative, hero images, social posts, mood boards |
| Skip if | You can hire a photographer and need rights-cleared, real-people imagery |
| Get them | midjourney.com · Google Imagen |
What we love
- Far cheaper than a photo shoot
- Brand control through style references
- Fast iterations
Trade-offs
- People still look slightly off in close-ups
- Some industries (healthcare especially) require real photography
TOOL 06
AI-assisted scheduling that doesn’t punish you for having seven platforms.
- Social media
- Paid plans
- Used for Kihan's own brand
We use SocialBu to run Kihan Marketing’s own social presence. It schedules across Facebook, Instagram, LinkedIn, X, TikTok and Threads from a single calendar, and the AI features can draft captions or expand a single post into a full week of variations. For client work we use different stacks per industry, but for our own brand this is the one we pay for.
How we use it
Repurposing a single long-form blog post into a multi-week social calendar — short carousels for LinkedIn, vertical clips for TikTok, threaded posts for X — scheduled from one screen. The AI caption assist turns one piece of pillar content into a string of platform-native ones.
| Best for | Multi-platform scheduling, repurposing one piece into many |
| Skip if | You only post on one platform and prefer native tools |
| Get it | socialbu.com |
What we love
- One calendar across every platform
- AI caption assist saves real time
- Fair pricing for small teams
Trade-offs
- Analytics aren't as deep as Sprout or Hootsuite
- TikTok integration occasionally flaky
TOOL 07
Custom AI workflows (our internal stack)
The unfair advantage you can’t buy on a SaaS pricing page.
- Internal tools
- Built in-house
- Antigravity Kit
The biggest gain we’ve seen isn’t from any single tool — it’s from stringing them together into repeatable workflows. We built an internal kit (we call it Antigravity Kit) that handles the parts of marketing that look the same every time: city-specific landing pages for service businesses, FAQ generation, brand-consistent design tokens, the works. A workflow that used to take an account manager two days now takes a few hours. That’s where margin comes from in an agency, and it’s where leverage comes from for a small business doing its own marketing.
How we use it
Generating city- and neighbourhood-specific landing pages for service businesses at scale — each one rewritten to be genuinely useful for that area rather than template-spun. The work that used to be the slowest and most expensive part of a local SEO engagement is now the easiest part of the workflow.
| Best for | Anything you do more than three times — turn it into a workflow |
| Skip if | You’re still in the “trying one AI tool” phase — come back later |
| Want one built? | Book a free audit |
How the seven tools fit together
None of these tools matter on their own. The leverage comes from the workflow that connects them. Here’s roughly how a single piece of work moves through our stack — for example, building a content engine for a new med-spa client.
Step 1 — Research. Perplexity gives us the regulatory and clinical context. Firecrawl pulls every competitor’s service pages in the city. We feed both into Claude.
Step 2 — Strategy. Claude analyses the gap — which services are over-supplied, which are under-served, which questions no one is answering well. We get a content map in under an hour.
Step 3 — Production. ChatGPT drafts the short stuff (meta tags, ad headlines). Claude drafts long-form pages. Both reference the brand voice file we built once at the start of the engagement.
Step 4 — Visuals. Midjourney and Imagen generate hero imagery and ad creative against a brand mood board.
Step 5 — Distribution. SocialBu schedules the social cuts across platforms. Custom workflows handle the city-specific landing-page variants.
The whole loop runs weekly. The client sees more output for less money than any other approach we’ve found. The market hasn’t fully priced this in yet — the small businesses that adopt it now will pull ahead of competitors who are still doing everything by hand.
Should you DIY this or hire an agency?
Honest answer: it depends on three things.
Time. Building this stack in-house takes consistent effort before it pays back. If you don’t have focused weekly time to invest in learning it, hiring an agency that already runs the stack will usually be cheaper than the cost of figuring it out yourself.
Volume. If you publish a piece of content occasionally, you don’t need a full AI stack — Claude or ChatGPT on their own is fine. If you’re trying to dominate a local search market with a wide library of pages, you need a workflow, not just tools.
Judgement. AI is excellent at production and bad at strategy. If you have strong opinions about your market and your customers, you can run this yourself with the tools above. If you’re still figuring out positioning and messaging, that’s where an outside perspective pays for itself quickly.
Want us to run this stack for your business?
We work with roofers, med-spas, realtors, property managers and four other local-service industries. Free 30-minute audit — we’ll look at your site, your competitors, and the three biggest gaps an AI marketing stack could close for you.
Frequently asked questions
Yes — but only if you pick two or three tools, use them every week, and connect them to a real outcome like leads booked or pages ranked. AI is genuinely useful for content drafting, competitor research, ad creative, and social scheduling. It is not yet useful for strategy or judgement calls, which is where most small businesses still need a human.
Claude or ChatGPT — both have a useful free tier and an affordable paid plan. Start there before adding a second tool. You will get more leverage from one tool used daily than from a shelf of tools you barely touch.
It will replace agencies that only do production work — writing posts, building basic pages, running cookie-cutter ads. It will not replace agencies that build strategy, judge what to test, manage budgets, and own a measurable result. The honest test: ask your agency what they do that AI cannot.
The tool subscriptions themselves are a small fraction of what most businesses already spend on marketing. The bigger cost is the time to learn them and the discipline to use them every week. Content and creative get cheaper to produce, which is usually where the return shows up first.
Google penalises unhelpful content, not AI content. Pages that solve a real problem, cite real sources, and reflect real expertise rank fine whether a human or an AI wrote the first draft. The difference is editing — AI gets you to a draft in minutes, but a human still needs to add experience, examples, and a point of view.
Any local service business with a sales cycle longer than a phone call — roofers, med-spas, realtors, property managers, dentists, contractors. The leverage comes from being able to produce localised content at scale, run smarter ad creative tests, and respond to leads faster than competitors who are still doing everything manually.
Kiran Dasgupta
Founder, Kihan Marketing · Seattle, WA
Kiran founded Kihan Marketing to bring AI-leveraged marketing to local service businesses that can't afford a big-agency retainer. He writes about marketing operations, SEO at scale, and what's actually working in the field.
