← Back to all free guides
A free starter pack by Mariah Brunner ↗ INSTAGRAM ↗ TIKTOK

Become The
AI Consultant
SMBs Actually Call

OpenAI just launched a $4 billion company to send its engineers into big companies. Every business below them still needs help and has no one to call. Here’s the 5-step playbook to step into that gap, with the exact prompt to run at every step.

In May, OpenAI launched the OpenAI Deployment Company with $4 billion to embed its engineers inside large enterprises. The dentist, the HVAC company and the three-agent brokerage down the street aren’t getting that call. They need someone who speaks their language and can set up AI that saves them real hours.

That someone can be you. Five steps, one prompt each, and you can run the first one today.

01 Pick One Niche

Don’t be the AI consultant for small businesses. Be the AI consultant for real estate brokerages in Texas, or dental practices with one to three locations, or local HVAC companies. The narrower the niche, the faster people trust you, because you speak their language from day one.

The right niche is one where you already understand the work. A past job, a past industry, a family business, a hobby you obsess over. Context beats expertise.

01

The niche finder

Run it once, screenshot the output, and pick the niche that makes you nod three times.

Niche Finder
You are my AI consulting strategist. Help me pick the ONE specific small-business niche where I'll build my AI consulting practice. This is the most important decision I'll make this year, so be honest, not flattering. My background (the more specific, the better): - Jobs I've held in the last 10 years (title, industry, what I actually did): [LIST] - Industries I've sold to, supported or built for: [LIST] - Family or close friends' businesses I've watched up close: [LIST] - Hobbies or side projects where I go deep: [LIST] - Communities I'm already in (industry groups, Slack, associations): [LIST] - Where I have an on-the-ground network: [PLACES] - Credentials or licenses: [LIST] - My current main income: [DESCRIBE] Do five things. 1. List 10 candidate niches where I have real context, ranked by the strength of my edge, with one line of reasoning each. If my "edge" is thin (I worked in retail 8 years ago), rate it that way. 2. For my top 5, list the 5 most tedious weekly workflows a typical owner in that niche does. Be specific ("following up with leads who ghosted after a quote", not "marketing"), and rate each 1-10 on how much AI could take off their plate. 3. For my top 5, estimate typical business size and whether they could realistically pay a flat install fee plus a monthly retainer. Search for what consultants or agencies serving that niche publicly charge and link what you find. If you can't find real numbers, say so instead of guessing. 4. Pick the 2 niches with the best mix of my edge, clear AI workflows, and businesses big enough to pay. Defend each in 3 sentences. If neither hits all three, tell me what to do about it. 5. For each finalist, write my one-line positioning: "I help [exact niche] use AI to [specific outcome] without [friction they know]." End with one paragraph of pushback: name anything in my background that's wishful thinking, and any niche I didn't list that I'd be better off in. Voice: warm but ruthless. I'd rather hear it now than spend a year in the wrong niche.
02 Build Real Fluency

You’re not becoming an ML engineer. You’re becoming the most AI-fluent person in your niche. Learn in this order:

  1. Daily reps. Use Claude or ChatGPT for your own work every day.
  2. Prompting. Anthropic’s free prompt engineering docs and Claude Academy courses are some of the best free resources out there.
  3. Connectors. Get Claude reaching into Gmail, Calendar, and the CRMs and accounting tools your niche uses.
  4. One no-code tool. Make, Zapier or n8n, for clients whose tools have no Claude connector.

You don’t need a paid course. The docs plus Claude itself get you most of the way there.

02

My 12-week fluency plan

Paste this once your niche is locked. Claude builds a week-by-week plan so you stop guessing what to study next.

Fluency Plan
You are my AI fluency coach. Build a weekly plan that takes me from where I am to confident, paid-consultant fluency in [MY NICHE] in 12 weeks. No theory I won't use. My inputs: - My niche and positioning line: [PASTE FROM STEP 1] - Hours I can really commit per week: [NUMBER] - Current level: [beginner / comfortable with Claude or ChatGPT / already building skills and connectors] - Tools I already pay for: [Claude / ChatGPT / Zapier / Make / n8n / none] - What intimidates me most: [e.g. "I've never written code", "pricing"] - What would prove I'm ready: [e.g. "I could confidently pitch a paid install if..."] Build the plan in five phases. Scale the workload to my hours. PHASE 1, REPS (weeks 1-2): the 5 prompting skills to drill and a 30-minute daily habit. Point me to Anthropic's free prompt engineering docs and Anthropic Academy courses, and say which to do first. PHASE 2, NICHE IMMERSION (weeks 3-5): the 20 most tedious recurring workflows in my niche. For each: one sentence, a rough prompt that would handle it, the tools or connectors involved, and a 1-10 automation score. By the end I should be able to pitch any of them in 60 seconds. PHASE 3, BUILD (weeks 6-8): the 3 workflows with the best pain-to-effort ratio. For each: what to build (skills, connectors, scheduled tasks, projects, plus any outside tool), in what order, and the metric the client would care about. By week 8 one of them runs end to end in my own Claude. Only name Claude features and connectors you're sure exist; if unsure, tell me to check the connector directory. PHASE 4, PROOF (weeks 9-10): a script for a 90-second screen-recorded demo of my workflow, and a one-page case study template for my first pilot client. PHASE 5, SHIP (weeks 11-12): 5 specific outreach actions a week to land my first free pilot: where my niche hangs out online and what to post, a LinkedIn DM script, one coffee a week (and how to find the right person), and a free workflow audit offer. End every week with 3 small deliverables, a 5-question self-check, and a "behind schedule" minimum version. Voice: a coach who's done this, rooting for me, who won't let me kid myself.
03 Build One Install

Don’t sell consulting hours. Sell an install. It’s the same workflow at the same price for every client. Hourly billing caps your income and punishes you for getting faster.

Five workflows almost every small business needs:

  1. Lead response and qualification. Every new lead gets a fast, personal reply and lands with the right person. Clients see the money fastest here.
  2. Customer service triage or front-desk chat.
  3. CRM cleanup and lead scoring.
  4. A content and SEO engine.
  5. AI-drafted vendor and operations emails.

03

Design my one install

Pick one workflow from the list and run this. The output is your installable product, ready to demo.

Install Designer
You are my AI install product designer. Take one workflow and design a repeatable install I can sell over and over to businesses in [MY NICHE]. Same scope, same price, same deliverables every time. My inputs: - Niche: [PASTE] - The workflow: [e.g. lead response / customer service triage / CRM cleanup / content engine / operations emails] - What I know about the tools this niche uses (CRM, phones, email, booking software): [LIST OR "STILL FIGURING OUT"] - My build skill level: [beginner / intermediate / advanced] Build the product spec in seven sections. 1. BEFORE AND AFTER: one vivid paragraph on how the business handles this today, and one on how it runs with my install. Use the owner's language, not consultant-speak. 2. DELIVERABLES: a concrete list (for example: one Claude project, two or three skills, the connectors it uses, an alert for high-value leads, a written SOP for the team, a 60-minute training). 3. TECHNICAL BUILD: step by step, in order. Only use connectors that really exist in Claude's directory (tell me to confirm in the directory before I promise anything). If a tool has no connector, give the real workaround (a scheduled export, Zapier or Make, or pasting data in). 4. CLIENT ONBOARDING, DAY 0 TO 14: kickoff agenda, the 5 things I need from the client, build and test, a dry run with the owner watching, go-live and training. 5. METRICS AT DAY 30, 60 AND 90: three measurable numbers per check-in (for example, average response time to a new lead) and one qualitative win. 6. ONE-PAGE PROPOSAL COPY: their problem in their words, what I'm installing, the 30/60/90 outcomes, the price (install fee plus monthly retainer, left as [PRICE] for now), my guarantee if I offer one, and the next step. 7. WHAT WILL TRIP ME UP: three predictable problems on install #1 and the workaround for each. Include data privacy: what client data the install touches and what plan or settings the client needs. Keep it simple enough to deliver the same way every time. Repeatability is the whole point.
04 Land Your First 3 Clients

You need case studies to get clients, and clients to get case studies. The fix is three installs in about 60 days. Three is where your story shifts from "I did this once" to "I do this."

Here’s the ladder. Do your first one free for a friend or local business in your niche, in exchange for real numbers, a quote and an intro. Discount the second. Charge full price for the third. Run each as a 30 to 60 day pilot on one workflow, never "AI for your whole business."

04

Turn the install into a case study and outreach

Run this the day install #1 wraps. You get the case study, a LinkedIn post, a cold email, a demo script and your new pricing from one conversation.

Case Study Engine
You are my case-study writer and outreach strategist. Turn one finished install into the engine that brings me clients #2 and #3 in the next 30 days. My inputs: - The install: [WORKFLOW + NICHE + CLIENT TYPE] - The client (named, with permission, or anonymized): [e.g. "a two-location dental practice in Austin"] - Real numbers from the project: - Hours saved per week: [NUMBER] - Money saved or earned per month: [NUMBER OR "not tracked"] - Response time: [BEFORE] to [AFTER] - Errors or misses: [BEFORE] to [AFTER] - Owner quote, exact words: [QUOTE] - What the client didn't love: [DESCRIBE OR "NONE"] Build five deliverables. 1. ONE-PAGE CASE STUDY (under 400 words): client snapshot, the problem in their words, what I installed, one big number up front, three supporting numbers, the owner quote, and what we're doing next. 2. LINKEDIN POST (under 200 words): open with the specific number, three short paragraphs (situation, install, result), and a soft call to action for other [NICHE] owners to DM me. No emojis, no "follow for more". 3. COLD EMAIL TO 10 PROSPECTS (under 100 words): a subject line with the real number, who I am in one line, the outcome in one line, what I'd do for them, and one clear ask (a 15-minute call). 4. 60-SECOND DEMO SCRIPT, in 4 beats: the before in the client's words (0-10s), the install live (10-30s), the after with numbers on screen (30-45s), the call to action (45-60s). 5. PRICING FOR CLIENT #2: whether to raise my install fee and retainer now that I have proof, with two reasons, and the exact sentence I'll use to quote it. Rules: use only the numbers I gave you. Never round up or invent results. If a number is missing, leave a [BLANK] and tell me to get it.
05 Price Like An Installer

Stop quoting hourly. Charge a flat install fee plus a monthly retainer for keeping it running and improving it. Set the number from the value you create, the hours saved and leads closed, and raise it each time you add a case study.

Once you have several installs with clear results, you can start tying part of your fee to measured outcomes. Earn that by shipping installs first.

05

Build my pricing and proposal

Run this before your first (or fifth) paid proposal. You get the price with the math behind it, the proposal, objection replies and the follow-up email.

Pricing And Proposal
You are my pricing strategist and proposal writer. Recommend the right price for this client and write the proposal I'll actually send, plus objection replies and the follow-up. My inputs: - Niche: [PASTE] - The install I'm selling: [WORKFLOW FROM STEP 3] - Case studies so far: [NUMBER, with a one-line outcome each] - This client: [NAME, ROUGH REVENUE, EMPLOYEES, HOW THE DEAL CAME IN] - Their pain from the discovery call, in their words: [DESCRIBE] - Budget they mentioned: [AMOUNT OR "NOT DISCUSSED"] - Decision timeline: [WHEN] - Who else decides: [PARTNER / SPOUSE / CO-FOUNDER / NONE] - My floor (the lowest I'll go): [AMOUNT] Build five deliverables. 1. RECOMMENDED PRICE: install fee and monthly retainer. Defend each with (a) where I am in proof (case-study count) and (b) the value math for this client, showing your work ("they get 80 leads a week; if 5 more close at $X each, the install pays for itself in N weeks"). Say whether to offer a payment plan. 2. ONE-PAGE PROPOSAL (under 600 words): their problem in their words, what I'm installing, the 30/60/90 outcomes with named metrics, price and payment terms, timeline with go-live date, my guarantee if I offer one, and one next step with a date. 3. OBJECTION REPLIES: the three objections this client is most likely to raise (at least price, "will this work for us?", and "we're slammed, can we wait?"), with a 2-sentence reply each that sounds like me on a call. 4. 48-HOUR FOLLOW-UP: three sentences plus a subject line, warm and direct, confident enough to walk away. 5. WALK-AWAY LINE: if they push below my floor, two sentences. The first acknowledges their constraint. The second offers to cut scope, so the price drops but my rate doesn't. Voice: warm, confident, direct. Not stiff consulting-firm, not bro marketer. Use only the numbers I gave you or math you show.

One Last Thing

If you only do two of these, do Step 1 today and Step 4 within 60 days. Picking the niche and shipping the first install are the moves that change your trajectory. Everything else is preparation.