You're an Oklahoma City or Tulsa business owner. You keep hearing that AI could save you 10 hours a week, or write your marketing, or handle customer questions automatically. Every LinkedIn post is an "AI consultant" now. You don't know how to tell real from rebrand.
Here's the honest filter. What AI consultants actually do, what most of them can't, and what to ask before you sign a contract.
What "AI Consultant" Has Come to Mean
Three kinds of people are using that title right now:
1. The Rebranded Generalist
A marketer, a web dev, a coach who added "AI" to their LinkedIn in 2023 and now calls themselves an AI consultant. They know ChatGPT. They may have built one or two automations using Zapier + OpenAI. They can teach you prompts.
What they can't do: build custom AI that actually fits your business, debug why a RAG chatbot hallucinates on your docs, integrate AI into your existing software, or evaluate whether an AI pitch from a vendor is real.
2. The Agency Repackager
A digital agency that added "AI services" to its website. Someone on the team uses ChatGPT occasionally. They'll sell you "AI strategy" engagements that amount to PowerPoints. Implementation goes to a junior who'll outsource the actual coding.
What they can't do: own a technical project end-to-end. You'll spend as much time managing them as doing the work yourself.
3. The Actual Engineer
Someone who writes code. Understands how LLMs work beyond the API surface. Can ship production systems. Has shipped AI features that are running and used, not just demoed.
In Oklahoma City or Tulsa, there are only a handful. It's worth knowing which type you're talking to.
Red Flags
- "I don't code, I just prompt engineer." Prompt engineering is real, but it's 10% of building useful AI. If they can't code, they can't ship anything that integrates with your systems.
- Their portfolio is all screenshots of ChatGPT conversations. Not projects. Not running systems.
- They can't explain RAG, embeddings, or vector databases. You don't need to understand these, but they do.
- They quote $20K for a "strategy engagement" and the deliverable is a slide deck.
- They promise AI will fully automate anything. Real AI almost always augments humans, doesn't replace them.
- They can't name the models they've used or the tradeoffs between them. GPT-4 vs Claude vs Gemini isn't trivia; it shapes what you can build.
Green Flags
- They ask about your existing systems first. What software are you using? Where does data live? Who's going to use the AI tool? Because AI is only as useful as the context around it.
- They offer a paid small project before a big one. Two weeks, a few thousand dollars, real deliverable. If it works, you scale up.
- They're honest about what AI can't do. If they say "AI can do everything," they're either lying or don't know.
- They can point to things they've built that are in production. Not demos. Not screenshots. Actual running systems that real people use.
- They write or speak about AI. Blog posts, talks, code repos. Signals they think deeply about this, not just pitch it.
What to Actually Ask
Before you sign anything, put these questions in front of the consultant:
- "Show me something you've built that's running right now." Not a demo. A real deployment.
- "Walk me through how an LLM-powered app actually works under the hood." You'll learn a lot from how they answer.
- "What would you do not with AI if we worked together?" If they can't name anything AI is bad at, they're selling.
- "What happens when your AI is wrong? Who catches it? What's the blast radius?" Production AI always has this question.
- "What does a first project together look like?" Real consultants propose small, specific scopes. Rebrands propose "AI transformation."
- "What does it cost to run this after it's built?" API costs, hosting, maintenance.
If they can't answer these, they're not the right fit.
Local vs Remote: Does It Matter?
For most AI work, remote is fine. A consultant in San Francisco can build you the same thing as a consultant in OKC.
But local has real advantages:
- Same time zone. Response times matter when something breaks.
- Ability to meet in person. Especially at the start of a project, when you're aligning on what "useful AI" means for your specific business. Much easier in a conference room than over Zoom.
- Cultural fit. Oklahoma business culture is different from coastal tech. A local consultant isn't going to pitch you a 6-figure "AI-native transformation." They're going to build you something that works within your budget.
- Referrals stay in your network. Successful projects produce word of mouth. Local word of mouth compounds.
For projects over about $15K, local meaningfully matters. Under that, remote is usually fine.
What AI Work Actually Looks Like for a Local Business
Real projects I've seen in OKC / Tulsa:
- Document extraction: Law firms, insurance agencies, real estate. AI pulls structured data from PDFs and forms. Saves paralegal and admin time.
- Customer service chatbots: Trained on the company's FAQ and docs. Handles tier-1 questions. Escalates the rest.
- Internal knowledge base RAG: Employees ask questions in plain English, get answers from company documentation. Bigger lift than a chatbot, bigger payoff.
- Marketing content drafting: AI drafts social posts, email campaigns, blog outlines. Marketer edits and publishes. 10x throughput.
- Sales prep automation: AI summarizes prospect companies, drafts personalized outreach, flags likely interest. Sales rep reviews and sends.
- Internal ops automation: Excel/Google Sheets work, weekly reports, data consolidation. Unsexy but the highest ROI for most small businesses.
These are the AI projects that make money. Not the ones you see on Twitter.
How Much Should AI Work Cost?
Honest ranges for a business in OKC or Tulsa:
| Project type | Typical cost |
|---|---|
| Quick automation / Zapier + LLM setup | $1K–$5K |
| Custom chatbot on your docs | $5K–$15K |
| Document extraction pipeline | $5K–$20K |
| Internal RAG system for knowledge base | $15K–$40K |
| Custom AI embedded in existing software | $20K–$75K |
| Full AI platform or product | $75K+ |
Ongoing monthly costs (API + hosting + maintenance) typically run 10–30% of the build cost annualized. Plan for it.
What I Actually Do
I'm an AI engineer based in Oklahoma City. I've built production AI systems—intake copilots, anomaly detection, forecasting models, AI-generated reporting—including a collision repair analytics platform running across 100+ shops.
When someone in OKC or Tulsa asks me about AI, the first conversation is always about whether AI is even the right tool. Sometimes it's not. A well-designed spreadsheet or a $100/month SaaS product beats a $20K AI project for a lot of problems.
If AI is the right tool, the second conversation is scope. Start small. Ship something real. See if it works. Then scale up.