Why I Turn Down Some AI Projects

I've turned down more AI projects than I've accepted. Here's what those rejections taught me about what actually works for Central Florida small and mid-market businesses — and what doesn't.

I got a call last spring from a logistics company in Sanford. The owner, Dave, had heard about AI voice agents and wanted to replace his entire customer service team of eight people. He’d already budgeted $15,000 for the project. He was ready to sign that day.

I told him no.

It wasn’t the first time I’d turned down work, and it won’t be the last. Over the past two years, I’ve said no to roughly half the AI projects that come across my desk. Some of those decisions surprised the business owners. A few even upset them. But every rejection taught me something about what seperates a useful AI implementation from an expensive mistake.

If you’re a business owner in Central Florida thinking about using AI — maybe you’ve heard about it from a friend, a conference, or a vendor — I want you to understand why I say no. Because the reasons I reject projects are probably the same reasons your own AI plan might need a second look.

1. The Problem Isn’t Actually a Problem

The most common reason I turn down a project is simple: the business doesn’t have a real problem worth solving with AI.

A few months ago, a real estate agency in Winter Park reached out. They wanted an AI chatbot for their website to answer questions from potential home buyers. Sounded reasonable. But when I asked what questions those buyers were actually asking, the owner couldn’t tell me. She just knew that other agencies had chatbots and felt like she was falling behind.

I asked her to track incoming calls for two weeks. What’d we find? Three calls per day from buyers on average, and most were just scheduling showings. The office manager handled them in under five minutes each. Total daily time spent: maybe 15 minutes. Building and maintaining a chatbot to save 15 minutes a day — when the office manager had plenty of capacity — didn’t make financial sense. I told her to save her money.

That’s the first question I ask every potential client: What specific, measurable problem are you trying to solve? And I mean actual numbers. Not feelings. Not trends. Not a vague sense that you should be doing this. If you can’t answer that with data, you’re not ready to buy AI yet. You’re ready to do some thinking first.

2. The Data Is a Mess

AI runs on data. Not hope. Not good intentions. Not a vague sense that you’ve got alot of customer information somewhere. I’m talking about clean, structured, accessible data.

A medical billing company in Lake Mary came to me wanting to use AI to automate claims processing. They had years of claim forms, denial letters, and payment records. The owner was frustrated by how many claims got rejected and how long it took to fix them. He wanted an AI that could learn from past denials and automatically correct future submissions.

Great idea on paper. But when I looked at their data, it was a disaster. Claim forms were scanned PDFs with handwritten notes scattered throughout. Denial letters came from dozens of different insurance companies, each with different formats, codes, and language. Some records lived in a legacy database that nobody fully understood anymore. There was no consistent way to even tell the AI what a denial looked like.

I told him that before we could build any AI solution, we’d need three to six months cleaning and standardizing the data. That wasn’t what he wanted to hear. He wanted a quick fix. Here’s the thing, though: I don’t do quick fixes with bad data, because they don’t work. The AI would just automate the mess — faster, but still wrong.

If your data lives in spreadsheets with inconsistent columns, handwritten notes, or siloed systems that don’t talk to eachother, fix that first. AI can’t fix bad data. It only amplifies it.

3. The Expectations Are Fantasy

Sometimes the problem isn’t the technology or the data. It’s what the business owner believes AI can actually do.

I met with a restaurant group in downtown Orlando that wanted an AI to handle all their social media marketing, generate new menu items, and predict which dishes would be most profitable. By next month. They’d seen demos of AI writing posts and creating recipes, and they figured it would just happen automatically.

I walked them through what realistic AI implementation looks like for a restaurant. Could you use AI to analyze past sales data and suggest which menu items to feature? Sure. Could you automate responses to common customer questions on social media? Absolutely. But generating entirely new menu items that account for ingredient costs, prep time, local supplier availability, and customer preferences? That’s a research project, not a plug-and-play tool. And expecting it in a month? Not happening.

The owner was disappointed. Some software vendor had told him AI could do everything. I told him that vendor was selling a dream, not a solution. Look, if you hear promises like “set it and forget it” or “works out of the box with your existing systems,” be skeptical. Real AI projects take time, testing, and iteration. Anyone who tells you otherwise is either naive or dishonest.

4. The Business Isn’t Ready for Change

This is the hardest one to spot, and honestly, it’s the most common reason I walk away.

A construction company in Apopka wanted AI to automate their project estimating. They had three estimators who put together bids for commercial projects. The process was slow, error-prone, inconsistent. AI could definitely help there.

But when I talked to those estimators, they were openly hostile to the idea. They saw AI as a threat to their jobs. The owner hadn’t communicated why he was considering it — hadn’t explained that the goal was to let them focus on more complex bids, not replace them. There was no training plan, no roadmap for adoption. The culture wasn’t ready.

I told the owner that if we deployed AI without addressing their concerns, we’d get resistance, sabotage, and ultimately failure. The technology would work fine. The people would reject it. I offered to help him build a change management plan first, but he wanted to move fast. I declined.

AI adoption is 20% technology and 80% people. If your team isn’t on board, if you haven’t explained the why, if there’s no plan for how roles will evolve — you’re not ready. And that’s okay. It just means you need to do the human work first.

I’ve turned down more AI projects than I’ve accepted. The common thread? The business didn’t have a clear problem, clean data, realistic expectations, or a team ready for change. Fix those first, and the technology becomes simple.

5. The Vendor Has Locked Them In

Every now and then, a business comes to me after they’ve already bought something. A salesperson convinced them to sign a contract for an AI tool that promised the moon. Now they’re stuck with a system that doesn’t work and a vendor who won’t let them out.

A property management company in Heathrow had purchased a year-long subscription to an AI-powered leasing assistant. The software was supposed to handle tenant inquiries, schedule tours, and screen applicants. It did none of those things well. It misunderstood basic questions, scheduled tours for the wrong times, flagged every applicant as high-risk. The company was spending $2,000 a month on a tool that made their work harder.

When they called me, I looked at the contract. Auto-renewal clause. 90-day cancellation notice. They were locked in for another nine months. I couldn’t help them with that tool. I could only advise them on what to look for next time.

If you’re considering an AI vendor, read the contract carefully. Look for long-term commitments, data portability clauses, and what happens if you want to switch. The best AI tools are the ones you can walk away from. If a vendor won’t let you leave, that’s a red flag.

6. The Project Doesn’t Need AI at All

Sometimes the best solution isn’t AI. It’s a simpler tool, a process change, or just doing the work differently.

A boutique hotel in Mount Dora wanted an AI system to personalize guest experiences — recommend activities, adjust room settings, send tailored offers. Sounded fancy. But when I looked at their operation, they had 12 rooms. The owner knew every guest by name. She already asked them what they liked and made recommendations over breakfast. The personalization was happening naturally, without any technology at all.

I told her that AI would add complexity without adding value. What she actually needed was a better booking system and a simple email automation tool to send follow-ups after stays. Total cost: a few hundred dollars a month. She saved thousands.

Not every problem needs an AI solution. Sometimes a spreadsheet, a checklist, or a better process is all you need. Don’t let the hype convince you otherwise.

What These Rejections Mean for Your Plan

If you’re reading this and thinking about AI for your business, I hope you see that my rejections aren’t about the technology being bad. They’re about the conditions not being right. And the good news? You can create those conditions.

Start by asking yourself the hard questions: What problem am I really solving? Do I have clean data? Are my expectations realistic? Is my team ready? Have I looked at simpler alternatives? Am I locked into a bad vendor?

If you answer those honestly, you’ll either find that you’re not ready — which is fine — or you’ll find that you’ve got a clear path forward. And if you want help thinking through those questions, I offer a free AI Readiness Assessment that walks through exactly these issues. It’s not a sales pitch. It’s a diagnostic. Most people who take it realize they need to do some prep work first.

I also offer a Fractional AI Officer service for businesses that want ongoing guidance without hiring a full-time executive. And if you’re curious about the terminology, my AI Glossary explains the common terms in plain English.

When you’re ready to move forward, I can help with specific implementations like AI voice agents or Microsoft 365 Copilot rollouts. But only if the conditions are right.

Because sometimes the smartest thing I can do for a business is say no. And the smartest thing you can do for yours is listen.

I've turned down more AI projects than I've accepted. The common thread? The business didn't have a clear problem, clean data, realistic expectations, or a team ready for change.

Frequently asked questions

Why would an AI consultant turn down a project?

I turn down projects when the business doesn't have a clear problem to solve, the data is too messy, expectations are unrealistic, the team isn't ready for change, the vendor contract is too restrictive, or the solution doesn't actually need AI. Taking on a project that's likely to fail doesn't help anyone.

How do I know if my business is ready for AI?

Start by asking: Do I have a specific, measurable problem? Is my data clean and accessible? Are my expectations realistic? Is my team on board? Have I considered simpler alternatives? If you can answer yes to all of those, you're probably ready. If not, focus on the gaps first.

What should I do before buying an AI tool?

Define the problem with numbers, clean up your data, talk to your team, and research alternatives. Read the vendor contract carefully — watch for long-term commitments and data lock-in. Consider starting with a small pilot before scaling.

Can AI work with messy data?

No. AI amplifies whatever data you give it. If your data is messy, AI will just produce messy results faster. Clean and standardize your data before implementing any AI solution.

How do I handle employee resistance to AI?

Communicate early and often. Explain that the goal is to help them, not replace them. Involve them in the planning process. Provide training and a clear roadmap for how their roles will evolve. Change management is often harder than the technology itself.

What's the biggest mistake businesses make with AI?

Buying AI because it's trendy, not because it solves a real problem. That leads to wasted money, frustrated teams, and abandoned projects. Always start with the problem, not the technology.