Everyone’s racing to “do something with AI.” But not every AI play pays off.

Pressure from investors, competitors, and the media has pushed many companies to act fast. But urgency without clarity burns cash. AI introduces real risk: organizational change, compliance exposure, and data dependency (whose data is perfect?). For leaders operating under margin pressure, that is not a side bet worth making.

Does AI for your business really matter?

Yes, AI matters. But the real question is: how do B2B businesses turn “AI” into revenue? In my view, there are only three options:

1. Operationalize it. To make your company faster and leaner.
2. Sell it. As in, build an AI product or service to deliver to customers.
3. Show up in it. Make sure large language models (LLMs) like ChatGPT, Gemini, and Perplexity recommend your business when buyers search.

(A fourth is emerging, that I call “Shape it,” which means training or influencing AI systems using your expertise. But that is just an advanced version of showing up. Please feel free to comment if I’m missing another!)

All three have value, but only one offers a realistic path to compounding revenue. And that’s #3. Here’s why:

1) Operationalizing AI improves efficiency but rarely drives growth

A lot of companies jump into AI projects thinking they’ll unlock quick wins, but most stall out before they make a real impact. RAND found that over 80% of AI initiatives collapse before full deployment, nearly twice the failure rate of standard IT projects. MIT found 95% of generative AI pilots show no measurable profit impact because they never make it into daily work.

That means efficiency gains rarely reach the P&L (at least in the short term). There are plenty of cautionary tales when it comes to using AI inside operations:

• Automating support emails, sales outreach, or ad personalization can save time, but one off-brand line or awkward message can lose buyer trust.
• AI-driven forecasting sounds impressive, but with incomplete data, it produces incorrect answers more quickly.
• Internal tools built on company data can backfire if confidential information isn’t handled carefully.
• Automated workflows or sensor-triggered systems sound great (and I’m a huge fan of tools like Zapier), but they’re fragile. When a model changes or a trigger breaks, the whole process stops.

Even when AI “works,” mistakes can multiply quickly. A small error can quickly escalate into a big problem. The takeaway is simple: human oversight still matters. Add in the time it takes to retrain teams and rebuild workflows, and you’re looking at a long-term investment, not a quick win.

Operational AI belongs in the roadmap, just not the revenue forecast.

2) Building and selling AI is expensive and slow

Creating an AI product (or service) feels innovative, but beware of the capital trap. To compete, you most often need:

• Engineers with deep model experience, most of whom are already working at big tech firms.
• Clean datasets (that few companies truly have).
• Serious computing power and constant model retraining as AI evolves.

Meanwhile, as you build Google, Microsoft, NVIDIA, OpenAI, Meta, and more continue to release faster, cheaper, best-in-class technology. For mid-market firms, the math rarely works. The giants outpace your innovation before your product even launches, and your core business slows down while it waits for returns that may never come. Selling AI might make sense as a long-term R&D play. It’s just not the lever that moves revenue in the next two quarters.

Whether you are an AI product or service, competition is high. Add in retraining teams and rebuilding workflows, and it’s a long game, not a quick win. Does slapping “AI” on what you do make it more buyable? I believe buyers are fatigued with so many AI claims.

3) Showing up inside LLMs is your best bet for growth.

Showing up inside LLMs means your business is named, cited, or recommended when a buyer asks AI for answers. Google still leads search, but large language models such as ChatGPT, Gemini, and Perplexity are growing quickly.

Why this matters for revenue

• Gartner predicts that by 2026, traditional search volume across Google and Bing will drop by 25%.
• Forrester reports that 89% of B2B buyers have adopted generative AI for self-guided research. These users are already qualified because they have done their research “right in ChatGPT (and the likes).”
• Semrush data shows AI-assisted search traffic converts at up to 4.4× the rate of traditional organic traffic (and Forbes reports up to 9×).

If an AI system delivers the first answer to a buyer’s question, that engine effectively becomes your first salesperson. It shapes perception, shortlists vendors, and influences who gets contacted. If that system accurately describes your solution, it builds trust before you ever speak to the buyer. If it gets you wrong, it becomes your worst salesperson, one you can’t see, correct, or retrain.

The reputational risk you don’t see

That is the new reputational risk. AI is already presenting answers about your business (or saying it cannot find any information at all), whether you like it or not. If that answer is incomplete, outdated, or pulled from a competitor, it distorts how buyers see you.

This emerging discipline, whether you call it AI Discoverability, Generative Engine Optimization (GEO), Artificial Intelligence Optimization (AIO), or Answer Engine Optimization (AEO), all point to the same goal: making sure AI systems can understand, trust, and surface your business as part of the answer the moment your buyer asks the question. Call it whatever you want, I’m tired of all the buzzwords, too.

What we’re testing next

We’ve been digging into what it actually takes for LLMs to recognize, cite, and trust a business. There is still a lot we don’t know, which makes it even more intriguing. (I’m in geek mode as we speak!) Over the next few months, we’ll be sharing what I’m testing and learning while I clock more than 100 hours of strategic and technical training.

This isn’t a trend. It’s the next evolution of how businesses build credibility and turn it into revenue.

The Zero to Visible Experiment

We are running a Zero to Visible experiment: taking a cold-start B2B brand and showing, step by step, what it takes to get named, cited, and recommended by LLMs with public benchmarks and real steps. If you want a weekly update with the experiment results as we publish them, subscribe to the newsletter below.