The 10% Rule: When AI Becomes a Cost, Not a Multiplier
Key Takeaways
- AI is only worth it if it delivers a 2x gain. A 10 percent improvement that costs more than it saves is not an efficiency. It is an expense.
- The subsidy phase will not last. AI tools are cheap now because investors are funding adoption. When pricing normalizes, only the tools delivering real ROI will be worth keeping.
- Fix the process before adding AI. If a workflow is full of exceptions and manual patches, AI will not fix it. It will scale the mess.
- The winners will not have the most AI tools. They will have the fewest tools that each deliver transformational impact.
- Stop asking how fast your team is adopting AI. Start asking what they are doubling and cut anything that cannot answer that question.
Every week, another tool. Another seat licence. Another API meter ticking quietly in the background of some workflow nobody has fully mapped. The AI stack in most companies today is not a strategy. It is a pile of subscriptions.
I have spent a lot of time this year looking at what AI actually does for us inside Kentico, and for our partners and for the companies we sell to. And one thing has become clear. The conversation we are all having about AI adoption is the wrong one.
Everybody is asking: how fast can we adopt?
Nobody is asking the harder question: when does this stop being worth it?
The Real Math Behind AI Productivity
Let me put some numbers on that. Imagine someone on your team earning 100,000 a year. Thanks to AI, they become twice as productive. You now get the output of two people from one salary. Those AI tools cost you 50,000 a year in tokens and subscriptions. The math still works. You have saved 50,000 on the second headcount you did not need to hire.
Now change one number. Instead of 2x, AI makes that same person 10 percent more productive. You are still paying 50,000 for the tokens. You have just paid 50,000 to save 10,000 worth of output. That is not an efficiency gain, it's a cost. This is the calculation almost nobody in the industry is doing.
How Kentico Measures Real AI Value
I am not against AI. Far from it. At Kentico, our developers are moving up to 30 percent faster on technical work using AI tools. Our marketing team is half the size it was two years ago and producing more output now than it did then. Our legal team is executing contracts 20 percent faster because AI is doing the first pass on redlines. We recently needed security awareness training software and instead of buying it, we vibe-coded our own. That one decision saved us somewhere between 20,000 and 50,000 a year.
These are real gains. I see the numbers every month. But every one of them passes a simple test. The AI is not a 10 percent improvement. It is a fundamental unlock.
When the gain is real, the math speaks for itself.
The Subsidy Phase Won't Last Forever
There is a second pressure coming that most leaders are not pricing in. Right now, AI tools are cheap because investors are paying for your usage. The big providers are subsidizing adoption to win market share. That phase ends. It always ends.
I lived through it with cloud computing. Everyone assumed the price of AWS and Azure would keep falling. It did not. As soon as enterprises were locked in, pricing stabilized. Then it climbed.
AI will follow the same curve. The question is not whether token costs go up. The question is whether you will still be able to justify the spend when they do.
The 2x Rule: A Simple Test for Every AI Tool
So here is the rule I use and the one I ask our teams to use. Before you buy the tool, before you scale the license, look at the workflow you are trying to automate and ask one question. Does this give us a 2x, or does it give us a 10 percent?
If it is a 2x, move fast, invest and lock it in.
If it is a 10 percent, stop. It is not an efficiency. It is an expense waiting for the subsidy to end.
This is not about being skeptical of AI. It is about being disciplined with it. The companies that win the next five years will not be the ones with the longest AI tool inventory. They will be the ones that cut AI spend as ruthlessly as they cut any other line item that fails to pay for itself.
Fix the Workflow Before You Add AI
There is a deeper point underneath the math. If your workflow only gets 10 percent better with AI, the AI is not the problem. The workflow is. Too many exceptions. Too many special cases. Too much process nobody has had the courage to simplify.
I see this in our own sales contracts. Every deal has its own exceptions, and that is exactly why we cannot automate renewals the way we want to. The workflow has to become simple enough for AI to amplify. Until it does, no tool will save you.
Start with the process. Then apply AI. Not the other way around.
Stop Counting Tools. Start Counting Returns.
I will keep saying this. AI is not a strategy. It is a force multiplier. A multiplier only works when there is something worth multiplying. If the underlying work is friction, exceptions, and manual patches, AI just multiplies the mess and sends you a bill for the privilege.
Here is my challenge to the other CEOs and leaders reading this: stop asking your teams how fast they are adopting. Start asking them what they are doubling. And when something is not doubling, stop paying for it.
Grow fast, or do not grow at all. That is the real choice in front of every software company today. In an AI-first world, the companies that understand this calculation will compound. The ones that keep adding tools because the demos looked good will spend a lot of money proving the obvious.
The obvious is that 10 percent is not a multiplier. It is a cost.
Frequently Asked Questions
Get ready for your upgrade
Upgrading from Kentico Xperience 13 to Xperience by Kentico gives your marketing and development teams a unified and future-proof platform with native AI that reduces complexity, and helps you create more impactful customer experiences.
