The more research comes out about the actual effects of AI in business, the more I wonder...
Are we in love with the wrong AI?
Maybe it's like wanting and finding a man in finance (trust fund, 6'5", blue eyes).
Except sometimes they say stupid things or act like the worst-of-the-worst boyfriend/employee...
Sometimes they take too long or forget what we just said or we spend time correcting them instead of just doing the work.
But we ignore all that because everyone says they want it/have to have it/there's cool memes about it.
And they're just nice to look at (i.e. we enjoy cognitive breaks instead of having to concentrate when we engage with generative AI).
Hope vs Reality: AI Use In Business
Interesting research from Metr in July 2025 is what first got me pondering whether this the love affair my clients really want?
Whenever someone talks about "adopting" AI in their business, they always seem to mean the trendy, cybernetic collaborator.
The one who listens, encourages, empathizes.
The one who talks to us as if we matter, builds us a vision of what we could be, and is constantly getting in better shape, doesn't snore, and knows exactly where to find that article we want to quote while we look widely around our desk searching for a long gone post-it that we're sure we left around for just this very purpose.
AI productivity gains, typically measures around the uptick in Generative AI use (ChatGPT, CoPilot, Claude, etc.) aren't quite meeting expectations nor even our perception.
Which brings us to the Metr study...
Because even self-reported productivity gains aren't matching reality.
The study researched how early-2025 AI tools affect the productivity of experienced open-source developers.
The developers used AI in a collaborative fashion on a total of 246 tasks.
Ahead of the study, forecasts among the economic experts and machine learning experts put the average expected productivity gain around 35-40%. The experienced developers were a little more conservative but estimated a 10-30% increase in productivity using generative AI.
Where it gets interesting though is that after the project work, some projects with AI and some not, the developers said they were slightly less productive than expected but still around a 20% increase.
Except their productivity was observed objectively in addition to their estimates. And the observed result was no where near what the developers quoted. Instead of a 20% increase, developers had an average 20% decrease in productivity.
The Perception Gap In AI Adoption: What We Want vs What Is
There's more to say about why this happens but the perception gap is particularly fascinating to me.
It means the following:
Do we want generative AI to make us more productive?
Yes.
Do we expect generative AI to make us more productive?
Yes.
Do we believe generative AI makes us more productive?
Yes.
But do we objectively get more productive using generative AI?
Quite possibly no.
Add to this that if 95% of big businesses aren't getting their AI to completion in all the ways that matter...
(See this other July 2025 study from MIT Nanda)
Are you really going to be the one to change him/it?
("But c'mon, Rachel... he's 6'5", in finance, and HOT. At least he's not cheating on me, right?")
The problem is that this kind of AI relationship is surface level and prone to pitfalls.
Bridge the AI Profit & Productivity Gap With Fundamentals
When it comes to the businesses who get results from AI, the details are far less cover-model flashy... at least at first glance.
Real AI growth is being driven where there is alignment between:
We can't slap a bandaid on or get some arm candy and expect a business to transform.
All three of these are relevant for change.
You can get the process and the tools in place, but if the culture isn't attuned to or coached to adapt to changes, you can't expect full adoption.
And if you add in something new and get people using it, but you're still stuck dealing with the same roadblocks, what's the point?
And if you know the goals but your processes are ill-defined or your data is inconsistent and sitting in separate unlinked places, you'll struggle to get out of the starting gate.
Plus there's the perspective problem itself...
Solve the Perspective Issue: Get Outside Help With AI Initiatives
You may know what you want changed or fixed (you have goals!), but being so deep in the business, you might not see how the process can be transformed to meet those objectives.
You wouldn't be alone in this struggle.
It's likely why the data shows that 67% of AI initiatives that do reach deployment with measurable returns (the thing that matters) use external help - experts from outside your business who understand processes and how to implement best practices across your organization and tool stack - versus 33% for internally driven initiatives. (MIT Nanda again)
HINT: this might just be something we can help you with! Check out our Fractional COO services or schedule a consultation to start exploring how outside expertise could help your business scale.
Win With AI Through Business Operations
As much as results should should the day, we humans also want the flash.
People want the arm candy.
Companies want the off the shelf... (80% have explored or piloted the general LLMs (e.g. ChatGPT) but only about half that have deployed it.)
And corporate budgets will go toward marketing and sales.
(Even though the data tells us that it's the behind the scenes infrastructure investments that pay off long term.)
We might want generative AI to be our "man in finance."
But like any good long-term relationship, it takes work to get the things that matter.
Just like planning date nights, filling the gas tank, and unloading the dishwasher without being asked...
The fundamentals never go out of style.
Your data, processes, and cross-business functionality will be key to long term success and profit when adopting AI or scaling your business.
Think about it this way:
Be careful you're not holding onto relationship vapors.
Because the data is showing us... it's time to fall in love with cool nerds, clear goals, and a quality systematic approach.
