Whatever your business is, I’m sure you’ve spent time imagining what it will look like in a world with advanced AI.
Specifically, how AI could transform your processes, improve your product or services, or help you communicate with customers.
Could a future AI conceivably execute your core business functions itself? Automatically?
We’ve been doing this kind of thinking for years, and while the AGI talk has died down a bit after GPT-5—and Meta’s using their Superintelligence team (or some of it anyway) to work on Llama—there’s still no denying that AI is getting better fast.
Language-learning app company Duolingo knows this all too well.
An AI-first company, GenAI has helped them increase revenue by 41%, bookings by 41%, active users by 40%, monthly users by 24%, and paid subscribers 37% year over year, per Bloomberg. In early August, they raised their outlook and increased bookings while also increasing profitability.
Their value soared as a result of this data, as you’d expect. Until later that day.
GPT-5 arrived at the same time, and OpenAI’s demo showed the newest version of their chatbot whipping up a custom web app to teach users French. It was done in just three minutes, and at the same time, halved Duolingo’s gains in just hours.
[Our most recent PTP Report profiles this and other business news like it in the July–August AI roundup.]
None of this is new, of course, but today I revisit the topic in light of news around AI pilot failures, and against the successes we’re seeing in AI-driven innovation in 2025. For reinvention, I look specifically at what Google is doing.
It wasn’t so long ago we were all talking about how every business is really in the data business. It turns out we may all be in the AI trade as well.
But we’re not there yet.
The 95% and Promised AI Disruption in Business in 2025
As AGI talk fades, AI bubble talk has gotten louder.
This was especially true after MIT released a report showing that, despite $30–40 billion in spending, 95% of the enterprise AI pilots they studied failed to deliver meaningful ROI.
This has been quoted and thumbnailed everywhere, though the fine print is quite intriguing. I encourage you to read the report yourself.
The researchers conclude that this divide is more due to approach than either capacity or model quality. In businesses, generative AI is now improving individual productivity far more than organizational, and enterprise-wide systems end up rejected (60%) more often than adopted.
Patterns they profile include:
- True disruption has been limited so far (to just two of eight analyzed sectors).
- Bigger companies have fared far worse, despite leading in the pilots.
- Investments have favored the high profile and visible over higher-ROI areas.
- External partnerships have done far better than in-house builds (around twice the rate of success).
The study found that despite all the noise, only 5% of enterprises really have AI integrated in their workflows at scale.
While LLMs have a lot of flexibility, enterprise-crafted tools often fail to learn sufficiently, are too brittle, narrowly capable, or end up more trouble than they’re worth.
Unsurprisingly, size is also problematic, and a lack of speed kills.
Their study found mid-market top performers moved from pilot to full implementation in just 90 days, while enterprises on average took nine months or even longer.
A Look at AI Pilots and Business Failures
Models are only as good as context, and learning is critical to AI success. So when your tools can’t retain feedback and improve when applied to real workflows or integrate with necessary systems, they are bound to fail. No matter how much money you spend.
If your first AI solution out of the box is aimed at complex asks like managing clients on an ongoing basis or covering work that spans weeks in duration, you are missing out on what AI thrives at right now.
As this report points out, AI has already won over busywork, even if you’re not capitalizing on it. 70% of workers prefer to use AI for drafting emails and 65% for initial analysis.
Startups are thriving with successful AI business models by focusing use on critical, well-defined workflows—like sales, marketing, and customer service.
The report finds AI is thriving in voice systems, call summarization and routing, document automation, and code generation for repetitive engineering tasks (or three strikes and you automate).
In other words, get the wins that you can count on and grow from there.
Adoption of course matters, too. Empowering managers in the know over technicians or labs is essential, as is effective training, and measuring visible outcomes that can align with key KPIs.
The Google Search AI Transformation
This is all fine and well, but this article is about reinvention—building your business with AI—and yet I am discussing quick wins and bottom-up automation of repetitive tasks.
Pilots are not the same thing as transformation, and to that end, no one seems to have read the AI trajectory more clearly than Google.
With a trillion-dollar business that is a foundation of the internet—accounting for 90% of the global market on search and so deeply used that the brand name is a household action—it would have been easy for them to just wait and see.
But when thinking of businesses that AI could replace, search is one of those things that comes up quickly.
With LLMs trained on vast quantities of scraped internet data, they seem well suited to not only do the searching for us but even act as our personal maître d’.
Why click and navigate pages if an AI agent can go do it for me (at many times the speed), or the LLM already has it all in mind? And better yet, it picks what I like, and delivers it in the way I need, when I need it?
Rather than just protect what they do, Google continued to work to stay ahead of the game.
Their AI Overviews was an initial step at merging AI and search in one, and AI Mode takes it further, planning, synthesizing, and answering.
The immediate concern from a business point of view is that this means a loss of clicks and power, and Google is arguably struggling with anti-trust cases now because they’ve done too good a job at amassing power over how sites are accessed.
But now Google is changing search before anyone changes it ahead of them. Instead of just connecting us to options, their focus includes selection, too, helping us digest the content, and even acting on the information.
Financially, this has been successful for them to date, with search beating expectations (by some 12%) and AI Overviews monetizing at the same rate as traditional search while also growing in volume.
They’re also working in ads and shopping.
But aren’t there long-term concerns?
One Core Problem with AI in Search Engines
This move is quite different than just changing the UI.
Content publishers who used to get their traffic now do not. And this risks cutting off the very fuel that powers the engine. If no one is visiting your site and you are receiving no compensation for generating content, why continue?
As Cloudflare’s Co-Founder and CEO Matthew Prince told Time for his profile in their Time100 AI 2025, AI Overviews has made it 10 times harder for news sites to get traffic than it was a decade ago.
And for search via chatbots like GPT-5 or Claude, this number is far worse. AI increasingly doesn’t guide you to the source, it replaces it.
And search, or Google’s core business, with it.
On the content front, Prince has advocated for his company’s default blocking of AI scrapers as a way to force more compensation for content creators.
Note that Cloudflare also brokers some of these arrangements.
AI companies overall have shown their own willingness to explore this, through contracts like The New York Times and Amazon’s or through direct funding efforts like the one recently announced by Perplexity.
But search, as we knew it pre-AI, will almost certainly occupy a far less central place in all our lives.
Google vs AI Competitors
Alphabet, Google’s parent company, is all in on AI, and faring well in what could have been a race for survival. And with their strong numbers, they’ve only upped their AI spend, promising some $85 billion to data centers and continued infrastructure growth.
Alphabet was already one of the three cloud hyperscalers, conveniently controlling the very systems that bring AI to us.
They bought DeepMind in 2014, enjoyed early successes like AlphaGo, but then fell behind and had to aggressively race to catch back up.
And they’ve weathered numerous AI-related failures that have been big news, from results that recommended adding glue to pizza or eating rocks to their image generation fiasco. Yet their aim has been unwavering, and they continue to surge ahead of competitors.
[Here as well you can take a look at our roundup for more on Google’s other recent advances, like Nano Banana.]
McKinsey on AI and Rewiring Your Business
McKinsey & Company is another business that’s seen the AI writing on the wall.
With some 40% of their business now advising on AI and tens of thousands of AI agents at work, they’ve attempted to rewire their business with AI and, like Google, stay ahead of the transformation in consulting.
They’ve also written extensively about it, in numerous reports like March’s The state of AI. As with the MIT report, this stresses oversight, training, leadership engagement, and the importance of continuous improvement.
They also advocate for dedicated adoption teams, feedback loops which ensure improvement, clearly defined road maps, tracking well-defined KPIs, incentivizing employees, and trust-building, both for employees and customers.
Rewiring is their alternative to piloting, though obviously it is not as easy. AI becomes the operating system instead of the application or interface. Their aim is to have an AI agent for every employee, and they’re well on their way.
Disrupt or Face Disruption: Reinventing Your Core Business with AI
At PTP, we’re an AI-first global tech recruiting and consulting firm, and we’ve been working through our own best approaches to these opportunities. That’s meant bringing AI into recruiting in ways that add value and don’t degrade the quality or access to talent.
It’s also meant pioneering AI solutions across the board, from sales to marketing to communication.
I continue to think about the possibilities of AI transformation in a larger sense, but we also see first-hand where it’s winning big for us now—improving our time-to-hire without reducing quality and helping us provide new and varied options for customers.
Not every company has the power of Alphabet, able to read the tea leaves and acquire a leading AI firm an entire decade before it would become commonplace.
But we can visualize what AI may be and forecast its impact on our core business—as more than just a bolt-on tool. Instead, it will become a fundamental aspect of what we are doing in the future, and how we’re doing it.
Conclusion: How Companies Adapt to AI May Define Them Tomorrow
It’s risky spending money gambling on the future of AI in business strategy. But what we must do is take advantage of what it does well now so we can get our minds around its unique benefits and challenges.
It’s commonly said that today’s GenAI systems are “weird,” in that they’re nothing like the tech we are used to. In a guest essay in yesterday’s New York Times, AI luminary and outspoken LLM-skeptic Gary Marcus warned about an over commitment to scaling, while not doing enough to research alternative approaches to AI. At the same time, he points out several that are currently underway (like Google DeepMind’s efforts to build world models).
With such a new and emerging technology, it is almost impossible to know for sure what is coming and when. But it is coming.
And by the same token, it’s essential we not only imagine the possibilities, but adopt now the solutions that are working well.
References
The state of AI, McKinsey
Duolingo stock soars after earnings, forfeits roughly half its gains after OpenAI GPT-5 demo and If AI eats search, Google is still all in: Morning Brief, Yahoo Finance
The GenAI Divide: State of AI in Business 2025, MIT NANDA
Matthew Prince, Time 100 AI 2025
Glue pizza and eat rocks: Google AI search errors go viral, BBC
The Fever Dream of Imminent Superintelligence Is Finally Breaking, The New York Times
