Is AI Actually Changing How Companies Work?
PwC surveyed 4,454 CEOs across 95 countries and the answer is not encouraging.
Written by Dana Condrea, Director of AI Services at Hype
in 2025, worldwide spending on AI hit $1.5 trillion. every earnings call mentions it. every team has a roadmap slide with “AI integration” somewhere on it. the money is moving. but what did companies actually get back from all of that?
PwC surveyed 4,454 CEOs across 95 countries and the answer is not encouraging.
56% of CEOs report getting nothing measurable from AI investments (PwC)
95% of generative AI pilots fail to deliver P&L impact (MIT)
80%+ of all AI projects fail, 2x the rate of regular IT projects (RAND)
42% of companies abandoned most AI initiatives, up from 17% one year prior (S&P Global)
only 6% of organizations qualify as AI “high performers” (McKinsey)
for most companies, the honest answer to the question is: no. they’re not getting anything meaningful done yet.
despite all of this, 85 to 92% of organizations say they plan to increase AI spending. not hold it steady. increase it. so why does the money keep flowing?
because the proof that AI works does exist. it’s just not inside most companies.
General Catalyst built a $1.5 billion roll-up engine around this exact thesis. they acquire fragmented service businesses, automate operations with AI, and some of these companies doubled their EBITDA margins within 12 months. PE firms across the board are running the same playbook. buy underperforming businesses, strip out inefficiency with AI, flip or scale. tech M&A hit nearly $480 billion in deal value by mid-December 2025, up 75% year over year, with almost half of the larger deals involving AI-native companies or citing AI as a core value driver.
on the product and engineering side, the numbers are just as real.
Klarna: revenue per employee went from ~$300K (2022) to ~$1.3M (early 2025). AI replaced the workload of 700 agents. resolution time dropped from 11 minutes to 2 minutes
Shopify: headcount cut from 11,600 to 8,100 while revenue grew 20-40% annually. CEO made AI usage a baseline performance expectation
Anthropic: 70-90% of code is AI-generated company-wide. PMs don’t write PRDs. they build the first version of features directly in the codebase using AI. their head of Claude Code ships 20-30 pull requests a day running 5 parallel agent instances and hasn’t written a line of code himself in over two months
General Catalyst portfolio: companies doubling EBITDA margins within 12 months of AI-driven restructuring
so the proof of concept isn’t missing. AI can compress teams, cut costs, and move faster than most executives thought possible two years ago.
but look at where all of those examples come from. PE firms with small operating teams who rebuild companies from scratch. AI-native startups that were built around these workflows from day one. founders and engineers who treat AI as the default, not the add-on. none of them are a 2,000-person company trying to bolt AI onto processes that were designed in 2015. and that’s the gap.
BCG’s 10-20-70 framework breaks it down: 10% of AI success comes from the algorithms, 20% from the technology, and 70% from people and processes. McKinsey’s 2025 research confirms it. workflow redesign has the single biggest effect on whether a company sees actual EBIT impact from AI.
AI adoption bottleneck in numbers
68% of CFOs cite AI skills gaps as their top challenge to ROI (RGP)
only 10% of CFOs fully trust their enterprise data (RGP)
only 22% of employees say their org has communicated a clear AI plan (Gallup, n=23,068)
the tools are available. the orgs aren’t set up to use them.
and that leads to the part most companies don’t want to hear. AI talent commands a 23% wage premium across the board (PwC). senior AI engineers and ML leads sit at $350K+ total comp. McKinsey found that 48% of smaller businesses haven’t hired for a single AI role in the past year. RSM’s survey shows 70% of mid-market firms openly say they need external support.
you can’t solve this by posting a job. you need someone who understands AI deeply enough to look at your existing workflows, identify what should be automated, rebuild the processes, and actually ship it. those are high-agency individuals who are fluent in AI and also understand business operations well enough to know what to change. they’re rare, expensive, and every company is trying to hire the same profile.
for most companies right now, AI spend is a number on the accounting closing sheet that doesn’t connect to anything measurable on the other side. the proof that AI can transform a business is real. but replicating those results inside an existing organization requires one of two things: either invest seriously in finding and retaining people who can rebuild your workflows from the inside, or bring in external specialists who do this full time and have already figured out what works.
without one of those two, the pattern stays the same. more budget. same results.
that’s not a prediction. that’s what 4,454 CEOs already reported.
sources:
worldwide AI spending 2025 ($1.5T):
56% of CEOs getting nothing from AI: PwC 29th Global CEO Survey (n=4,454)
95% of gen AI pilots fail: MIT NANDA GenAI Divide report
80%+ AI project failure rate: RAND Corporation
42% abandoned AI initiatives: S&P Global (n=1,006)
6% AI high performers: McKinsey 2025 State of AI
85-92% plan to increase spend: Deloitte 2025 / McKinsey 2025
General Catalyst $1.5B roll-up engine:
tech M&A $480B deal value:
Klarna RPE and AI metrics: CEO Sebastian Siemiatkowski, Time interview
Shopify headcount and revenue: public reporting
Anthropic 70-90% AI code, Boris Cherny:
Anthropic PMs building first versions:
BCG 10-20-70 framework: BCG AI Radar 2026
workflow redesign and EBIT impact: McKinsey 2025 State of AI
68% CFOs cite skills gaps / 10% trust data: RGP CFO Survey (n=200)
22% employees heard clear AI plan: Gallup (n=23,068)
23% AI wage premium: PwC 2025 Global AI Jobs Barometer
48% smaller businesses no AI hires: McKinsey 2025
70% mid-market need external help: RSM 2025 AI Survey