Since OpenAI debuted ChatGPT in November 2022, Artificial Intelligence has dominated the conversation of what is possible in the future, and created urgency in businesses to deploy its potential.
A recent chart from Pitchbook shows just how much Artificial Intelligence investment has overshadowed all other sectors.
However, the investment has not easily lived up to its potential. A report from MIT Media Lab and Project NANDA found that 95% of AI Projects are getting zero in return.
The paper's findings highlighted two key takeaways:
- Enterprise paradox: Big firms lead in pilot volume but lag in scale-up
- Investment bias: Budgets favor visible, top-line functions over high-ROI back office ones
In this report, I take an alternative approach to identify how the largest companies are investing in AI, and where they have been most successful in implementations.
To create this data, I used the quarterly earnings transcripts of S&P 500 firms from Q1 2024 through Q2 2025 to measure the progress of AI implementation.
Key Findings from Earnings Transcript Analysis
- Growth of AI initiatives has been steadily increasing. By Q2 2025, more than 60% of the S&P 500 firms had announced an AI initiative.
- Firms reported tangible returns from these initiatives 21.4% of the time, nearly 4x that of the MIT / Project NANDA report.
- While firms are deploying more projects in every area by Q2 2025, they have launched more projects aligned to improving Core Operations than Front Office tasks such as Sales, Marketing and Customer Service - a core recommendation from the MIT paper's findings.
- Projects in Voice AI, Document Processing and Code Generation produced an impact on a greater percentage of their initiatives than the average, which aligned to findings from the MIT paper.
- Success varied by sector; despite Financials deploying the most AI projects, they had the lowest percentage of products described with any impact (19%).
By Q2 2025, 323 firms from the S&P 500 discussed an AI initiative
As expected, there was strong growth in firms mentioning they were 'Using AI' in quarterly earnings reports. By Q2 2025, 65% had mentioned a project initiating the use of AI in their firm.
When we look at the numbers reported within each quarter, we see a huge jump from Q3 2024 to Q4. Part of this is the role Q4 earnings plays in both summary of the previous year and projection of what's to come, but it keeps that pace throughout 2025.
This data trends along with similar data found in the US Census Business Trends and Outlook survey. The survey also tracks AI progress by asking two questions:
- During the next six months, do you think this business will be using Artificial Intelligence (AI) in producing goods or providing services?
- In the last 2 weeks, did this business use Artificial Intelligence (AI) in producing goods or services?
Positive responses to the first question increased 130% from just 6% in Jan 2024 to 14% by Sept 2025.
And it wasn't just intention; the growth in firms answering having used AI on a transaction or service in the last two weeks increased from 6% to start the year to nearly 10% by August 2025.
S&P 500 Companies described Returns on their AI Investments 21.4% of the time
When Companies discussed an AI project, they included measured returns 21.4% of the time. While this does not mean the projects did not see a return on investment if it was not discussed in the Earnings Transcript - this serves as a reasonable proxy due to the incentive to show investors the technology is giving an edge.
Since only publicly declared projects are included, this may not necessarily be an optimistic counterweight to the MIT report. In fact, many of the returns were only provided with qualitative statements but were still included in the report.
Despite the increasing experience of launching more AI projects, the S&P 500 Companies don't appear to be getting better at seeing returns on AI projects as time has gone on.
Companies discussing launched AI projects increased by 36% from Q1 2024 to Q1 2025
In 2024, companies more frequently described AI initiatives in early stages. But by 2025, more projects were described as fully deployed. One hypothesis could be that the shift to more Core Operations projects found more targeted and easier to scale projects that could launch faster.
Firms shifting focus to AI in Operations
MIT Media Labs' paper indicated firms were more likely to invest in AI in Marketing and Sales, but less likely to see a return than Operations. Our research shows firms shifting to more focus on using AI in Operations.
While projects in these different categories were roughly at parity in Q1 2024, Core Operations became the most common area where firms were investing in AI projects.
Note: Not all projects were able to be categorized based from the discussion and were removed from this view. Uncategorized would be the highest number of projects in each quarter.
Higher success in MIT's recommended project categories
From the MIT paper: "Successful categories from our sample included: Voice AI for call summarization and routing; Document automation for contracts and forms; Code generation for repetitive engineering tasks"
The Earnings Transcripts mentioning AI investment projects were categorized across these three project types and split out from the rest. We found projects in these 3 categories were more likely to mention a return on that investment than those in different categories ("Other").
Financials lead in volume, lag in returns
In the first Quarter of 2024, 13 Companies in the Financials Sector announced AI projects, putting it roughly at parity with the other leading sectors in Consumer Discretionary and Health Care. By Q2 of 2025, Financials more than doubled to 30 projects announced and 10 more than the next nearest sector.
Top Financials Projects
| Company | AI Application | Impact |
|---|---|---|
| Nasdaq, Inc. | AI Copilot tools for Developers | Up to 90% reduction in Alert research time |
| Mastercard | Fraud detection solution 'supercharged' by Gen AI | Fraud detection rate improvement of 20% on avg. |
| Hartford | AI-driven underwriting logic for coverage suggestions | 75% quote bindability within minutes |
But despite this activity, it was the poorest performing Sector when looking at the percentage of projects that mentioned impact or returns with 15.9%.
Part of this may be Financials already having implemented AI for opportunities like Fraud Detection and have moved to harder-to-implement projects.
While 40% of Code Generation projects were announced with measured impact - in line with all sectors - they made up just 5% of the projects.
26.5% of projects in the Code Generation, Voice AI, and Document Automation categories were announced having impact, compared to just 8.1% in all other categories.
Standout Example: PPL Corporation
In operations, PPL announced the integration of AI into its business operations and data center support yielded $150 Million in savings.
A big theme in many of the projects announced by the Financials sector were attempts to bring AI to the customer experience. The attempts to personalize the Customer Experience may be long-term impactful but had yet to materialize.
By contrast, Utilities was the highest in mentioning Returns at 30.1%. In this sector, companies were focused on proven AI capabilities like Vision to analyze footage for signs of fire.
Limitations
This research was intended to understand how the largest and most well-capitalized companies are implementing AI into their business by taking advantage of investor earnings transcripts.
There are limitations in its measurement, especially compared to the MIT report which it is responding to.
Mentions vs. Investment $
We are only counting an AI project when it has been mentioned by the firm, but we are unable to know how much investment is required from each project.
MIT Media Labs / Project NANDA's paper discusses the investment being put toward Marketing & Sales which does not show up as strongly in this analysis. But it may just be the case that the Marketing & Sales AI projects are more expensive and slow to roll out, while the Operations projects have rolled out more quickly and with initial ROI.
Unmeasured Activity
I find the concept of counting how often a company mentioned a return on their AI investments in these calls to be an interesting way to track the success, but it is only measuring what is being offered.
Much business activity does not make it into these calls. That only 21.4% of AI projects discussed are mentioned with measured impact leaves me with a hypothesis that the 5% rate is likely if you include all projects which did not scale up and launch.
AI Synthesis
This project was created with the aid of LLM tools for categorization and synthesis purposes. As a result of this, some categorizations were based on judgement between the AI and human reviewer (me).