AI models now meet or exceed human performance across many tasks, including multitask language understanding, PhD-level science questions, and competition-level mathematics. Despite these incredible technological advances, evidence suggests firms have been hesitant in deploying AI at the at the enterprise level. Just over 9% of firms are currently using AI, according to the Census Bureau’s Business Trends and Outlook bi-weekly survey.
Of course, many firms may be using AI in incidental or insignificant ways and thus not reporting such use. Many employees may also be using AI (in the form of GPT and copilots) without business leaders’ awareness of such use.
Still, the Census Bureau’s survey – which reaches approximately 1.2 million employer businesses over a given 12-week period – is highly relevant as the results arguably reflect the most direct and significant use of AI for business purposes.
Of course, many firms may be using AI in incidental or insignificant ways and thus not reporting such use. Many employees may also be using AI (in the form of GPT and copilots) without business leaders’ awareness of such use.
Still, the Census Bureau’s survey – which reaches approximately 1.2 million employer businesses over a given 12-week period – is highly relevant as the results arguably reflect the most direct and significant use of AI for business purposes.
Note: The BTOS sample consists of approximately 1.2 million employer businesses over a given 12-week period. Thus, each bi-weekly collection goes to a sample of about 200,000 businesses. The average biweekly response rate over the period of collection for AI-related core and supplement content is about 16%, resulting in a sample of about 164,500 businesses for our main analysis sample utilizing the supplement. Following standard Census Bureau procedures, the sample is weighted so that estimates are representative at the national, state, sector and firm size level.
Are some sectors dragging down national statistics on enterprise AI-use?
The answer is yes, but even key sectors are venturing into organizational-deployment with caution. The chart below shows the share of businesses by sector reporting using AI to generate goods and services in the last two weeks. Some sectors have negligibly small shares of businesses reporting organization-wide deployment of AI. Some examples include accommodation and services, and transportation and warehousing– a key reason that economy-wide statistics of AI-use by organizations remains low (and one key reason why the Census Bureau data may show lower AI-usage rates compared to private-sector surveys asking similar questions, amongst other sampling differences).
Beyond sectoral differences, what is more striking is that AI-organizational deployment remains low across basically every sector. Even the information and professional services sectors remain cautious in deploying AI with only around a quarter of enterprises reporting AI-use. This is surprising given the plurality of compelling and demonstrated enterprise use cases, ranging across data infrastructure platforms and management, business intelligence generation, machine-learning models, agentic workflows, and strategy-use cases.
Let’s take a look at what we do know about the firms who are deploying AI, and what can we learn about why are firms might be holding back despite substantial gains in AI technical capabilities.
Organizational challenges may be limiting AI adoption
AI technologies can diffuse through the economy either via the entry of new businesses using AI or the adoption of AI by existing businesses. The data shows a “U-shaped” adoption curve with both small and large businesses deploying AI at a higher frequency than mid-sized businesses, illustrating the frictions associated with organization change when AI technologies are deployed.
Existing businesses can benefit from AI only to the extent they can change the way work is organized
For example, businesses may need to retrain or retool their existing workforce in order to effectively utilize AI. According to Census Bureau data, among firms expecting to use AI in the next six months, training existing staff (41.9%) and developing new workflows (37.6%) are still the most common adjustments businesses expect they will need to make to integrate AI-use into their organizations.
In anticipation of AI use, companies also reported they will need to make changes to data collection or data management practices, use vendors to install/integrate AI, purchase computing power, or hire staff trained in AI.
Small businesses may be able to navigate these frictions and more readily integrate new technologies as their business processes may be more nascent or they may even be able to design the business around the new technology, for example AI-using start-ups. Compared to large businesses, mid-sized companies may be limited by leaner budgets, smaller teams, and more limited IT support – mid-sized companies may particularly be able to benefit from partnerships with consulting enterprises offering AI deployment capabilities and services (offering industry experts and technology practitioners with AI deployment experience, de-risked capex investment, and helping prepare organizations for the changes AI brings), or else they risk being left-behind by larger competitors that have economies of scale on their side.
In anticipation of AI use, companies also reported they will need to make changes to data collection or data management practices, use vendors to install/integrate AI, purchase computing power, or hire staff trained in AI.
Small businesses may be able to navigate these frictions and more readily integrate new technologies as their business processes may be more nascent or they may even be able to design the business around the new technology, for example AI-using start-ups. Compared to large businesses, mid-sized companies may be limited by leaner budgets, smaller teams, and more limited IT support – mid-sized companies may particularly be able to benefit from partnerships with consulting enterprises offering AI deployment capabilities and services (offering industry experts and technology practitioners with AI deployment experience, de-risked capex investment, and helping prepare organizations for the changes AI brings), or else they risk being left-behind by larger competitors that have economies of scale on their side.
What are businesses telling us about why they’re holding back
Enterprise-level AI deployment is going to be determined by the costs versus the benefits that business leaders assess to be associated with these new technologies. There are a variety of other reasons business report holding back (per supplemental questions on AI use from the Census Bureau survey back in early 2024).
For many businesses, the return-on-investment is still unclear relative to the potential costs. Many businesses cite that AI is not applicable to their business, but they may not appreciate the technology’s capabilities. Relatedly, many businesses report various cost-related concerns, citing expense or lack of skilled workforce (with higher costs associated with hiring the necessary talent) and risk-related concerns such as privacy and security, laws and regulations, and concerns about bias.
For many businesses, the return-on-investment is still unclear relative to the potential costs. Many businesses cite that AI is not applicable to their business, but they may not appreciate the technology’s capabilities. Relatedly, many businesses report various cost-related concerns, citing expense or lack of skilled workforce (with higher costs associated with hiring the necessary talent) and risk-related concerns such as privacy and security, laws and regulations, and concerns about bias.
Reasons for Not Planning to Use AI
| Reason | Firm-Weighted | Employee-Weighted |
|---|---|---|
| AI is not applicable to this business | 80.9% | 76.2% |
| Lack of knowledge on AI capabilities | 7.3% | 8.8% |
| Concerns about privacy / security | 6.6% | 8.9% |
| AI is not a mature enough technology | 6.1% | 8.1% |
| Other | 4.5% | 5.8% |
| Too expensive | 4.1% | 4.6% |
| Lack of skilled workforce | 2.9% | 3.6% |
| Concerns about bias | 2.8% | 2.6% |
| Lack of required data | 2.2% | 2.4% |
| Laws and regulation preventing or restricting AI use | 1.2% | 2.5% |
| Previous or current use of AI did not meet expectations | 0.9% | 0.5% |
Source: Census Bureau Business Trends and Outlook Survey Supplemental Content December 2023 – February 2024
What can we expect regarding AI deployment?
We expect the cost versus benefit calculation for enterprise-wide AI deployment will continue to evolve as the technologies continue to mature and advance and as AI-expertise diffuses through the economy. As the chart on the left shows, individual use of generative AI has continued and been exponential. -Enterprises are likely to be not far behind (in part because individual AI use for work both limits enterprise level productivity gains and potentially leaves businesses more exposed to data breaches and regulatory risks).
Necessity is often the mother of invention
As shown below, US productivity gains typically accelerate coming out of economic downturns as firms discover ways to gain efficiency through hardship. With incredible advances in technology in recent years, firms may be able to navigate macroeconomic challenges better than ever and come out with better productivity and efficiency. There may be many reasons to be concerned today about the global economic outlook, but as AI technologies continue to diffuse through the economy, there are also many reasons to be optimistic.
