Insights from MIT Technology Review

In August 2024, The MIT Technology Review Insights team published a comprehensive playbook for crafting AI strategy, based on a global survey conducted in March 2024. In Technology Years, August 2024 is ancient, but it’s still worth looking at some of the goals from 2024 and how (if) they can be achieved in 2025.

The research captured insights from 205 executives and technology leaders across eleven industries, including financial services, manufacturing, IT, healthcare, and more.

Given that 88% of the respondents were from the C-suite and organizations generating over $500 million in annual revenue, the report offers a unique window into how large enterprises are approaching AI adoption.

The Current State: A Clear Implementation Gap

The current state of AI presents an interesting paradox: while 95% of companies are using AI in some capacity, only 5.4% have successfully scaled it to develop products or services enterprise-wide. This stark contrast, highlighted in the MIT report, reveals significant challenges organizations face when moving from pilots to full-scale deployment.

The report's findings paint a detailed picture of this implementation gap. First of all, although nearly all surveyed companies (99%) expect to use AI in the future, 76% are currently stuck in limited deployment, e.g., AI applications confined to just one to three use cases.

This raises an important question? How can organizations move past the experimentation phase to achiever broader organizational adoption? And what issues or factors are currently standing in the way?

Second, and somewhat in the face of the previous point, it’s interesting to note that approximately half of the surveyed companies expect to fully deploy AI across all business functions within two years. This anticipated scaling makes 2025 a crucial year for establishing the necessary foundations for enterprise-wide AI adoption.

Back to the gap.

The MIT report identifies several key barriers:

• Data Quality Challenges: Half of the respondents cite data quality as their most limiting factor

• Infrastructure Limitations: Legacy systems and outdated architectures impede scaling efforts

• Governance Concerns: 45% of companies (and 65% of those with revenues over $10 billion) identify governance, security, and privacy as major brakes on deployment speed

• Resource Constraints: Medium-sized firms particularly struggle with budgetary limitations, with 47% citing this as their primary obstacle

This implementation gap is more than just a series of technological challenges; it represents a broader organizational struggle to transition from isolated AI experiments to integrated, enterprise-wide solutions that deliver consistent business value.

To bridge this gap requires not only technical expertise but also strategic vision, organizational change management, and a careful balance between innovation and risk management.

Piece of cake, right? (Sarcasm.)

Building the Foundation: Data Quality is Non-Negotiable

Data quality emerged as a critical concern in the MIT report, particularly for larger organizations. The study found that companies with revenues exceeding $10 billion were the most likely to cite both data quality (52%) and data infrastructure (55%) as primary obstacles, compared to overall survey averages of 49% and 44% respectively.

The report emphasizes several key components for success:

• “Data liquidity” - ensuring seamless access and integration across sources

• Quality assurance frameworks

• Robust IT infrastructure

• Effective data governance policies

Investment Trends & ROI

The report provides some interesting insights into AI investment patterns, too. Spending in 2022-2023 was modest for most companies, with only one in four increasing their spending by more than 25%; however, 2024 showed a dramatic shift:

• 90% of companies planned to increase AI readiness spending

• 40% planned increases of 10-24%

• One-third anticipated spending increases of 25-49%

The report highlights how companies are developing new ROI methodologies that go beyond traditional metrics. For instance, Motorola's framework tracks task completion times with and without generative AI.

Klarna has quantified their AI's output as equivalent to 700 customer service agents.

Security & Governance

One of the most striking findings from the report is that 98% of organizations are willing to forgo first-mover advantage to ensure secure and safe AI deployment. This caution is particularly pronounced among larger organizations: 65% of companies with revenues over $10 billion cite governance, security, and privacy as their primary concerns in AI deployment.

(It’s also amusing that 2% of organizations were like: Eh, safe deployment isn’t for us…)

The Partnership Strategy

Matt McLarty, CTO at Boomi, and Kevin Collins, founder and CEO of Charli AI, emphasize the importance of strategic partnerships. As Collins notes in the report, "For most organizations, building their own large language models is very expensive and the value is time-limited."

Instead, the research suggests success often lies in:

• Strategic partnerships

• Fine-tuning existing models

• Focusing on application rather than creation

• Leveraging industry-specific solutions

Looking Ahead: 2025 and Beyond

The MIT report notes that 2024 is (was) a crucial year for establishing AI foundations, with many organizations aiming for enterprise-wide deployment by 2025. However, it’s yet to be seen how successful this has been.

The research indicates that success will depend on organizations' ability to:

1. Address data quality challenges head-on

2. Develop robust governance frameworks

3. Create clear ROI measurement methodologies

4. Balance innovation with security

5. Build strategic partnerships

The report makes clear that the gap between AI experimentation and enterprise-wide implementation, while significant, can be bridged with the right approach. As McLarty emphasizes, "The biggest factor holding back AI implementation is people not knowing where to start."

What do you think?

Random Fun Fact of the Week

The average smartphone today has more computing power than all of NASA's combined computing power in 1969 when it sent astronauts to the moon. The Apollo Guidance Computer operated at 0.043 MHz, while modern smartphones operate at over 2.5 GHz.

Contact us at NorthLightAI.com to learn how we can help you build a stronger data foundation for your AI future.