How Companies Are Balancing AI Innovation with Risk

Global spending on generative AI (GenAI) is forecasted to reach US$644 billion in 2025, which is a staggering 76% increase over 2024. The message is clear: Companies feel that if they fall behind in adopting AI, they will face existential threats. Despite its ubiquitousness, however, the use of AI continues to cause concerns, from security risks to operational issues, for enterprise leaders. 

There has always been uncertainty about whether the risks and potential downsides of AI are worth the purported benefits. Innovation promises transformative gains in customer experience and efficiency, but the potential risks of threading AI inextricably into existing operations could have severe consequences.

Technical leaders face a delicate balancing act: How can companies pursue AI innovation at scale without exposing themselves to unacceptable levels of risk?

New research commissioned by Akamai and conducted by Forrester offers some answers about how companies are achieving this balance today. By surveying 400 senior AI and enterprise cloud strategy decision-makers, the study uncovers how organizations are moving fast enough to capture value, while building the guardrails necessary to protect their businesses.

The Forrester research makes it clear that enterprise leaders are weaving AI directly into core business objectives, making it the centerpiece of business strategy. Even so, leaders are careful to prioritize low-risk, high-reward applications. 

Respondents identified three top goals driving AI investment:

1. Improving customer experience (76%)
2. Enhancing operational efficiency (76%)
3. Strengthening customer retention (71%)

Other studies have consistently shown that improvements in customer experience (CX) correlate directly with increased revenue, reduced churn, and stronger competitive differentiation.

According to McKinsey & Company, for example, “...80 percent of the value creation achieved by the world’s most successful growth companies comes from their core business — principally, unlocking new revenues from existing customers.”

So, instead of using AI because it's required to stay competitive, enterprise leaders are careful to tie the new technology to measurable outcomes that affect revenue and growth. This explains why the highest-priority AI use cases are those closest to the customer. 

More than half of organizations are using AI applications for personalization (53%), automating customer resolution processes (53%), and answering customer questions (52%). These applications offer relatively low risk compared with more experimental use cases, but deliver fast, tangible return on investment.

By focusing on immediate CX improvements, companies can generate returns that fund further exploration, while also preparing teams and infrastructure for more ambitious initiatives.

Cybersecurity was a concern for 63% of respondents. AI systems introduce new attack surfaces and expand existing ones for threat actors. Threats like model poisoning, prompt injection, and data exfiltration are already being observed in practice. When customer-facing applications rely on AI for decision-making, even small vulnerabilities can cascade into significant breaches.

As governments enact new AI-driven regulations, such as the EU AI Act and U.S. frameworks and sector-specific guidelines, companies face noncompliance hurdles as they scale AI. Unlike traditional software, AI introduces unique issues like explainability, auditability, and bias mitigation. Falling short of the regulations can lead to penalties, legal exposure, and halted deployments.

AI failures tend to become public. A biased recommendation engine or hallucinating chatbot can quickly become a headline. Customers’ data privacy can be breached. For customer-facing use cases, a misstep can be more than a technical issue — it can damage the brand. Trust in companies can be costly to rebuild.

Poorly implemented AI can cause churn, lost opportunities, and wasted investment. For example, if a retail AI chatbot gives misleading financial advice, the company may face not only regulatory fines and litigation but also consumer backlash, which would undermine the retention gains AI systems were meant to achieve in the first place.

The consequences of these risks have the ability to snowball into critical business disruptions for a company, which is why leaders must be careful with AI adoption and must carefully balance the risks and rewards of innovation.

Instead of going all in on ambitious GenAI projects, many organizations are starting small. By rolling out AI in controlled use cases (like customer service automation), they gain experience without jeopardizing critical operations. This iterative model allows them to test, learn, and scale responsibly.

Building and maintaining AI infrastructure is complex and enterprises increasingly rely on managed services for model training and inference. This reduces operational burden and makes specialized expertise more accessible. 

By outsourcing risk-heavy tasks, organizations also minimize exposure while accelerating deployment. The key is to choose services from companies with similar security and risk mitigation standards as your own.

The survey data also shows that enterprises are experimenting with cloud native and open source AI technologies. Cloud native architectures offer scalability and resilience, while open source technology provides flexibility and speed, both of which are necessary for AI adoption and technology updates. The payoff is agility without long-term vendor lock-in.

Given the complexity of AI adoption, many organizations are prioritizing vendors with proven AI expertise. These partners help enterprises navigate not just technology but also compliance, scalability, and security. A strong vendor ecosystem supports AI risk mitigation strategies.

Perhaps most important, organizations are recognizing that AI governance cannot sit solely within IT. Leading enterprises are creating cross-functional governance frameworks involving compliance, legal, and security teams from the outset. This holistic approach ensures that AI innovation aligns with regulatory requirements and ethical AI standards.

The main pattern that has emerged from the data is that companies are achieving balance through structure; that is, they are building the scaffolding that allows innovation and risk management to progress in parallel.

In addition to operational structure, balanced AI adoption requires solid infrastructure. In fact, according to the Forrester study, 55% of enterprises cite technology and platform gaps as their top AI adoption challenge. 

Traditional cloud platforms were not designed for the unique demands of performance-sensitive AI workloads like inference at scale. Latency, cost inefficiency, and inadequate data locality controls all create risk — from degraded customer experiences to higher exposure in compliance audits.

Enterprises are rethinking their infrastructure strategies by exploring alternatives built specifically for AI performance. They are doing so by using platforms optimized for distributed workloads, low-latency inference, and security by design. 

Put simply, infrastructure isn’t just the foundation for AI technology; it’s an active part of the risk equation. Choosing the wrong foundation can amplify vulnerabilities. Choosing the right one can transform AI from a liability into a long-term competitive advantage.

The journey of enterprise AI is evolving quickly. Early adoption centered on narrow use cases, then the generative AI wave expanded to content creation, analytics, and automation. 

Now, we’re on the cusp of the next phase: agentic AI. Agentic AI makes up systems with autonomy that act on behalf of humans and make decisions based on existing context. This evolution promises further gains, but also, of course, unprecedented risks.

Each phase of AI increases both the potential upside and the stakes of failure. Enterprises that succeed will be those that are currently building organizational fluency through iterative adoption by investing in resilient, performance-optimized infrastructure and embedding governance and compliance from the start.

Now is the time for technical decision-makers to reassess AI solutions and cloud strategies, particularly their infrastructure choices, to ensure they can scale responsibly and with ethical considerations in mind.

For a deeper look at how 400 senior decision-makers are approaching AI development and adoption, download the full Forrester report, State of Enterprise AI: Gaining Experience and Managing Risks.