# What We Can Learn From the Cloud Journey to Understand Today’s AI Adoption Curve > Source: > Published: 2026-08-04 17:27:45+00:00 One of the greatest challenges organizations face today is adopting artificial intelligence (AI) to strengthen cybersecurity. The rapid rise of AI has generated both excitement and concern as organizations begin implementing it within their environments. If we rewind 15 or 20 years, there are several parallels we can apply from the cloud maturity journey to help us navigate this major technological shift today. **Market Complexity** Cloud computing emerged when organizations were hyper-focused on building data centers. Early adopters of cloud computing sought ways to make life easier and scale quickly. Moving to the cloud eliminated the hardware problem and the challenges with building virtualization farms – cutting months of procurement and deployment, not to mention the cost of ongoing management. However, customers had to evaluate a plethora of options: - Delivery models including Infrastructure as a Service (IaaS), Platform as a Service (PaaS), and Software as a Service (SaaS). - Deployment models, including public, private, and hybrid, while maintaining on-prem infrastructure. - Different cloud vendors with different strengths and weaknesses. - Storage services and sharing applications. Similarly, early adopters of AI are drawn by the promise of a fast, easy way to complete tasks and create things they couldn’t before. They also have a range of platforms and vendors to choose from, each with its own sweet spot. What’s more, companies further along the adoption curve are starting to expand beyond AI models to agentic AI, using different AI agents and piecing tasks together to complete entire workflows. History has shown that when usage increases amidst market complexity, we start to see unintended consequences. **Security ** As organizations increasingly adopted the cloud, they quickly realized security challenges that needed to be addressed, especially access and configuration management. Overly permissive sharing policies and access control lists allowed unauthorized access to sensitive data in S3 buckets or Azure storage blobs and left the door open to hackers. Additionally, losing control and visibility when moving data from onsite servers to the cloud raised compliance concerns. Similarly, as AI adoption expands, users are starting to encounter biased information and hallucinations, which raise questions about data sources, transparency, and the ability to verify outcomes. Compliance risks are also a concern. Sending data offsite for processing, analysis, or storage may not meet privacy and security requirements and may expose the organization to risk. As we have already seen in the early days of enterprise AI adoption, users have unintentionally uploaded sensitive data to public models resulting in unauthorized access and disclosure of sensitive data. **Cost** The shift from hardware amortization to subscription models for cloud resources reduced the need for large, upfront investments but introduced variability. Cloud providers charged based on the amount of storage used in their environments, so organizations optimized the amount of data stored by shifting from one provider to another. Over time, service providers adjusted their models to promote retention, charging less for storage and more for data transfer in and out of their cloud. This approach increased uncertainty, forcing organizations to re-evaluate their cloud strategies altogether to optimize their operational costs. [Token economics](https://www.finops.org/insights/token-economics-the-atomic-unit-of-ai-value/) has introduced similar uncertainty to AI usage. Initially, AI was largely free to encourage users to explore the technology by completing minor tasks. Soon, token models opened additional functionality and enabled new usage-based costs. As a result, organizations focused on prompt engineering to drive efficiency and manage consumption. As models keep evolving, additional factors like context and memory, model selection, output length, and retry and orchestration overhead, increase costs and make forecasts unreliable. The latest models exacerbate uncertainty by imposing upcharges for API integration and GenAI models with complex inference capabilities. Once again, organizations are forced to re-evaluate their spending models as the changes make their costs less predictable and harder to budget for. **Organizational Roadblocks** When the cloud was first introduced, the skills gap, fear of vendor lock-in, and cumbersome security questionnaires slowed adoption. Organizations, rightfully so, were worried about the location, use, and protection of their data. Today, similar organizational roadblocks hinder AI adoption. - On a technical level, teams need to learn new skills such as prompt engineering to improve ROI from AI models and model training and monitoring to ensure output reliability. - To prevent vendor lock-in and maximize value, teams must understand which models align best with specific business use cases. In some instances, organizations choose to develop their own large language models (LLMs) and AI agents. However, proprietary tools are notoriously costly to develop and maintain. - AI security questionnaires are essential tools to validate that AI models operate in a secure and compliant manner. Just as with cloud computing, auditing what, how, and where data is used and protected within the AI models is an important hurdle to overcome. **Recommendations** Considering cloud maturity patterns, organizations looking seriously at adopting AI more broadly and evaluating opportunities within cybersecurity and network operations should consider these factors: **Identify your primary business use cases.** Don’t apply AI to everything. Start with valid business cases – important tasks that either can’t be done now or need to be done more efficiently and reliably than humans alone can do, such as reducing vulnerability risk and accelerating compliance reporting and remediation. Align your AI tool selection accordingly.**Proactively manage security risks.** Understand what data leaves your organization, who can access it, and what could happen if it were exposed more broadly or if accessed by unauthorized individuals. Ensure that the AI vendor’s policies around governance, risk, and compliance match your own. Identify suitable controls and safeguards to keep your data private and secure.**Focus on the true costs of using AI.** There is a quantifiable value from completing a task 10x faster, but the costs factored into the equation must lead to a verifiable ROI. Ensure you understand the total cost of ownership and make informed decisions accordingly.**Ensure accountability for reliability and accuracy.** At the end of the day, the AI might be used to increase efficiency, but it’s the person using it who will be accountable for the output. Therefore, the output from the AI must be explainable, and human-in-the-loop oversight to review, override, and adjust data and processes is necessary to improve the AI’s performance and the reliability of its outputs. Organizations that adopt AI with careful and deliberate strategies – with an emphasis on responsibility, safety, and security – will maximize the value of AI to the business. Just like the cloud, they’ll soon wonder how they ever managed without it.