The delivery of computing resources, including storage, processing, databases, and software, over the internet on a pay-as-you-go basis rather than through owned hardware. Cloud computing is the infrastructure layer that makes AI tools accessible to agencies without requiring data center investment.
Also known as cloud infrastructure, cloud services, cloud platforms
Cloud computing provides on-demand access to shared pools of computing resources managed by cloud providers such as Amazon Web Services, Google Cloud, and Microsoft Azure. Resources are provisioned and released programmatically, allowing workloads to scale up during peak demand and scale down when not needed. This elasticity is what makes training large AI models economically feasible: a job requiring thousands of processors can run for hours and then release those resources rather than requiring permanent ownership.
Cloud services span several layers. Infrastructure as a Service provides raw compute and storage. Platform as a Service provides managed environments for application development. Software as a Service delivers fully managed applications through a web interface. Most AI tools agencies use are delivered as SaaS products running on cloud infrastructure, with the cloud layer invisible to the end user.
The foundation models powering modern AI tools were trained on cloud infrastructure at scales that would be impossible for any single organization to maintain independently. The shift to cloud has democratized access to AI capabilities that previously required research institution-level resources.
Agencies do not manage cloud infrastructure directly, but they are responsible for advising clients on AI tool selection, data handling, and technical integration. Cloud computing concepts surface constantly in those conversations: where is client data stored, what happens to it when a subscription lapses, how does the tool scale during campaign peak periods?
Data residency affects compliance. Cloud providers offer regions in different geographies, and client data stored in certain regions may be subject to different legal requirements. For clients in regulated industries or with international operations, asking which cloud region an AI tool’s data is stored in is a legitimate due diligence question, not a technical detail to defer to IT.
API-based AI tools are cloud-dependent. Every AI tool that calls an external API is dependent on that provider’s cloud uptime. Agencies building AI into client workflows should understand the SLA guarantees of the tools they are integrating and have contingency processes for when those APIs are unavailable.
Cost models need to be understood before deployment. Cloud-based AI tools often price by usage: tokens processed, API calls made, compute hours consumed. Agencies deploying these tools at campaign scale need to model usage volume before committing, because unexpectedly high usage can produce unexpectedly high bills that were not in the client budget.
An agency recommends a cloud-based AI content generation tool to a financial services client. Mid-deployment, the client’s compliance team asks where the generated content and input prompts are stored, how long they are retained, and whether they are used to train the vendor’s models. The agency does not have immediate answers and must go back to the vendor for documentation. The experience leads the agency to add a cloud data handling checklist to their AI tool onboarding process: data residency, retention policy, model training data usage, and SLA guarantee are all questions that should be answered before a client contract is signed.
The generative AI foundations module of the workshop covers how today’s models work, what they can and can’t do, and how to choose between them.