Software audits have always been about visibility
For years, organizations have viewed software audits as a licensing issue.
The process was relatively familiar. A software vendor would request deployment information, review infrastructure, analyze usage data, and compare actual utilization against contractual entitlements.
The details varied by vendor, but the underlying principle remained the same: the vendor needed visibility into how its software was being used.
Historically, obtaining that visibility was often difficult. Today, that is changing.
Artificial intelligence is creating new forms of telemetry, new sources of usage data, and unprecedented visibility into software consumption.
While much of the conversation surrounding AI focuses on productivity and innovation, organizations should also recognize another reality: AI may significantly increase the amount of information vendors can collect regarding product usage.
For software publishers, that visibility creates opportunities. For customers, it creates new governance and compliance challenges.
Understanding telemetry
Telemetry refers to information collected about how software is used. Examples include:
- Feature utilization
- User activity
- System performance
- API calls
- Configuration settings
- Service consumption
Most modern software platforms already collect telemetry. Organizations encounter telemetry every day through SaaS applications, cloud platforms, monitoring tools, and subscription services.
Telemetry is not inherently problematic. In many cases, it improves product reliability, security, and support.
However, telemetry also provides vendors with insight into customer behavior. The more detailed the telemetry becomes, the greater the visibility.
Software audits have always relied on visibility
Historically, one of the biggest challenges facing software vendors was determining how products were actually being used.
Many traditional software environments were deployed entirely on-premises. Visibility was limited.
Audits often required:
- Manual discovery
- Customer-provided inventories
- Configuration reviews
- Deployment analysis
This process could be time-consuming for both parties.
Cloud computing changed that dynamic. AI is accelerating it further.
AI generates new categories of usage data
AI systems create significantly more activity than traditional software applications. Examples include:
- Prompt submissions
- Model interactions
- Agent actions
- API requests
- Vector searches
- Context retrieval operations
- Token consumption
- Workflow execution logs
Every interaction creates data. Every workflow creates data. Every agent action creates data.
Many organizations focus on what AI can do. Fewer focus on what AI records.
AI features create new visibility
Consider a traditional software application. A user logs in. The user performs a task. The application records limited information.
Now consider an AI-enabled application. The platform may record:
- Which AI features were used
- How frequently they were used
- Which users initiated requests
- Which models were accessed
- How much data was processed
- How many tokens were consumed
- Which workflows were executed
The amount of available information increases dramatically.
For vendors, this information provides valuable product insights. For compliance teams, it creates a much larger data set that may eventually influence licensing reviews.
Consumption models depend on telemetry
Many AI services are priced using consumption-based models. Examples include:
- Token usage
- API requests
- Agent execution
- AI credits
- Inference workloads
- Compute consumption
Unlike traditional perpetual licensing models, these services require continuous measurement. The vendor must track utilization to calculate charges. That tracking naturally increases visibility into customer behavior.
Organizations adopting AI should recognize that many consumption models are fundamentally dependent upon telemetry. Without telemetry, the billing model cannot function.
AI agents create additional audit signals
AI agents introduce another layer of complexity.
Traditional software usage is driven by people. AI agents may operate continuously.
An agent can:
- Retrieve information
- Execute workflows
- Query databases
- Access APIs
- Update records
- Trigger additional processes
Each action creates additional records. As organizations deploy more autonomous systems, the volume of telemetry expands accordingly.
From a vendor perspective, agent activity may provide significantly more insight into product utilization than traditional user activity ever did.
The expanding definition of "usage"
Historically, software audits often focused on relatively straightforward questions:
- How many processors are deployed?
- How many users have access?
- Which products are installed?
AI introduces additional dimensions. Organizations increasingly need visibility into:
- Feature usage
- Agent activity
- AI service consumption
- Model utilization
- API interactions
- Data processing volumes
The concept of software usage is becoming broader and more complex.
This does not necessarily mean audits will become more aggressive. It does mean vendors may have access to more information than ever before.
Why this matters across multiple vendors
This trend is not limited to a single software provider. Major vendors are embedding AI throughout their product portfolios:
- Oracle AI Services and Database 23ai
- Microsoft Copilot and Azure AI
- Salesforce Agentforce
- ServiceNow AI Agents
- SAP Joule
- Adobe Firefly
- Workday Illuminate
Each platform introduces new AI capabilities. Each capability creates additional utilization data. Each utilization record contributes to a broader picture of customer activity.
The specific licensing implications vary by vendor. The visibility trend remains consistent.
Governance is becoming more important
The challenge for organizations is not merely understanding AI. It is understanding AI utilization.
Questions leadership teams should begin asking include:
What AI services are currently enabled? Many organizations discover AI functionality after deployment has already begun.
What telemetry is being generated? Understanding what data exists is essential.
How is consumption measured? Consumption-based models often differ significantly from traditional licensing structures.
Who is monitoring utilization? Without oversight, organizations may struggle to identify cost drivers or compliance risks.
What contractual rights govern collected data? Technology teams should understand both operational and contractual considerations.
The next evolution of software audits
Software audits are unlikely to disappear. If anything, they may become more data-driven.
Historically, vendors often had to reconstruct software usage after the fact. Future environments may contain detailed records of:
- AI interactions
- Feature utilization
- Agent activity
- Service consumption
- Platform usage patterns
This does not automatically create compliance problems. It does increase the importance of governance.
Organizations that maintain visibility into their own environments will be better positioned to understand what vendors may see as well.
The bottom line
Artificial intelligence is changing more than user experiences. It is changing how software utilization is measured.
Every prompt, workflow, agent action, and AI interaction creates additional telemetry. That telemetry helps vendors improve products, deliver services, and support consumption-based pricing models.
It also creates new forms of visibility.
For organizations, the lesson is straightforward. AI governance is not simply about model accuracy, security, or ethics. It is also about understanding how AI changes software utilization, software measurement, and software compliance.
The companies that establish visibility into AI usage today will be far better prepared for the licensing, cost, and governance challenges that emerge tomorrow.
