Home > Blog > Best Practices for Managing AI Spend: Lessons Learned from Early Adoption

Best Practices for Managing AI Spend: Lessons Learned from Early Adoption

AI Is Not Just a Technology Decision. It's a Financial Management Discipline.

As businesses rapidly adopt generative AI tools, many have discovered that the first challenge is adopting the use of AI at scale. The next challenge is understanding, governing, and optimizing the cost of that usage.

As Stoneridge has expanded our own use of Microsoft Copilot, agents and other AI solutions, we’ve had valuable reflective conversations across leadership, IT, and operations about what it takes to manage AI spend effectively. Those discussions revealed several practical lessons learned that might help any organization navigating a similar journey. 

1. Start with Visibility Before Governance

One of the earliest priorities for AI adoption should be establishing usage visibility. 

Before implementing hard limits or restrictions, organizations need to understand: 

• Who is using AI 
• How frequently it is being used 
• Which scenarios generate the highest consumption 
• Overall organizational consumption patterns 

Resist the urge to govern before you understand. Start by measuring how AI is being used across your organization. Establishing visibility into consumption patterns provides the foundation for effective budgeting, governance, and long-term optimization. 

Best Practice:

Spend the first month measuring before optimizing

2. Establish a Budget Early, Then Refine

As we expanded our own use of AI, we quickly recognized the need for a defined monthly spending threshold. 

Rather than trying to determine the perfect budget from the start, we established an initial limit and used real usage data to guide our decisions. As we gained a better understanding of adoption patterns and business value, we adjusted our approach accordingly. For organizations on a similar journey, the lesson is simple: start with a reasonable budget, learn from the data, and refine as your understanding grows. 

Best Practice:

• Set an initial organizational spending cap
• Review usage monthly
• Adjust based on demonstrated business value rather than theoretical estimates

3. Understand Which AI Activities Actually Cost Money

A key lesson was that not every AI interaction carries the same cost. 

Key learnings highlighted the distinction between: 

Lower-Cost Activities:

• General AI chat 
• Drafting content 
• Internet-based research 
• Simple prompt interactions 

Higher-Cost Activities:

• Reasoning across enterprise data 
• Cross-document analysis 
• File comparison 
• Meeting and email synthesis 
• Agent-based workflows 

The highest cost scenarios tend to involve AI reaching across organizational knowledge repositories and performing multi-step reasoning. 

Best Practice:

Educate users on selecting the appropriate tool for the task instead of defaulting to the most expensive AI workflow. 

4. Create a "Use the Cheapest Effective Tool" Mindset

An effective model emerged: 

The principle is simple: 

Don't use enterprise reasoning when simple retrieval is sufficient. 

Using AI only when analysis or synthesis is required can significantly reduce spending without sacrificing productivity. 

5. Track Usage at the User Level

Usage reporting quickly identified significant differences in consumption among users. 

Rather than assuming excessive usage was misuse, we looked at business context behind the data. In many cases, the heaviest consumers were performing legitimate, high-value work. This reinforced the idea that usage metrics should drive conversations, not punishment. 

Best Practice:

Review high-consumption users to understand value creation before taking corrective action.

6. Tie AI Consumption to Business Outcomes

A recurring theme in the AI adoption reflection was the need to evaluate spending through a business lens. 

Questions leaders should ask: 

• Is AI reducing project effort? 
• Is it accelerating activities? 
• Is it improving documentation quality? 
• Is it generating reusable intellectual property? 
• Is it improving client outcomes? 

If AI produces measurable value, increased spending may be justified. 

Best Practice:

Measure cost and value together.

7. Plan for Greater Cost Visibility

As AI becomes embedded within business processes, organizations will eventually want greater visibility into how costs are distributed across projects, teams, and initiatives. 

One challenge we identified is that many platforms can show costs by user, but not always by project or business activity without additional tracking processes. Until reporting capabilities continue to mature, organizations may need to supplement platform reporting with their own processes to gain a more complete picture. 

Best Practice:

Develop a future roadmap for: 

• Project attribution 
• Cost visibility across teams and initiatives 
• Measuring AI usage against business value 
• Department and organizational budgeting 

Even if implementation comes later.

8. Watch Credit Allocations Carefully

One unexpected lesson came from administrative configuration. 

Credits had been allocated to multiple environments, which unintentionally reduced the available organizational pool and accelerated pay-as-you-go consumption. After correcting the allocation strategy, prepaid credits became available again. 

Best Practice:

Regularly review: 

• Environment allocations 
• Reserved credits 
• Department-specific credits Consumption policies 

Small configuration choices can have significant budget impacts.

9. Create Cross-Functional Governance

The most successful discussions on AI adoption involved collaboration between: 

• IT 
• Operations 
• Finance 
• Service leadership 
• Executive leadership 

AI adoption impacts all of these groups and cannot be managed effectively by IT alone. 

Best Practice:

Create a recurring AI governance forum that reviews usage, value creation, risk, and spend.

10. Treat the First Year as a Learning Year

Perhaps the most important takeaway was the team's mindset. 

Start somewhere. Learn. Adjust. 

Rather than attempting to predict every cost or perfectly engineer governance upfront, we focused on gathering data, educating users, measuring outcomes, and continuously refining their approach. 

Organizations that approach AI with a similar iterative mindset will likely achieve faster adoption and better financial outcomes than those attempting to control everything before learning. 

What We Learned About Managed AI Spend

As we reflected on our own AI adoption journey, one thing became clear: effective AI cost management isn't about controlling every dollar from day one. It's about creating visibility, learning from real-world usage, and making informed adjustments over time. 

Successful AI cost management comes down to five core principles: 

1. Measure before you optimize. 
2. Set spend limits and review regularly. 
3. Use the least expensive tool that solves the problem. 
4. Connect consumption to business value. 
5. Continuously educate and govern as adoption grows. 

AI adoption and cost management are continuing to evolve, so if you have questions about managing AI spend, governance, licensing, or building a sustainable AI strategy, the Stoneridge team is always here to help!

Our Verified Expert

Michelle Schumacher

Michelle Schumacher is the Chief Financial Officer for The Stoneridge Companies, overseeing the strategic financial direction and IT strategy for Stoneridge and Levridge. Prior to Stoneridge, Michelle spent seven years at Microsoft where she and her team were responsible for complex technical accounting for Microsoft’s cloud and AI product offerings, along with Microsoft’s cloud infrastructure. During her 5 years in Microsoft’s Corporate Accounting group, she was part of the team early adopting the new revenue standard (ASC 606) for Microsoft’s approximately $100B in revenue. Prior to joining Microsoft in 2011, Michelle’s career was spent in public accounting, both at Eide Bailly and PwC.

Read More from Michelle Schumacher

Under the terms of this license, you are authorized to share and redistribute the content across various mediums, subject to adherence to the specified conditions: you must provide proper attribution to Stoneridge as the original creator in a manner that does not imply their endorsement of your use, the material is to be utilized solely for non-commercial purposes, and alterations, modifications, or derivative works based on the original material are strictly prohibited.

Responsibility rests with the licensee to ensure that their use of the material does not violate any other rights.