Most organizations have turned to AI for the same reason: to automate manual processes to reduce costs. The Telarus Tech Trends Report found 84% cite this as their primary investment driver.
The reality is more complex.
For example, organizations implementing AI to automate and reduce manual processes often find cloud infrastructure costs surge to support processing-intensive workloads. In other words, what starts as an efficiency pursuit becomes a budget expansion.
“If a client has cloud-based technology or they’re trying to incorporate something for their portal, they would need a decent amount of cloud infrastructure behind it to power that,” explains Owen Hood of SpyGlass, Vice President of Analytics. “The cloud isn’t cheap.”
SpyGlass provides critical insight into the potential benefits and risks of using AI for technology expense management (TEM) to help plan a smarter path to achieve real savings and operational control.
The growing cloud infrastructure problem reveals a bigger issue with AI. Harvard Business Review recently found “AI-generated workslop,” or wasted output, creates more work instead of eliminating it.
In fact, Harvard reports that while workers are largely following mandates to embrace the technology, few are seeing it create real value. And while employees are using AI tools to create low-effort, passable-looking work, it ends up creating more work for their coworkers.
Using AI for TEM faces the same concerns, in addition to extra costs.
Organizations deploy AI tools expecting to save money and streamline operations. Instead, they run into huge computing demands that require:
These computing needs can drive up monthly cloud bills fast. Companies managing dozens of technology providers find that AI overhead grows with complexity, not savings.
And it’s important to remember AI systems are only as good as the data they are trained on. If the underlying data is inaccurate, incomplete, or biased, the AI’s outputs will reflect those flaws. IT professionals must ensure that the data fed into AI systems is clean, accurate, and representative to avoid misleading results.
The false promise of AI hits hardest for budget-conscious organizations. The companies trying to use AI to manage tight budgets end up facing bigger infrastructure cost increases compared to their current technology spending.
These infrastructure costs represent the beginning. The real danger lies in where AI tools consistently fail, creating three critical blind spots that turn efficiency investments into budget drains.
Did You Know?
Thomson Reuters
“Sensitive financial data processed by AI systems faces increased cybersecurity threats, including data poisoning and adversarial attacks meant to destabilize models and leak private information.”
“You’re never going to get any kind of results by taking a legacy-type invoice and having AI try to interpret it,” Hood explains.
The problem is widespread. Most organizations maintain a blend of old and new technology services, where legacy billing platforms hide crucial information within:
These systems demand expertise that automated tools can’t replicate.
AI tools excel at single invoices. Reality is messier.
“Most of our clients have 20 invoices with 20 different providers,” Hood notes. “Our value is that we have extensive experience in analyzing all of them together.”
This comprehensive cross-provider analysis reveals savings opportunities that AI tools miss entirely. While an automated system might optimize one vendor’s services, it can’t identify redundant services across platforms or make strategic recommendations spanning multiple providers.
One of the most notable issues AI raises is the “black box” problem, or the lack of transparency in how complex AI models, especially those using deep learning techniques, make decisions. While users and operators can see the inputs fed into the AI system and its outputs, they can’t see or explain the internal reasoning, calculations, or logic the system used to arrive at an answer or conclusion.
Industries most affected by black box AI risks include healthcare, finance, criminal justice, insurance, recruitment, and autonomous transportation. In these fields, decisions made by AI can have significant consequences for individuals, organizations, and regulatory compliance.
One of the most serious “black box” consequences is compliance. AI-generated recommendations lack contextual understanding for compliance-sensitive environments like:
Worse yet? Finance teams can’t trace AI logic or verify compliance implications. This creates audit vulnerabilities that human oversight prevents. It also leads to budget exposures that organizations discover too late.
The discovery process reveals where human expertise consistently outperforms AI tools. It’s about asking the right questions before analyzing a single invoice.
“I think the main issue is you have to determine what your goal is,” Hood explains when discussing AI audit limitations.
AI systems can’t access organizational context, yet this background is what determines which technology audit recommendations will actually work.
SpyGlass begins every engagement with a comprehensive discovery meeting that Hood calls essential. “Many times, when clients hire us, they have specific objectives they’re trying to achieve,” he explains. This personalized approach ensures recommendations align with actual organizational capabilities, something automated analysis can’t replicate.
“If your goal is to get as thorough as possible, you really need a professional who can look at the entire picture. There are too many variables in play that will not be in the data.”
Owen Hood, Vice President, Analytics, SpyGlass
AI tools excel at pattern recognition and basic cost categorization. Smart organizations recognize these strengths while acknowledging their limitations.
The strategic advantage lies in hybrid thinking.
Organizations succeeding in technology expense optimization don’t choose between AI and human expertise. They combine both strategically.
While AI is a powerful tool for prediction and analysis, it can’t replace human judgment. Leading organizations use a hybrid approach: Automated AI is used to screen for scale, while incorporating human judgment for high-impact decisions, exception resolution, and continuous improvement.
Evidence from a field study led by Stanford University and MIT provides strong evidence supporting the statement that AI plus human collaboration consistently outperforms AI-only or manual review in audit quality, policy enforcement, fraud detection, and user experience within accounting and expense management workflows.
“If your goal is to get as thorough as possible, you need a professional who can look at the entire picture,” Hood explains. “There are too many variables in play that will not be in the data.”
Finance leaders implementing successful TEM programs focus on three operational priorities:
So the choice isn’t whether to use AI for technology expense management. The choice is how to combine AI capabilities with the human expertise and trusted advisors necessary for strategic optimization.
Ready to see what AI tools miss?
Get a SnapShot Audit that combines advanced analytics
with personalized expertise.
