How to cut runaway costs by optimizing tech spend

Handing you a blank map and saying “Look out for inefficiencies in your tech stack!” isn’t good enough. Inefficiency must be identified, addressed, and analyzed. That’s what we’re going to do here together. 

There’s inefficiency in the cloud, where overprovisioning, underutilization, and misconfiguration can run rampant and derail productivity. There are inefficient data tools that are used or not used across your IT environment, affecting adoption rates (the tech fatigue is real). And then there’s the seen and often unforeseen data-use expenses that eat budgets like candy.   

Without a proactive approach, data initiatives will always face obstacles and become bloated, making them prime targets for defunding or cancellation. As a Data Leader, you need to know what those problem areas are and what to do about them. 

Thankfully, no matter what industry you’re in, there are a few areas to search. But first, our next Data Leader Scenario provides a powerful proof point.

Data tools: all good. Fragmented, duplicate data tools: not cool. The fix: reduce the bloat. 

According to one study, nearly 70% of employees add more than 20 hours per week looking for tools and information spread across multiple silos.¹³ Yikes! But that’s the truth about data tool fragmentation, making it difficult for data teams to access and use technology throughout your organization. Surprise, surprise: Time-consuming tasks make it harder for people to move fast and make informed business decisions. Just ask the marketing team at Georgia-Pacific. Even worse, fragmentation could leave you vulnerable to compliance issues if you’re working with private or sensitive data. 

And when 25% of data teams already report difficulty implementing AI tools,¹⁴ data tool fragmentation just makes things even more inefficient, adds overlapping costs, and hurts AI adoption rates. 

1. Audit your data and analytics: Look for tools that overlap in purpose or across departments (especially across departments) — these overlaps happen over time as functionalities or projects are added. Often, teams aren’t working with strict data quality or governance that monitors and controls how and which tools are used. Where do you even begin with an audit? Skip down to the Data Leader Tip for pointers.

2. Use enterprise-wide data and analytics solutions: With end-to-end visibility and scalability, structured and unstructured data becomes easier to access and understand, whether it’s stored in lakehouses and warehouses, or even in open lakehouses. Plus, in addition to enabling cross-functional capabilities, having one solution helps reduce maintenance, licensing, and integration fees. That’s money you can redistribute elsewhere: to things like hiring more people or investing in ongoing training.

Next up, cloud inefficiencies. In general, we’re referring to resources that are over-provisioned, underutilized,or misconfigured. In other words, you aren’t using what you have, you don’t have enough, or what you have isn’t working hard enough, leading to anything from increased costs and performance issues to security vulnerabilities. Fun fact: Even in 2024, 78% of companies still had less than 75% of their cloud spend allocated,¹⁵ meaning any positive management steps you make now put you ahead. 

One area to look at more closely is the cost of analytics queries. This is one of those areas where not all tools are created equal. Some solutions offer in-memory query caching and indexing. Not only does this minimize the number of queries sent to the underlying data warehouse, avoiding additional costs, but it also results in faster query execution, improved availability, and better user experience.  

The truth is, data is only getting more expensive. GenAI queries, for example, use huge amounts of compute power, resulting in heavy investment in back-end infrastructure. Factor in the increased demands on GenAI and agentic models, and you’re looking at higher energy costs than any other IT-related initiative by 2027.¹⁶  

As part of the governance policies discussed in Part 2, some organizations include energy costs and ways to mitigate uncontrolled spending. Platform features like real-time monitoring and automated alerts on usage anomalies can support regular audits to avoid excess expenses. Plus, most data vendors are now offering pricing plans that ensure your team only pays for the resources you use, rather than paying a fixed rate per user per month. If you’re not already in a usage-based pricing model, talk with your platform provider, as it can help eliminate dead costs. 

It’s a double-edged sword that Data Leaders know all too well: Do I focus on performance at the expense of costs, or do I rein in costs at the expense of performance? Our answer: Don’t compromise on either performance or costs.  

AI is hungry for compute. But many ML models degrade quickly, requiring costly retraining. Fragmented data inflates storage needs, and AI-ready data preparation is often an afterthought. 

Now, companies are trying to find the sweet spot between moving fast and being smart with their money. It’s not just about cutting costs; it’s about spending in the right places and making sure teams aren’t over-provisioning or duplicating resources. Check out how Steinemann approaches this in our next Data Leader Scenario. 

As Steinemann shows, companies of all shapes and sizes are adopting AI tools to stay competitive, and that means one thing: Spending is going up. A lot. And not just on the AI tools themselves — cloud costs are rising too. 

Cloud platforms are the go-to place for running AI models, storing massive data sets, and spinning up computing power on the fly. They make it easy to experiment and scale fast, but they can also lead to sticker shock when the monthly bill hits. As AI projects grow, so does the need for serious horsepower, and cloud usage (and cost) tends to follow. 

Also, keep in mind that once all your information is “in,” you need to prep it for AI, which could mean even more fragmented solutions and costs, unless, as with Steinemann, these solutions are built in.

In an ideal world, you could feed AI models any dataset, and it could figure out the good stuff from the bad. Sadly, that’s not the case (at least for now). Until then, data should meet certain criteria so AI can use it right away, earning a time savings of 50%, or $2.5 million in value, when it comes to data.¹⁸ Remember from Part 2: Being AI-ready means having data that’s diverse, timely, accurate, secure, discoverable, and consumable. 

Besides, who wants to spend hours manually cleaning or moving data around? No one. Tools with low-code or no-code AI features can handle a lot of this grunt work — like data transformation, deduping, and formatting — so your team can focus on actual insights. Bonus: They also reduce mistakes and speed everything up. 

Another thing people don’t like waiting for: batch processing. At least not when you can get insights as the data comes in. If you’re using streaming tools like Kafka or Kinesis, setting up real-time ingestion means your AI models always have fresh data to work with. It’s faster, cuts down on extra storage, and helps your team react in the moment. 

Finally, avoid data waste at all costs. If everyone has access to all your data, there’s a high chance you’re duplicating data, running up storage costs, and increasing compute costs when querying the cloud data warehouse. A simple step like setting up role-based access helps keep things nice and tidy. Just make sure the right teams have what they need. That way, you’re not paying for unused or irrelevant data sitting around in your systems.

Here’s the thing about cost optimization. It needs to happen now. It needs to happen later. It needs to happen often. And if that doesn’t happen, then you risk costs getting complicated again, fast. 

But let’s also be clear. This strategy isn’t just about money. It’s about creating a sustainable business model from a scalability and growth perspective, because without a firm plan in place, it’s difficult to see, understand, and navigate the unseen forces that drive data and AI inefficiency. 

Our final Data Leader Scenario highlights the value of building a sustainable cost strategy. 

AI workloads are always hungry. And if you’re not careful, models can burn through your budget faster than you can say “Hey, stop eating up my budget.” A lot of teams throw way too much compute at the problem or retrain models more often than they need to, without really stopping to ask, “Is this actually worth it?” It’s like revving a sports car in traffic: flashy, but not super-efficient. 

Instead, you can leverage flexible pricing models and cost tracking, not subscription models that have fixed rates regardless of usage. Also known as consumption-based pricing, this offering is a relatively new and popular option for organizations looking to only pay for what they use. Its popularity makes sense, considering how much flexibility (e.g., scalability) it offers businesses without adding risk. 

Data silos are another huge cost culprit. When your data pipelines and analytics tools aren’t in sync, things get clunky. You end up moving the same data around more than necessary, repeating queries, and watching shadow IT costs quietly pile up in the background. All that fragmentation makes it hard to see the true cost, and even harder to keep overhead under control. 

Consider consolidating your tools into one, AI-driven platform. Then use that platform to create real-time dashboards that can help track data usage and spending and detect cost anomalies before they chew up irreplaceable budget. Then, throughout various departments, create more applications designed to spot unknown or unexpected spending before it snowballs, and uncover hidden pockets of efficiency across your business.  

The reality is that tech budgets are spread all over the place: marketing’s got a chunk, IT has some, and maybe even HR’s hoarding a bit. It all adds up. Without clear leadership, it’s a bit of a free-for-all. That means teams end up making random investments, and it’s a territorial nightmare to figure out how much is being spent. 

Opportunity is knocking. Seize the moment by helping to ensure that your AI and cloud initiatives align with your business goals. Make sure all project stakeholders have a thorough understanding of the expenses associated with AI development and cloud deployment costs for storage, software, data transfer, and operations. And check that you’re able to quantify those expenses against cost savings, revenue increases, and other KPIs.

Don’t blink. If you do, you risk letting data users (aka everyone) run wild with your budget — or worse, derailing your profitability, productivity, and reputation all the way up to the executive level. 

All good, though. Ensuring you get nothing but the best from your data, integration, and AI initiatives requires a keen sensitivity toward inefficiency and thoughtful planning. You got this.