In the rush to embrace AI, many businesses have prioritized deploying AI wherever and whenever they could make a case for doing so. Ensuring that AI solutions were necessary and cost-effective tended to be less of a focus during the AI adoption stage.
But as AI applications, services and agents become commonplace components of IT estates — and as AI accounts for an increasingly large share of enterprise budgets — it is becoming harder to justify AI investments that don’t create real value.
That is driving the need to rationalize AI assets — in other words, to assess them, identify instances of AI waste or suboptimal use and take steps to maximize the efficiency and ROI of AI.
Read on for guidance as we unpack what AI rationalization means, why it’s important and which actionable steps business and IT leaders can follow to align AI investments with organizational needs.
What is AI asset rationalization?
AI asset rationalization is the practice of evaluating an organization’s AI systems to determine how much value they bring to the business.
If an AI resource is under-delivering, the organization should either find ways to boost the value it derives from the investment or — if improvements aren’t possible — decommission it.
Why AI asset rationalization matters
Rationalizing AI assets is important for enterprises because AI accounts for an increasingly large share of enterprise IT spending, even as many executives still struggle to identify meaningful business value from their AI investments.
This mismatch likely stems in part from poor optimization of the way enterprises are leveraging AI due to issues like:
-
Paying for high-cost AI solutions when less expensive but equally capable ones are available.
-
Purchasing AI products that offer redundant or overlapping functionality.
-
Failing to maximize the number of users or processes that benefit from the AI investments a company has made.
-
Failing to adapt business processes to maximize the value of AI deployments.
AI asset rationalization addresses these issues by providing a way for companies to assess their AI investments and the way they are using them, then determine how to leverage AI assets in more impactful and cost-effective ways.
AI asset rationalization example: Customer service chatbots
As an example of AI asset rationalization, imagine a business that has implemented a generative AI customer experience chatbot designed to resolve customer queries without requiring manual intervention by staff. To rationalize this resource, the business should assess:
-
How much it cost to build or buy the chatbot.
-
How much the company spends managing and maintaining the chatbot.
-
How much value the chatbot creates — in terms of staff time savings, customer experience improvements or other metrics that directly affect business outcomes.
Too many handoffs to human agents
A rationalization assessment of the chatbot might find that a majority of customer interactions that start via the chatbot end up having to be redirected to human agents. This would imply that the chatbot is under-delivering because it’s not achieving its intended goal of minimizing the time that staff spend on customer support.
In response, the business might modify the chatbot. For example, it could connect it to a better large language model (LLM) as a way of improving the accuracy and effectiveness of information shared with customers.
Misalignment between AI models and workflows
Alternatively, the organization might determine that the problem isn’t with the chatbot’s AI technology but rather with a misalignment between what the chatbot can do well and how the business is using it. It could be the case, for instance, that the business receives complex customer requests that no chatbot — even one powered by the most modern, full-featured LLM — can handle reliably without human assistance. In that case, the business might conclude that the chatbot is a losing investment and choose to stop using it.
Inefficient business processes
Another possibility is that the business processes surrounding how the chatbot is used are the problem. For example, the chatbot may fail to resolve customer issues automatically in many cases because doing so requires triggering other workflows (like pulling data from a CRM system) that are not fully automated, and the bot has to escalate requests to human staff members to collect this data manually. To resolve this issue, the business would need to revisit its process automations to ensure that all of the systems with which the chatbot interacts are able to work as efficiently as the chatbot itself.
AI asset rationalization is different from traditional IT rationalization
The practice of rationalizing AI assets is part of a broader discipline known as IT rationalization — which refers to making strategic decisions about how an organization uses its IT assets in general.
That said, even at a business with a strong culture of IT rationalization, AI asset rationalization can be easy to overlook due to factors such as:
-
The novelty of AI investments. Unlike other types of IT assets (like servers and applications), LLMs, AI agents and other AI-based solutions have appeared within enterprise IT estates within just the past few years. Processes for rationalizing them do not yet exist within all businesses.
-
Unique AI cost-management challenges. Challenges like the difficulty of predicting AI model costs make it tougher to assess AI-related spending than it is to evaluate spending on most other types of IT services.
-
Evolving AI use cases. Even at organizations that have moved from the AI experimentation stage to production deployment of AI resources, use cases and user engagement surrounding these solutions continue to change. Change complicates rationalization because it means ROI assessments may not remain constant.
-
Changing AI prices. The price of AI products and services may also change over time — particularly as AI vendors raise prices in a bid to increase profitability (or reach profitability in the first place). Increasing prices could mean that an AI asset that a company deems rational at one point is no longer a smart investment.
What to consider during AI asset rationalization
Given that AI assets are a relatively new type of resource for businesses to rationalize, playbooks surrounding AI rationalization best practices are still evolving.
Nonetheless, by adapting the fundamentals of IT rationalization to meet the unique challenges of AI asset rationalization, businesses can implement effective AI asset rationalization practices starting today.
Considerations and priorities for rationalizing AI include:
-
Total cost of ownership. Cost calculations should factor in subscription fees, token costs, staff time spent deploying and maintaining AI products and any other expense related to AI solutions. Products that cost more to own and use need to create more value to justify their expense.
-
User engagement. In general, having more employees and/or customers using an AI product or service suggests that it may be creating more value and is an important asset for the organization — although this is not necessarily the case. It’s also possible that users frequently access an AI product simply because management tells them to, for instance, or because they’re using it for non-work-related tasks.
-
Engagement duration and frequency. The amount of time users spend with AI products provides additional context that can help determine whether a product generates real value. Frequent access events coupled with short engagement periods may be a sign that users want to experiment with a new AI solution but are struggling to obtain real value from it — so they abandon their sessions frequently.
-
Duration of deployment. Evaluating how long an AI product has been available to stakeholders is another important contextual data point. Utilization statistics related to newer products can be misleading, either because the products are not yet well known among users (which is an indicator that the organization should invest in AI asset awareness and education) or because users are flocking to test a new tool, but few will stick with it for the long term.
-
Integration status. Integration status refers to which systems an AI tool connects to or integrates with. Generally, more integrations are a sign of higher value because they imply that an asset has become an intrinsic part of solution stacks and processes. However, just because an integration exists doesn’t necessarily mean users are leveraging the integration routinely, so it’s important to examine actual workflows to determine how AI assets fit into them.
-
Security and compliance status. AI assets that pose security and compliance risks are less likely to create value for organizations — although it’s possible the risks can be managed effectively given greater investment in AI governance and compliance.
-
Vendor dependencies and lock-in. AI investments typically create more value when they do not lock organizations into a particular vendor ecosystem. To that end, the AI rationalization process should consider the extent to which an AI product or service requires the business to use other products and services from the same vendor.
-
Future-proofing. Given that AI solutions are evolving rapidly, it’s important to assess how well an AI investment can keep pace with technological change. For instance, does the vendor have a track record of updating the product with new capabilities or taking advantage of more powerful models? Or is the solution likely to become outdated in a year or two?
How and when to rationalize AI assets
Given the rapid pace of AI adoption across enterprise environments, now is the time for CIOs and other business leaders to implement an AI rationalization strategy, if they have not already. To do so, they must determine:
-
Who participates in AI rationalization. Ideally, stakeholders should include AI experts, who are qualified to understand how AI technology works and what it is capable of doing, as well as representatives of the business functions that use AI products. The latter can provide perspective on how employees and customers are actually engaging with AI. Stakeholders who bring financial expertise, too, can be valuable to help assess the ROI of AI investments.
-
When to schedule AI asset rationalization. Ideally, the rationalization process for AI should occur frequently, especially for businesses still evaluating and testing AI tools and services. IT rationalization typically occurs no more than once per quarter, but it may be beneficial to rationalize AI assets more frequently — such as once per month — so the organization can identify and mitigate suboptimal AI investments and workflows before they become entrenched.
-
Which factors matter most. We laid out key considerations for AI asset rationalization above, but the factors that your organization chooses to assess should reflect its overall AI strategy. For example, if you’re already committed to a specific AI vendor, assessing lock-in risks for AI assets may be less important than for an organization that is still weighing its options when it comes to which AI vendor ecosystem it will operate within.
No matter how organizations choose to approach AI asset rationalization, what matters most right now is simply having a plan for assessing and optimizing AI investments.
It’s much easier to correct oversights and minimize the disruption stemming from product abandonment when solutions are still new — as AI products are for many enterprises today than it is to wait until the business has already become wed to suboptimal AI products and workflows.
