Telecom & Technology Consulting
When developing a technology budget, CIOs initially place items into one of two buckets: Need to Have and Nice to Have. From there it’s a matter of choosing what best aligns with the organization’s strategic goals and business priorities.
The same process, at high level, applies when creating an AI-specific budget. But since we’re all still getting a grip on AI’s potential, the allocation of dollars must be undertaken with more prudence than a traditional tech budget.
To begin, be sure that your AI strategy connects to measurable business results. This is stating the obvious, of course. The difference here is that AI implementation doesn’t yet deliver consistently predictable metrics. Accordingly, try using existing metrics to assess success, and adjust accordingly over time. While not ideal, you do require hard targets before moving into the full-scale deployment phase.
As you’re considering the budgetary possibilities, do so in the context of AI governance, which requires you to consider a host of issues like setting data-access rules and accuracy guidelines. This is critical: While putting AI governance on the back burner is often viewed as a compliance risk, it’s actually a budgetary challenge. Waiting to wrap everything around AI governance can actually create funding gaps and, ultimately, stall progress.
Standards are fundamental when developing any category of technology. Unfortunately, some have overlooked that time-honored rule regarding AI-generated code. That can be a costly mistake when crafting your AI-specific budget.
According to a recent survey, over 70% of developers use AI coding assistants, yet they prioritize quality (75%) over costs (38%). As such, CIOs are wise to craft rules for choosing between cost and quality. Also keep in mind that over 90% of developers are skeptical of AI-generated code, with nearly 50% consistently reviewing this type of code before deployment. Extra, unforeseen budgetary needs can creep in without code quality baked in from the start.
Also spend time considering which AI model(s) needs a line item in the budget. In many cases, developers select models due to personal preferences instead of objective performance or cost factors. Known as “frontier models”, these are typically high-priced and aren’t required for every AI-developmental task. Instead, allocate money for an assortment of models with a less-expensive variety including small language, custom, and industry-specific.
AI agents can bring additional, unanticipated costs. That’s because those agents may not be monitored, causing price increases as well as streaks of inefficiency. Remember that AI agents typically have variable costs — AI providers either offer outcome-based pricing models while others employ a license model with a fixed number of tokens. As a result, tech leaders need to monitor costs, develop AI observability standards, and set rules for managing differences in performance.
Budget can be affected if you develop agents without a solid change management plan. The reason: deploying AI agents is merely a single step in the multi-pronged change management process. This is why the largest overspending on AI rarely occurs due to the actual technology. Rather, it stems from deploying AI without proper oversight including that for redesign and AI governance.
To avoid overruns with cost, you must identify who (and what department) is responsible to review AI agent suggestions. By doing so, you promote confidence in the switch to autonomous decision-making for high-volume recommendations.
In the age of AI, it’s easy to rely on that very technology to develop any plan, be it for a product and even a budget. The price, however, can be too devastating when we assume the prompt will always deliver a clean, strategically aligned result that accounts for all contingencies. You simply can’t afford that. Before submitting the AI budget, prompt everything using your human mind.
