The value of cheaper AI depends on how effectively a company turns it into better decisions and more productive work.
The declining cost of AI
For businesses, AI is becoming more affordable. This is encouraging. But measurement is also important. Businesses can increasingly access comparable AI capabilities at lower cost, although savings vary by model and use case. Three developments help explain this improving affordability of AI:
- Lower token prices. Tokens are the units of information, such as words or parts of words, processed and generated by AI models. Providers commonly charge for direct model access according to the number of tokens used. Some newer models offer lower prices: Anthropic, for example, introduced Claude Opus 5.5 with standard input and output token prices 20% below those of Opus 5.
- More computing performance per dollar. Research by Epoch AI estimates an average annual improvement of 49% in computing performance per dollar spent on AI chips between 2023 and the end of 2025. This helps reduce the hardware cost of providing a given amount of computing capacity.
- More efficient processing. Companies can reduce costs by reusing previously processed information, a technique called caching, and submitting suitable work in batches when immediate answers are unnecessary. Anthropic offers these options, including a 50% discount for batch processing.
These savings do not mean every cost of using AI is falling. Data, integration and human review remain part of the total cost. The business opportunity depends on whether cheaper processing lowers the cost of reaching a useful, trustworthy answer.
Turning lower costs into business value
As AI becomes more affordable, business leaders should ask: What can we now do profitably that was previously too expensive or time-consuming?
Consider a consumer packaged goods (CPG) firm investigating why shoppers have stopped buying a particular brand. The investigation could require combining purchase data, consumer research and promotional results. AI could help analysts query these sources, summarize consumer responses and identify patterns for further investigation. If these tools reduce the effort required, teams could investigate more frequently and examine smaller shopper segments. Lower processing costs could make previously unexplored questions economical to investigate.
Lower costs make more analysis affordable
Lower AI costs can reduce the expense of existing analysis and make additional investigations worthwhile. For example, an analysis previously reserved for a major brand or an annual planning cycle may become worthwhile for a smaller product line, a regional market or a specific shopper segment. With cheaper AI, teams can test more explanations for a change in sales and refresh their findings more frequently. Total spending could still rise if growth in the number of tasks more than offsets savings on each task.
Why cheaper analysis does not guarantee better decisions
However, cheaper AI reduces only part of the cost of answering a business question. Companies still need reliable data, connected systems, people who can check findings and the ability to act. Human judgment remains essential: a convincing explanation may fit the available data without establishing what caused the change.
For example, a promotion by a competitor might coincide with falling repeat purchases, but a product that is out of stock could also explain shoppers leaving. Analysts must question assumptions, consider competing explanations and assess the reliability of findings. Business leaders must then decide whether the evidence supports action or further investigation.
The importance of trusted consumer intelligence
The usefulness of AI therefore depends on the quality and relevance of the available evidence. Teams need to know whether the data represents the relevant shoppers, how recently it was collected and where gaps in coverage could affect conclusions. Purchase data can reveal which shoppers are changing behavior, what they are buying and where they are spending.
Consumer research adds context by exploring whether shoppers report seeking better value or feeling dissatisfied with a product. Comparing these reported motivations with actual purchases helps teams test explanations, while recognizing the limits of each source.
Turning cheaper AI into competitive advantage
As AI becomes more accessible, a firm’s competitive advantage may increasingly depend on proprietary knowledge, trusted consumer evidence and people who can turn findings into action. A practical way for a business to assess this opportunity is to select one recurring decision and measure whether using AI improves the outcome after accounting for data, technology and review costs.
The value of cheaper AI depends on how effectively a company turns it into better decisions and more productive work.