Only 5% of companies report seeing a return on their generative AI investments, according to researchers with MIT's Project NANDA, despite enterprise spending on the technology totaling between $30 billion and $40 billion. The finding underscores a widening gap between the ambitious promises that have driven corporate AI adoption over the past three years and the tangible outcomes most organizations have actually achieved.
Many companies that invested in AI solutions have seen improvements limited to specific workflows or individual departments, rather than the strategic advantages across the entire organization that executives anticipated. Expectations for higher margins, reduced labor costs, and increased shareholder value led businesses to ride the wave and invest in a wide range of AI tools, from simple automation for research and data management to platforms that aggregate data in preparation for AI applications. In some cases, those expectations have been met, but only on a narrow or departmental scale, or as part of testing. In others, they have fallen short.
The gap between hype and tangible results is now forcing a shift in how AI is bought and sold. During early testing phases, it was common for technology and department teams to evaluate and test AI solutions independently. That step remains important for understanding where tools may fit in the larger technology stack. But as company objectives for AI adoption become clearer, C-suite leaders are becoming more involved, and the nature of the AI buying process is changing. Greater executive involvement in decision-making is now required for future AI investments, with a new evaluation and assessment process emerging to incorporate quantitative KPIs based on an enterprise view of how AI should support the broader organization.
A central problem is that many AI tools developed over the last few years have not been vertically aligned, meaning they can be applied across many industries. Through testing, buyers have determined that generic tools often lack the industry-level context, accuracy, and usability needed to meet specific enterprise goals. During initial scoping meetings, suppliers should be asked to demonstrate industry alignment through integrations with vertical platforms, custom AI tools or agents, compliance with industry standards or regulations, and customizable reporting on industry-specific KPIs.
In retail and consumer packaged goods, for example, brands continue to face supply chain challenges exacerbated by tariffs, geopolitical conflicts, and climate change. Purpose-built AI for retail is designed to deliver accurate, connected industry data to overcome those challenges, promoting agility to quickly adjust orders, assortments, promotions, and display placements when needed to prevent out-of-stock items, reduce waste, and improve collaboration with retail partners. One large CPG brand, for instance, detected an early risk in its distribution network that would have impacted 6,200 cases across five SKUs. The combination of deep, industry-specific data and strategically selected AI tools used to leverage large retail data sets helped the company avoid a costly error.
Once industry expertise is confirmed, customization becomes the next critical area to explore. Off-the-shelf AI tools may work as point solutions, but achieving strategic advantages across an entire organization requires customization, especially for AI agents, connectivity, and reporting. Buyers should begin the technology discussion by clearly outlining business goals and KPIs, existing data and platforms, and where current systems fall short or where knowledge gaps exist. This sets the foundation for expectations of AI deliverables. A good supplier will understand that today's sales engagement requires a «build as you buy» approach, where demonstrating customization during the sales meeting may be a baseline requirement.
A goal of 100% accuracy is also becoming a new standard for AI solutions. AI-generated data and analysis should precisely align with buyer KPIs, and buyers should ask for evidence of success. If a CPG brand knows a product is being phased out, for example, it can build an AI agent to automatically pause manufacturing, logistics, and marketing for the product and develop a phase-out and discount plan to move remaining product off the shelf. That scenario requires complete accuracy to ensure each step of the process is followed, data is delivered for reporting, and humans can step in to adjust along the way.
Rather than requiring internal team members to search for data, develop reports to solve a problem, and manage workflows, cross-functional AI agents can quickly analyze data across departments, recommend a solution, and with human guidance, automatically take action. Importantly, AI agents can overcome silos common in large organizations and work efficiently across departments toward specific outcomes, helping teams overcome the inefficiencies of organizations structured by functions and limited by insights.
As C-suite leaders take a more active role in AI buying decisions, the emphasis is shifting to asking the right questions, requesting examples of customization, requiring proof of data accuracy, and illustrating how AI can reduce workloads while addressing specific industry, company, and team requirements based on enterprise KPIs. By applying a smarter evaluation process to AI solutions, leaders can unlock strategic value from AI investments across the organization while delivering better outcomes for end users.