MIT’s The Gen AI Divide study reveals that 95% of generative AI projects create little business impact or ROI. In this post, we’ll break down the critical mistakes that limit the success of AI projects and what makes the great projects deliver outstanding ROI.
We wanted to provide some thoughts and context regarding the recently released MIT Study, entitled The Gen AI Divide: The State of AI in Business 2025. The study has attracted a great deal of attention as more and more companies try to leverage AI for business transformation.
RapidCanvas partners with hundreds of companies to help them define and deploy successful AI applications that drive material ROI. In the past 12 months, we have met with 340 business leaders and key employees, who are actively piloting or seeking AI solutions. These conversations have significantly informed my perspective on the challenges and opportunities facing organizations today around AI.
The MIT report reveals a stark reality with which many executives are all too familiar: despite $30-40 billion in enterprise investment, 95% of generative AI projects are delivering little or no measurable business impact. That statistic is, of course, a great headline grabber, and confirms what we and many others have suspected for some time: Quantification is incredibly valuable.
What is surely even more valuable is understanding how to ensure your company ends up on the good side of the divide. That’s the primary focus of RapidCanvas, and we have many case studies that demonstrate how our Agentic AI + Human Experts model helps achieve outstanding results in weeks instead of months.
The Problem: A $40 Billion Wake-Up Call
MIT's Project NANDA analyzed 300 AI deployments, conducted 150 interviews with business leaders, and surveyed 350 employees to uncover the truth behind AI's promised transformation. The findings reveal that only 5% of AI pilot programs achieve rapid revenue acceleration, while the vast majority stall and deliver little to no positive ROI. Many projects become stuck at critical bottlenecks, preventing them from advancing beyond the pilot phase.
The study exposes what NANDA Initiative researchers call "The Gen AI Divide" - a growing chasm between the minority of organizations driving ROI from AI and the overwhelming majority struggling to make AI work for their businesses.
Often, companies aren't failing because they lack intelligence or resources. They're missing out because they make some combination of eight critical mistakes that prevent AI projects from delivering real business value. Let's examine each problem area and explore high-level strategies to overcome it.
Challenge 1: Projects Not Aligned to the Company's Immediate Goals
Many organizations pursue AI for its own sake, rather than focusing on specific company business objectives. Common misalignment patterns include exploratory fishing expeditions without clear targets, projects that are too broad in scope, and unfocused initiatives that try to solve multiple problems simultaneously.
The MIT study also found that more than half of generative AI budgets are devoted to sales and marketing tools. While marketing and sales projects can be fruitful, they are often not the most lucrative opportunities compared to tasks plagued by inefficiencies such as underutilized data, overreliance on spreadsheets/BI dashboards, and manual processes. Many companies find, for example, that back-office automations offer the greatest value for initial projects.
Solution Strategy: Start with your most pressing business pain points. Identify processes that are currently outsourced, performed manually, or that create operational bottlenecks. Map AI initiatives to measurable business outcomes like cost reduction, revenue growth, or customer satisfaction improvements.
Challenge 2: Projects Not Scalable or Secure
AI Security and scalability concerns plague many AI implementations from the start, according to “The Gen AI Divide”. Organizations often build proof-of-concepts without considering enterprise requirements for data governance, compliance, and system integration. These can be real proof-of-concepts or de facto ones where employees adopt free personal AI tools without controls or safeguards. Chances are, you already have this problem in your organization, with people using free LLMs like ChatGPT and Claude to generate business plans, goals, MBOs, emails, and other content without safeguarding company IP or secrets.
Solution Strategy: Control AI access and data exposure from the outset. Create systems and processes that protect company IP and prevent data leakage while ensuring that the benefits of AI are available to all employees and functions. Design with scale and security in mind from the outset by establishing clear data governance frameworks, implementing robust security protocols, and ensuring your AI solutions can seamlessly integrate with existing enterprise systems. Consider cloud-native architectures that can grow with your needs.
Challenge 3: Too Focused on Build Versus Buy
MIT's research reveals a critical insight: purchasing AI tools from specialized vendors succeeds about 67% of the time, while internal builds succeed only one-third as often. Yet many companies default to building proprietary solutions.
Solution Strategy: Adopt a "buy first" mentality, but understand that simply purchasing a SaaS one-size-fits-all solution is a sure path to problems. The most effective approach combines Agentic AI with human experts, starting with proven solutions that are then customized to your specific processes, infrastructure, and data sources.
Build versus buy becomes a legacy concept. Internal teams provide institutional knowledge and context, while experts bring advanced AI knowledge that can be customized and applied to company challenges. You buy the AI expertise to build on your team’s skills and knowledge.
Evaluate existing solutions before committing to custom development, and partner with proven AI vendors who have solved similar problems for other organizations while offering the customization needed for your unique requirements. Reserve fully custom builds only for truly unique competitive advantages.
Challenge 4: Projects Unrealistic for Existing Teams
Many AI initiatives fail because they demand skills and capabilities that current teams simply don't possess. Organizations underestimate the learning curve and change management required for successful AI adoption. Hiring data scientists and other AI experts is an exceedingly difficult, time-consuming, and costly process. Solutions and approaches that require hiring can add months to project timelines and enormous costs.
However, there is an even greater benefit to Agentic AI than avoiding hiring challenges and delays. By democratizing AI, you shift the power dynamic, giving more freedom and responsibility to your existing teams. Teams gain explicit permission to try AI and automation without fear of compromising security or being penalized for taking shortcuts. Your team’s mindset changes from being rote process executors to why-thinkers. They ask questions like:
- Why does this matter?
- Why do we do this?
- Why do we do it in this way?
Agentic AI further transforms technology into teammates with complementary skills. Territoriality declines, and synergistic relationships form. People and agents work together to drive faster growth, enhance productivity, and unlock profitability.
Solution Strategy: Conduct honest skill-gap assessments before launching AI projects. Consider hybrid development approaches that combine external expertise with internal knowledge. For organizations with no data science resources, explore outsourced solutions that bring best-of-breed data science resources to your team’s industry and company expertise. This is an extremely powerful way to deliver results without requiring internal AI capabilities. Invest in training and change management during transition periods to drive the adoption of AI.
Challenge 5: Projects Not Prioritized Correctly
Prioritization is key to avoiding the issues identified in “The Gen AI Divide.” Organizations often fall victim to ‘shiny object syndrome’ and the desire for full transformation overnight. . Without proper prioritization, teams often tackle complex, high-risk projects first instead of building momentum with more achievable wins. This approach leads to early failures that can derail entire AI strategies.
Solution Strategy: Implement a systematic prioritization framework. Start with projects that offer high impact and low complexity. Build internal confidence and expertise through early wins before tackling more ambitious initiatives. Create a pipeline that balances quick victories with longer-term strategic projects.
Challenge 6: Inappropriate Project Budgets
Budget misalignment kills AI projects in two ways. Underinvestment causes disenchantment because it makes it appear that there are serious overruns and out-of-control costs when projects require more resources. Overinvestment wastes money while setting unrealistic ROI expectations. It sours an organization when incremental AI initiatives fail to deliver proportional returns, despite massive projects. Many organizations struggle to estimate the costs and expected returns of AI projects accurately.
Solution Strategy: Start small and build momentum through progressively larger investments. Develop realistic budget models based on successful implementations with similar characteristics. Factor in all costs, including data preparation, training, ongoing maintenance, and change management. Establish clear ROI thresholds and timeline expectations before project approval.
Challenge 7: Lack of Roadmaps and Accountability
AI projects need plans, timetables, milestones, and checkpoints, just like other initiatives. Don't get enchanted with the transformational power of AI and dive in without a plan. The MIT study attributes failures not to insufficient infrastructure or talent, but to the inability of AI systems to retain data, adapt, and learn over time. Many projects lack clear roadmaps that address these fundamental requirements.
Solution Strategy: Create project roadmaps that extend beyond initial deployment. Include plans for continuous improvement, user feedback integration, and system adaptation. Establish clear ownership and accountability structures, with regular milestone reviews to ensure transparency and accountability.
Continuous improvement warrants particular focus because it’s about more than small incremental changes in baseline performance. With Agentic AI, there are three broad dimensions to continuous improvement, all of which can dramatically enhance company processes, effectiveness, and results:
- Evolving AI capabilities: AI is advancing rapidly, and companies must either work with partners or plan internally to keep “upgrading” solutions in line with these evolutions. Agentic AI enables companies to “build in” technological change.
- Ongoing transformation: The benefits of Agentic AI aren’t a one-time event. Each launch creates opportunities for teams to enhance and push further up the value curve. The flexibility of Agentic AI enables follow-on transformation of processes and outcomes.
- Eliminating Vendor Lock-In: With Agentic AI, organizations no longer need to wait for SaaS vendors to roll out new features. With the right partner or internal capacity, they can quickly transform solutions on their own and eliminate dependence on SaaS providers and implementation firms.
Challenge 8: Missing Success and Failure Measures
Without proper metrics, organizations cannot distinguish between successful and failed AI initiatives. This leads to continued investment in underperforming projects and abandonment of potentially successful ones. Clear, pre-agreed metrics build commitment from the executive team through the rest of the organization by establishing proof and accountability.
Solution Strategy: Define success metrics before project launch. Some examples might include:
- New revenue streams/figures
- Hours saved
- Errors reduced
- Increases in gross margin
- Reduced out-of-stocks
- Throughput improvements
- Rework reductions
Establish both technical performance indicators and business impact measures. Create dashboards that track progress against baseline performance and provide early warning signs for course correction.
Assessing Opportunities and Results
As we evaluate AI initiatives, we must also recognize that success can come in multiple forms and degrees. We all want outstanding ROI from everything we do, but any sensible AI roadmap involves multiple projects that deliver varying ROI levels. AI transformation is a process requiring experimentation and learning. Success will include a broad range of projects and outcomes:
- High-return projects that deliver outstanding immediate and long-term ROI while laying groundwork for future initiatives
- Foundational projects that deliver lower immediate ROI but are essential for future gains, such as data integration efforts that enable multiple subsequent projects
- Projects with lower-than-expected returns that still serve as stepping stones to transformation and provide crucial learning for future efforts
The return on projects depends on your company's unique characteristics—your business needs and infrastructure differ from your competitors. This is a key reason why companies that buy pre-fab SaaS AI solutions are often disappointed in results. Having human experts to guide the AI roadmap and project development is critical for thebest results.
Transformation requires constant experimentation and acceptance that projects may exceed or fall short of ROI projections. The 'fail fast' concept - that is, learning from results to enhance future performance - is particularly relevant in AI transformation.
Successful companies prioritize efforts and implement sequentially to maximize ROI and gain insight into where the greatest value lies. What's important is starting now rather than waiting for guaranteed successes.
The Path Forward: Learning from the 5%
The Gen AI Divide: The State of AI in Business 2025 doesn't just highlight problems - it reveals what successful organizations do differently. Companies that succeed focus on empowering line managers, rather than just relying on centralized data science teams, to drive adoption. They select tools that can integrate deeply and adapt over time.
The study shows that successful AI implementations share common characteristics:
- Strategic Focus: They solve specific business problems rather than showcasing technology capabilities
- Realistic Scope: They start with achievable goals and build scale over time
- Strong Partnerships: They leverage external expertise rather than trying to build everything internally
- Continuous Learning: They implement systems that improve through usage and feedback
- Clear Ownership: They establish accountability structures that extend beyond IT departments
Breaking Through The Gen AI Divide
Your organization doesn't have to join the 95% that fail to see returns from AI investment. By addressing these eight critical challenges with proven strategies, you can position your AI initiatives for measurable success.
The question isn't whether AI has the potential to transform business - it's whether your organization will be among the 5% that make it work.
RapidCanvas has developed an Agentic AI + Human experts hybrid model that addresses many of the challenges outlined in "The Gen AI Divide." This approach has been tested with dozens of companies across various sectors, including supply chain, manufacturing, retail, financial services, CPG, real estate, energy/utilities, and others. The model typically delivers measurable ROI within 4-12 weeks, which is significantly faster than traditional AI implementation approaches.
The hybrid model recognizes that successful AI implementation requires both advanced technology and human expertise. Rather than forcing organizations to choose between building internal capabilities or buying generic solutions, this approach combines AI capabilities with experienced professionals who understand both the technology and specific industry challenges.
Learn how a RapidCanvas AI workshop can help you prioritize projects, set goals, and receive a roadmap for your AI journey in just two days. With 2026 planning starting soon and budgets being finalized, there’s no better time to define your highest-ROI AI use cases. In just 2 days, our expert-led workshop will help you pinpoint your highest-impact AI opportunities and equip you with a custom implementation roadmap—ready to deploy immediately. Get more information now.Read the other posts in our miniseries on The Gen AI Gap:
The Gen AI Divide: How to Prioritize AI Projects
The Gen AI Divide: Essential Elements of an AI Roadmap
