People
Tim Roberts
Partner & Managing Director, UK Country Co-Leader
London
MIT’s recent report on the State of AI in Business reported that 95% of GenAI deployments aren’t delivering real value, which should be a clear sign for every executive to pay closer attention to their AI investments. But is this a crisis, or simply an inevitable sign of an emerging technology finding its footing?
While we could examine some of the report’s finer details—and we acknowledge it is just one report among many—the research has clearly struck a chord with the business community. We agree with many of the conclusions and themes based on our own client work, and see the conversation as an important one for the industry. At AlixPartners, our experience across industries validates several key MIT findings. The “95% failure rate” MIT found is directionally correct with what we’ve seen, although it’s not a challenge unique to GenAI. While creating Practical AI for CEOs, we found that 80%+ of all AI projects (encompassing all ML, and therefore GenAI too) underperform expectations. It’s also true that the greatest ROI has generally come from traditional (non-GenAI) machine learning so far, but this is in part because GenAI is newer. We’ve already seen GenAI drive meaningful results in functions like marketing, customer service, and software development by unlocking efficiency and agility.
That being said, with decades of perspective and client-side pragmatism, we recommend business leaders take a disciplined, practical approach to investing in AI —rather than total disillusionment.
Practical execution: what's the most successful approach to AI investments? Where should you start?
In hundreds of client engagements, we’ve learned a tough truth: AI doesn’t fail for lack of ambition or budgets; it fails in execution. We also determined that the difference between hype and value comes from whether or not your business takes a disciplined approach in five areas:
- Start with a business objective to select initial use cases
Treat AI as a tool, not the strategy itself. Functional impact comes from integrating the right AI methods to improve and reinvent business processes that drive genuine business outcomes—not from launching isolated pilots or one-off apps. - Build a ‘risk-taking’ governance to adopt a VC approach to further ongoing investment in innovation
VC firms know that their investments are not risk-free – there are implications to making too many bets, too few bets, the wrong bets, indecisive bets … so establishing a sound decision-making process, driven by business opportunities, not tech ideas, is essential. - Build the right foundations
Iffy data quality, antiquated systems, and indecisive governance remain the biggest barriers to success. “Garbage in, garbage out” has never been so true. GenAI is only as useful as the structure, readiness, and trustworthiness of a company’s information infrastructure.
This includes investing in your people. Upskilling younger employees is especially important, as a recent Stanford paper shows that AI’s impact on employment is felt more strongly by entry-level employees whose work is most frequently replaced by AI. - Focus on scaling adoption effectively
Pilots and demos are easy; moving to scaled, workflow-embedded use is hard. The most successful companies don’t just roll out AI. They build buy-in through high-visibility wins, practical training, and leadership engagement. - Organize for success: Centralized leadership, decentralized action
There is no one-size-fits-all operating model, but we typically see winning strategies pair a small central task force—close to the C-suite—with empowered, cross-functional teams. Our recommended approach to organizational design accelerates knowledge-sharing and adoption, while putting the power to adapt AI to frontline problems directly in the user’s hands.
How should we equip our business to make the most of AI-driven efficiency?
The “shadow AI economy” called out by MIT is real: if companies don’t help their staff use AI safely and effectively, employees will do it anyway on their own accounts. Proving the ROI on everyday efficiency is difficult, but the trend for employees to use GenAI to make their projects smoother and faster is an inevitable development. People are using GenAI in their day-to-day lives, be it while shopping or doing research. It is natural for them to expect the same conversational and agile options to interact with the digital world at work, too. Leaders must strike a balance between enabling experimentation, upskilling their employees, and protecting data/IP. That means:
- Creating an environment in which people have access to coaching and training on when and how to use AI most effectively, with access to peer learning and open feedback to accelerate upskilling
- Establishing a culture of innovation by encouraging people to use AI to do things they already do better or faster than they previously did. This is far less risky than having employees try completely new things they can’t sense check.
- Proactively providing secure, sanctioned access to the best tools and training everyone to spot and avoid risky behaviors.
How aggressive should you be when it comes to investing in AI? When does it make sense to partner with others?
A practical buy-versus-build approach, not rigid adherence or a zealous approach to either, remains crucial if organizations want to quickly realize AI benefits and position themselves for longer-term transformation. It is important to find a balance between being overly cautious, given the costs involved with AI investments, or conversely, allowing a fear of being left behind to drive overly optimistic investments. Instead of chasing novelty, we’ve found that businesses get the best ROI by doubling down on what works and evolving as technology matures.
Exploit mature technologies for quick wins. This depends on your industry of course, but examples include ML-driven solutions for pricing, supply chain forecasting, and risk management, which have already proven their value at scale in retail, financial services, and manufacturing.
Deploy GenAI in its natural sweet spots. Content drafting, conversational chatbots, and code generation show notable impact in customer service, marketing, and software development—functions where GenAI can accelerate work without disrupting core operations.
Invest in the right foundations. Data curation, integration, and governance are all essential for effective use of AI today and in the future.
Take advantage of practical partnerships where they appear. MIT’s data show externally-built solutions deliver twice the success rate of internal DIY efforts, especially when tailored for workflow fit and continuous learning.
How will AI’s role in business evolve over the next 18 Months?
Change is accelerating, but so are expectations. Leaders must move quickly from experimentation to embedded, accountable business impact. The next wave of AI will be defined by agentic systems—AI that learns, remembers, and acts across complex workflows—not just in standalone apps. Generative and agentic AI are already transforming the way consumers interact with the digital world, and brands will need to incorporate GEO (Generative Engine Optimization) into their marketing to remain relevant.
Each industry will feel these shifts at different speeds, with some, like retail, already finding new revenue streams. However, every business should prepare for fundamental changes in how customers engage, how teams collaborate, and in the very nature of work itself by making sure it has the right strategy and information infrastructure in place to succeed with AI.
Bottom line: It’s time for discipline, not disillusionment
MIT’s report is an excellent reminder of an unexciting but essential truth in business—the path to value is usually found through focus, readiness, and relentless execution.
The future belongs to those who embrace this truth by investing in the boring but crucial fundamentals: building the right foundations, aligning technology to strategy, and leveraging AI to enhance—not substitute—their people's ambition and impact.
