Follow ZDNET: Add us as a preferred source on Google.
ZDNET’s key takeaways
- Only 5% of AI projects deliver. It comes down to the ability to customize.
- With partnerships in place, AI success odds double.
- Ask the right questions before deciding between building or buying.
There’s a tremendous gap between AI aspirations and actual successful projects — this was shown in the recent MIT study that found only 5 percent of generative AI projects have delivered measurable value to businesses. What is that top 5 percent doing differently? Their common denominator is that their technology teams are mastering the art and science of highly customizing AI to their businesses, while fostering partnerships versus go-it-alone approaches. They are going deep — very deep.
What successful AI projects do differently
Successful AI efforts “focus on narrow but high-value use cases, integrate deeply into workflows, and scale through continuous learning rather than broad feature sets,” according to the study’s research team, consisting of Aditya Challapally, Chris Pease, Ramesh Raskar, and Pradyumna Chari. “Domain fluency and workflow integration matter more than flashy UX.”
Also: 95% of business applications of AI have failed. Here’s why
Ultimately, it’s not about building or buying AI just to have AI — it’s about how the business can benefit from AI. Instead of “struggling with outdated SaaS playbooks,” professionals need to “capture enterprise attention through aggressive customization and alignment with real business pain points,” they added. “The standout performers are not those building general-purpose tools, but those embedding themselves inside workflows, adapting to context, and scaling from narrow but high-value footholds.”
It’s notable that “plug-and-play AI is a myth,” Paul McDonagh-Smith, senior lecturer of IT at MIT Sloan Executive Education, told ZDNET. (McDonagh-Smith was not directly involved with the study.) “Outside tools save time, but the real work is going to be with ‘plug-and-personalize AI’ where we customize AI tools to fit our workflows.”
The reason genAI tools such as ChatGPT succeed in pilots “is because of their flexibility,” he added. “However, they often fail in mission-critical work due to factors including a lack of memory, which reduces their ability to learn, adapt, and be customizable to the degree required to integrate effectively with our day-to-day workflows.”
Also: Stop using AI for these 9 work tasks – here’s why
Establishing strategic partnerships to move forward with AI makes a significant difference, the MIT study showed. The co-authors observed far more build than buy initiatives, with partnerships succeeding twice as often as internal development efforts. Such partnerships often provided “faster time to value, lower total cost, and better alignment with operational workflows. Companies avoided the overhead of building from scratch, while still achieving tailored solutions.”
When to build vs. buy
Still, AI proponents and developers need to weigh when it’s best to develop in-house versus working with partners such as vendors or network partners. When it comes to making such decisions, “the tipping point comes when speed, scale, or specialized expertise is called for, and your in-house teams aren’t ready to deliver to the timelines required by the business,” McDonagh-Smith said. “Building internally makes sense when the project is core to competitive edge, but we need to be careful, pride can come before a fall.”
While there is concern that using outside solutions will reduce opportunities for the needed customization, this fear is unfounded, McDonagh-Smith believes. “It’s not so much a case of ‘plug and play AI’ as it is ‘plug and personalize AI’ to fit existing and emergent workflows. I would argue that AI success isn’t dependent on the sourcing selection of our external AI tools but our internal ability to make them fit how our companies think, work, and act.”
Also: These CFOs are devoting 25% of their AI budgets to agentic AI
Other industry leaders agree that successful AI depends on individual circumstances. “Deciding on using in-house teams or outsourcing to other vendors depends on what the organization wants the AI to do,” said David Friend, CEO and co-founder of Wasabi Technologies. “If any technology, AI or other, is part of the company’s core differentiation or will allow the business to compete on price, it should be developed and managed in-house. If it is not a core part of the company’s offering, outsource it. What makes you different is what you should do yourself. What isn’t core to what you do, outsource it.”
This process needs to start with asking the right questions as well. “The question is not whether they can build the technology, but whether they should,” Adrian Murray, founder and CEO of Fisent Technologies, pointed out. “You may have very capable and well-funded technology teams, but your capacity is inherently limited and needs to be focused on the highest-value efforts. Teams need to focus their efforts where they can create differentiated value. Apply these technologies to specific business problems, not building core technology infrastructure that can be easily licensed from a solution vendor.”
The nature of the partnership relationship is also a deciding factor in AI success — it needs to be more than a transactional arrangement. “Top buyers treated AI startups less like software vendors and more like business service providers, holding them to benchmarks closer to those used for consulting firms or business process optimization providers,” according to the MIT report. This includes deep customization aligned to internal processes and data that are tied to outcomes.
Also: 4 ways to turn AI into your business advantage
“Ensure that existing and emergent workflows are decomposed, analyzed for patterns that will work — or not — with genAI, and then evolved, and recombined when genAI is ready,” McDonagh-Smith advised.
Grassroots AI adoption inside companies
Often, successful AI efforts start at the grassroots level, the study also showed. “Many of the strongest enterprise deployments began with power users, employees who had already experimented with tools like ChatGPT or Claude for personal productivity,” the study’s co-authors reported. They “intuitively understood genAI’s capabilities and limits, and became early champions of internally sanctioned solutions. Rather than relying on a centralized AI function to identify use cases, successful organizations allowed budget holders and domain managers to surface problems, vet tools, and lead rollouts.”
Also: How AI-enabled autonomous business will change the way you work forever
Agentic AI architectures are also emerging, supported by frameworks such as Model Context Protocol (MCP), Agent-to-Agent (A2A), and NANDA, which enable agent interoperability and coordination, the study found. “These frameworks form the foundation of the emerging Agentic Web, a mesh of interoperable agents and protocols that replaces monolithic applications with dynamic coordination layers.”
The culture shift required for AI
Working with AI vendors “can get you started and build important initial momentum, but the heavy lifting will be in threading AI solutions into your processes, policies, practices, and of course, your people and culture,” said McDonagh-Smith.