It’s Time for GBS Leaders to Act Like Venture Capitalists
Most AI Pilots Fail to Deliver ROI
CEOs and GBS leaders are looking to artificial intelligence (AI) as the next big thing to drive benefits for their organizations. However, the economic expectations differ from past digital transformations, such as robotics process automation (RPA). CEOs are looking for either revenue growth or hard cost savings, more so than improvements that “free up employees’ time to do more productive things.” For example, a July 2025 Wall Street Journal article was titled CEOs Start Saying the Quiet Part Outloud: AI Will Wipe Out Jobs.
Current reality poses a challenge for GBS leaders in meeting the AI expectations of their CEOs. A recent MIT study found that approximately 95% of generative AI pilots stall, delivering little to no measurable impact on P&L. While there are other classes of AI beyond generative AI (e.g., task agents, service agents, process agents), there is little evidence to suggest that pilot success rates are dramatically different in terms of driving tangible economic value.
How a Venture Capitalist Mindset Maximizes AI Impact
So, what is a GBS leader to do? Ignoring the push for AI isn’t a likely option, nor is changing CEO expectations. Hoping that AI success rates increase dramatically as one scales beyond pilots is just that, a hope. We suggest instead that GBS leaders take a proactive approach by managing their AI program like a venture capitalist (VC).
Think of the dynamics of VCs. VCs invest in start-up companies. Approximately 70-75% of venture-backed start-ups fail to return capital to investors, with some studies indicating that 30-40% may not return capital at all. But that doesn’t keep VCs from continuing to invest. Instead, VCs maintain a portfolio of companies with the expectation that a small percentage will hit it big, outweighing the failures of the others. GBS leaders need to manage AI investments similarly.
The key success factor for VCs is the careful selection of portfolio companies, followed by targeted support as these companies scale. The AI analogy to portfolio companies for VC is use cases. GBS leaders need to carefully select and dynamically manage a portfolio of AI use cases that have the potential for tangible economic value contributions. To manage the uncertainty associated with AI, the expectation is that a few use cases will outperform others and, therefore, on aggregate, meet the expectations of executive leadership.
Criteria for Selecting High-ROI AI Use Cases
Below are some practical screening criteria for selecting AI use cases.
- Potential economic impact. Avoid the risk of “spreading the peanut butter too thin” by selecting too many use cases, with many lacking any real potential for tangible economic impact. Although advantages such as increased efficiency or enhanced customer experience are appealing, they will not meet CEO expectations unless they deliver substantial and demonstrable economic value. One GBS leader described this scenario as “looking through the wrong end of the telescope.” His organization had launched a plethora of narrow use case projects focused on incremental improvements (e.g., looking from the large to the small end of the telescope). We worked with him to create a more targeted portfolio of use cases, examining end-to-end processes broadly and identifying each with potential for tangible cost savings, revenue generation, or other cash flow improvements.
2. Distribution of work. If cost savings are a primary objective, it is essential to consider not only what percentage of overall process work can be replaced by AI, but also how that work is distributed among employees. AI output is not perfect, requiring machine-human interaction to yield an acceptable accuracy rate. Look for use cases where one set of employees gathers information to be reviewed and potentially modified by another set of employees to complete a process. AI can then be deployed to replace the first set of employees, generating direct cost savings rather than merely augmenting the work of several employees, and thus making it easier to replace a full role.
3. Cost to achieve. Deploying and maintaining AI can be expensive. A recent article, Current AI Systems Cannot Replace Humans Easily: Their Total Cost of Ownership (TCO) is Enormous cites that it costs three to four times more to maintain an AI system than traditional software. Add project deployment, data management, and infrastructure costs, as well as oversight costs, to obtain a comprehensive cost view. For each use case, complete a thorough return on investment analysis.
4. Data availability and integrity. AI relies on data to learn. Generating the benefits anticipated by executives may involve applying AI throughout an entire process. However, the wider the process scope you attack, the greater the likelihood that multiple systems with multiple data sets will be deployed. Examine both the availability of data and its accuracy, as well as the level of data standardization required across systems.
5. Quality, regulatory, and ethical guidelines. There have been noted instances where AI has produced outputs that do not adhere to regulatory and ethical guidelines. At a more basic level, the current limit of AI output accuracy is approximately 92 percent, according to the article cited above, with many instances significantly less. Compare this to service level targets for most GBS processes, which range between 95 and 99 percent. Some issues with AI output quality can be addressed through human review of the output. However, consider carefully the risks and mitigation efforts associated with quality, regulatory, or ethical breaches when deploying AI for each use case.
6. Stakeholder acceptance and engagement. Just like a VC needs a company looking for investment, AI needs key process stakeholders (e.g., customers, employees, suppliers) to buy into the program. Look particularly at user trust and acceptance in selecting use cases.
How GBS Leaders Can Support AI Implementation at Scale
Once a use case is added to the AI portfolio, like successful VCs, GBS leaders need to support those use cases as they scale and dynamically manage the portfolio. Many GBS organizations have established their own AI Center of Excellence (COE) as a primary source of ongoing support. Others will rely on another AI COE in the enterprise. From whatever the source, key elements of ongoing support include:
- Program and project management
- External partner and provider relationship management
- Policies, methods, tools, and standards management
- Systems, applications, and infrastructure development and maintenance management
- Change management
- Financial and performance/benefits management
- Resource and talent management
- Risk, security, and compliance management
AI presents a unique opportunity for companies and GBS leaders to rethink their operating models to solve real problems. But expectations and risks are high. GBS leaders can learn from the portfolio management acumen of VCs to effectively manage benefits, expectations, and risks. To gain more excellent insights from our SSO Network, please join us for our upcoming Process Mining and Intelligence Virtual Summit.