- a better understanding of the underlying models,
- Experience integrating these models into various types of enterprise environments, as well
- Know about planned new features before they are officially announced.
On the other hand, however, there is the obvious disadvantage Vendor dependency: Even if future rollouts are not part of the contracted deliverables, an AI vendor’s FDE teams could have subtle influence on their customers’ future AI initiatives. This is also why Flavio Villanustre, CISO at LexisNexis Risk Solutions, urges IT managers to be careful: “FDEs are – also due to financial incentives from their employers – aimed at increasing the customer’s use of AI and creating ‘stickiness’ in terms of services. Even if these teams offer meaningful added value, it is recommended that users always consult unbiased experts who can unbiasedly evaluate competing solutions from different providers.”
According to Villanustre, this is particularly important at a time when the investor-subsidized business models surrounding AI tokens are starting to show cracks: “Given the current rapid pace of innovation, being flexible enough to change providers could create significant competitive advantages.”
Sanchit Vir Gogia, chief analyst at Greyhound Research, also sees the use of an FDE team by the AI provider as a double-edged sword: “FDEs are embedded in the customer’s environment, but they also represent the commercial interests of their employer. Giving the experts from AI providers excessive influence on implementation decisions could not only lead to increased dependency, but also to high prices, against which one may then not be able to defend effectively.”
In addition to potentially increasing vendor dependency, there are other aspects that IT decision-makers should consider before engaging their AI provider’s teams. The central question is how long the FDE teams will actually be needed. After all, a single implementation will hardly be enough. The resulting ones long-term costs is something that companies often overlook, according to John Sangyeob Kim, AI engineer at Solidroad: “The deployment itself accounts for perhaps 20 percent of the total cost. The remaining 80 percent goes to keeping the system running through model upgrades, as well as handling data movements and edge cases that only arise months into production.”
There are also risks that arise when the AI provider’s FDE team is no longer on site. For example, it is seriously underestimated, how comprehensive the insights arewhich external specialists gain during AI implementation, says Justin Greis, CEO of the IT consultancy Acceligence: “As part of the implementation, the FDE team receives insights into a variety of operational details. Although the data accessed is protected by NDAs and confidentiality clauses – the processes and workflows viewed are often left out.”
According to Greis, this problem is not limited to FDE teams from AI providers. Whoever supports the introduction of AI will gain significantly more knowledge than recorded in the Statement of Work, explains the manager: “This includes the actual, concrete work processes, undocumented exceptions, gaps in data quality, approval bottlenecks or security workarounds. This knowledge is as valuable as it is sensitive.”
