In Part 1 of our Cracking AI and Outsourcing Conundrums series, we discussed at a high level the challenges of requiring outsourcing providers to drive generative AI (GenAI) innovation while at the same time complying with companies’ AI policies. One of the challenges we identified was that many outsourcing agreements impose aggressive savings commitments, to be realized through the implementation of technology solutions that enable headcount or other cost reductions.
When implementing and managing GenAI technology solutions, most companies are in the early adoption stage and requiring providers to have robust quality programs in place to verify and monitor the quality and accuracy of the training models and the output of the Gen AI tools. Some providers are pushing back on the need to implement and maintain these programs on the grounds that, if they need to retain or add headcount to monitor quality and output, the productivity benefits of such tools diminish. Companies on the other hand are generally not comfortable lifting quality reviews as they navigate the application of increased regulation and potential data governance, security, and financial risks—noting that providers should not be offering ambitious solutions that are not fully compliant with regulations as well as internal and customer and user requirements.
In this Part 2 of our Cracking AI and Outsourcing Conundrums series, we dive into balancing the company’s need for enhanced quality checks with the desire (by the company and the outsourcing provider) to realize savings.
Gating Questions
Not all GenAI solutions are the same or at the same stage of maturity. Examples range from enhanced security monitoring (to supplement other solutions and resources) to replacement of manual financial control checks (such as duplicate pay). The appropriateness of the applicable quality program will differ based on the solution.
Some key questions to consider include:
- What functions and processes are at issue
- What data is at issue and how critical it is
- Whether any regulations are implicated
- If any internal or customer policies are affected
- Whether low quality data will impact the current or future training models, data lakes, and/or output
- The impact to the company if the output is incorrect
- The party responsible for low quality or inaccurate output
- The reliability of the tools and how much testing has been done of their accuracy
- Results/benchmarks that would allow for reduced quality checks
Realizing Benefits Beyond Headcount Reduction
Just as not all GenAI solutions are alike, the potential benefits of GenAI solutions differ as well. For example, a GenAI security solution may improve identification of security gaps or potential incidents, resulting in reduced data breach risk. A GenAI fraud detection solution may promise reduced financial leakage, or a customer experience solution may deliver more focused customer outreach and lead to increased sales. The benefits are less targeted on reducing headcount than on providing other meaningful business impact.
Considerations include:
- The promised benefits of the AI solution
- Benchmarks for the committed benefits, and whether actual benefits can be documented, tracked, and quantified
- What quality controls should be implemented
- Whether additional headcount or different skillsets will be required or any headcount can be reduced
- Whether additions or reductions in headcount are included in the business case
Conclusion
The purpose of this series is to identify the potential issues arising from the use of AI in outsourcing arrangements. As the use of AI, and the applicable regulations, evolve, it will be important to adapt the contractual terms to your particular situation.
Look out for more blogs in our Cracking AI and Outsourcing Conundrums series, where we consider the applications and implications of AI in outsourcing arrangements.