Scaling AI for finance a massive challenge for enterprises: Study

Gartner predicts half of finance AI projects will be delayed or cancelled by 2024

Half of current finance artificial intelligence (AI) deployments will be either delayed or cancelled by 2024, while the use of business process outsourcing (BPO) for AI will rise from 6% to 40% within two years, according to Gartner. The finance department of enterprises will face major barriers to scaling up the use of AI in-house and will increasingly turn to business process outsourcing (BPO) solutions to meet their digital transformation objectives.

Half of current finance artificial intelligence (AI) deployments will be either delayed or cancelled by 2024, while the use of business process outsourcing (BPO) for AI will rise from 6% to 40% within two years

“While finance departments have made reasonable progress in laying the groundwork for AI, the challenges come when attempting to scale up solutions that can manage the complexities of function-wide use,” said Sanjay Champaneri, senior director analyst in the Gartner Finance practice.

“The upfront costs of building scalable infrastructure in house, and the overreliance on stretched citizen developers, will lead many CFOs to rethink their current strategies,” he said.

Digital automation in finance often fails to meet the expected benefits outlined in business cases for deploying such technologies. Much of this is down to a lack of truly functional automated processes, according to Champaneri, who notes that a significant proportion of automation work fails and is rerouted to a human for manual input. Without correcting for this state of “fake automation,” finance departments will struggle to scale automated solutions, such as AI, effectively across the function.

Barriers to Scaling AI in Finance

As the number of AI solutions and users grow, so does the complexity in scaling efforts. CFOs who attempt to keep AI in-house will hit a productivity ceiling, as the complexity of maintaining projects tax internal resources and slow or prevent the deployment of new solutions.

According to Gartner three key barriers that finance departments will face when attempting to scale up their AI processes across the functions.

  • Costly upfront infrastructure – Building infrastructure in-house requires upfront investments for cloud hosting, acquiring new specialist skills for infrastructure maintenance and additional security investments required to manage an ever-growing user base.
  • Lack of bandwidth among citizen developers – AI models require continual monitoring and frequent retraining and configuration updates. These requirements divert citizen developers from their core tasks and stretch internal bandwidth.
  • Skill-gaps among citizen developers – The citizen developer role is not designed for the technical complexities required to synchronize IT systems and services, nor do they have the skill sets required in workflow management to adapt to frequent changing parameters.

“CFOs need help operationalizing AI, and also ensuring that their limited resources are focused on projects generating the highest return of efficiency,” said Champaneri.

“This reality will drive a significant growth in the use of BPO providers for AI, which offer ready-made solutions to overcome these barriers.”

Advantages of AI-Enabled BPO

Finance departments have thus far been slow to adopt BPO providers for AI, with just 6% currently utilizing an AI-enabled BPO. Gartner predicts this usage will rise to 40% by 2024, as the benefits of deploying a market ready solution for AI become more apparent to CFOs.

“CFOs are increasingly learning from past experiences of ‘going it alone’ with technologies such as robotic process automation (RPA), which have failed to achieve the expected ROI,” said Champaneri.

He adds that the case for offloading AI to an experienced BPO provider is even stronger, as it frees up internal resources while providing access to economies of scale that improve the accuracy of the technology and account for the complexities of scaling it up across the function.


Add new comment