Article | 13.4.2021
Is your company prepared to implement AI at scale?
As Robotic Process Automation (RPA) developer and consultant, I have witnessed the latest boom beginning roughly in 2015, where companies and organizations started adopting new business process automation tools to streamline their back-office work. Most of the early successes came from regular rule-based automations that have saved organizations countless hours of manual labor, often redirecting the menial routine tasks to software robots, focusing human efforts on more complicated and challenging work. As large bulk of these tasks were successfully automated, a new challenge emerged:
In comes AI and Intelligent Automation. As we at Sisua Digital started to develop our capabilities to solve this challenge by adopting 3rd party platforms to enhance robots’ performance, we faced new and unexpected hurdles that do not typically manifest themselves in rule-based automation projects. The exploration of these challenges forms the core of my master’s thesis, where I interviewed AI team leaders from five established Finnish companies that have managed to scale their AI operations successfully. The results are aimed to help companies to anticipate the challenges across industries, making AI more accessible.
Implementing AI should be viewed as a multi-disciplinary challenge
The motivations behind implementing AI might come from varying directions, which all are valid paths to adopt AI for the first time: company board setting new strategic goals on AI development, tech-savvy manager finding AI based tools a good fit to solve imminent business problems, or the pressure from competition that propels the industry standards. The decision to invest in AI is however not enough in itself to guarantee successful scaling, if the initiative is not backed up with sufficient strategic planning and resources. Implementing AI is a multi-disciplinary challenge and the success of the first pilot projects are dependent on different factors that can be further divided into 3 categories: technical, business and people related challenges.
Have your data readily accessible
Even though AI is not just purely a technical challenge to be solved, it is still highly data driven technology, and without solid existing data architecture, companies will have harder time to get going. Finding new use cases and rolling AI solutions into production also needs skilled workforce, which is in high demand. It is therefore no surprise that the companies with strong IT background are more likely to be the first ones to manage the scaling of AI as they can leverage their existing infrastructure to smoothly transition their operations to support AI implementations. (Paris et al., 2017) Building internal capabilities and knowhow is important but takes time and resources. The learning process can however be sped up by outsourcing the more challenging projects and seeking consultancy when needed.
Use cases, use cases, use cases!
But more notably from the business point of view, the company must have suitable problems for AI to solve and preferably a clear business case is required to back up the initiatives. Failed pilots on wrong use cases can easily dishearten the managers since the economic benefits are not materialized and the budgets are withdrawn before AI has been given enough time to gain a stable footing. For the early development and testing of AI, light POCs (Proof-Of-Concepts) as fail-fast, minimum-viable-product approach works best to enable focus on appropriate use cases. This way the scarce resources can be used intelligently, and scaling becomes easier.
Managing the rollouts properly is as important as technical skills to overcome the “last mile” problem
Facing opposition from the managerial level is one issue, but more crucially, the people working with AI must also be convinced of the usefulness of AI. Most unexpected challenges for AI implementors typically emerge when the well-functioning AI solution gains resistance from the employees. Educating the workforce on the benefits is therefore crucial and inspecting AI projects with purely technological lens will guarantee to miss these challenges when the solutions are rolled out to production. Change management principles help to overcome this “last mile” problem when the insights and gains from AI need to be integrated to the behavior of the people and organization as a whole. (Chui et al., 2018)
Despite the challenges, there has never been better time to start developing internal AI capabilities, as resources and guidance are plentiful. By consciously implementing AI, companies can set themselves up for successful adoption and scaling of AI to boost their technological capabilities and overall business performance.
Aleksi Hentunen, Senior Solutions Consultant at Sisua Digital
The gathered insights are based on the findings of my master’s thesis under the Master’s Programme of Industrial Engineering and Management at Aalto University. The researched sample consists of five well established Finnish organizations that have managed to scale their AI initiatives. If you are interested in the topic and would like to read the whole thesis, do not hesitate to contact via email at firstname.lastname@example.org
Chui, M. et al. (2018) Notes from the AI Frontier: Insights from hundreds of Use Cases.
Gerbert, P. et al. (2018) ‘The Big Leap Toward AI at Scale’, The Boston Consulting
Group, BCG Henderson Institute.
Paris, E. H. et al. (2017) Artificial Intelligence: The Next Digital Frontier?