The AI journey: From proving the concept to full-scale production

Publish date:

As Artificial Intelligence (AI) continues its evolution from a science fiction trope to a core component of the data-driven organisation, many business leaders find themselves increasingly frustrated by the challenges associated with developing AI applications.

In 2017, a third of the organisations surveyed for Capgemini’s ‘State of AI Report’ moved beyond the pilot and proof-of-concept stages for their AI applications (in a few or more use cases). In 2020, this rose to just over one in two organisations. However, just over one in ten of these organisations were able to successfully deploy their AI applications to multiple teams. So why did the majority struggle to take their applications to production?

According to those we surveyed, the three biggest challenges were:

  1. Lack of mid- to senior-level AI talent (70%)
  2. Lack of change management processes (60%)
  3. Lack of strong governance models for achieving scale (60%)

This highlights the clear need to generate a better pathway to industrialised applications incorporating AI. On the other hand, when reviewing the cohort that led in the adoption of AI within their domain it was found that most:

  1. Noted quantifiable benefits from their deployments
  2. Achieved benefits that either met or exceeded their expectations

Clearly there is a divide between organisations that struggle in this domain, and those that have taken the lead. This divide may be increasing, as a recent survey by the Harvard Business Review found that just 24% of their respondents claimed their organisation to be data-driven in past year – a reported decline from 37.8% over the year prior.

This was explored further in the report itself, concluding then that organisations needed to internalise four key principles to address the issue of scaling and industrialisation:

  • Kickstart the virtuous AI cycle: Continuously monitor model accuracy and business impact to amplify outcomes (Monitor & Amplify)
  • Build strong foundations that provide easy access to trusted, high-quality data, drawing on the right data & AI platforms/tools as well as Agile practices (Empower)
  • Build talent and collaboration with partners (Nurture)
  • Deploy AI through the right operating model and prioritise initiatives and ensure well-balanced governance, while at the same time embedding ethics (Operationalise)

Scaling AI with the four key principles

Scaling AI with the four key principles

Keeping these four principles in mind, our upcoming series of blogs will further explore how the inherent differences between pilots and industrialised applications can be used to identify how best to transition from one to the other. Specifically, we will discuss the following four focus areas:

  1. Value
  2. Operation
  3. Security
  4. Quality

We have chosen these four areas as we believe they represent the best ‘middle’ ground between the traditional pillars of project management and the Agile principles that are often used to incubate pilots and proof-of-concepts. For each of these focus areas we will present our analysis on what we believe can be used to assess application’s relative maturity, in addition to the transition from one maturity state to the other.

Through each blog we publish, we will also continue to ensure we provide reference to our reports, blogs, and other materials that we have published historically continuing to articulate our thoughts within this domain.

AI journey segments

                                                AI journey segments

 

Author


Danial Ali Khan

Danial is a consultant with experience in managing complex data and analytics services in Retail FS institutions, with a background in economics and an interest in how data is used within the business

Related Posts

Automation & AI

Assessing AI Value: Measuring the value of AI solutions

Date icon May 10, 2021

Why is it that businesses cannot find their feet with AI? It’s certainly not due to boardroom...