Episode IV: The 6 Dimensions of Scaling AI

From AI prototypes to scaled AI benefits — how to overcome the gap and make AI big in your organization

Applying AI in various fields offers great potential for organizations across industries. For example, for the global retail sector a $340 billion annual cost savings potential is estimated by 2022 through applying the right AI use cases in operations at scale. As the word is spreading, global spending of retailers on AI is estimated to climb to $7.3 billion per year by 2022.

Source: Capgemini (2018): Building the Retail Superstar: How unleashing AI across functions offers a multi-billion dollar opportunity

What are challenges for making AI big in organizations?

So, what’s the catch? — Well, realizing real business impact from AI in an organization is hard: According to a 2020 executive survey, about 73 % of leading organizations in different industries are convinced of that. It is not a very big surprise that only a low 15 % of those interviewed can say they have successfully deployed AI solutions at scale across their organization.

Interestingly, many of our clients — big organizations in different industries — do have first AI prototypes and pilot projects in place and they are well aware of the huge potential of AI applications. However, the actual challenge for most of them is to deploy AI at scale. Organizations are struggling to implement AI results and the insights generated into their business processes in a sustainable and widespread way to achieve the big benefits.

On the way to making AI big in your organization, there is a gap to overcome. This gap is characterized by several pain points. The most common pain points include:

· Use case selection. Picking too complex use cases first for AI application results in failure.

· Customer focus. AI solutions ignore customer demands and needs and do not “sell” right.

· Top management commitment. Executives are not driving deployment of AI in the business.

· Change management. People are not convinced of AI benefits and refuse to adopt new ways of working.

· Transparency. Many AI initiatives are isolated and “under the radar”, headquarters struggle to coordinate.

· Business-technology integration. Business and technology departments are decoupled.

· Data requirements. Data sources don’t meet requirements of quality and availability.

· AI deployment. Non-digital companies often lack suitable platforms to industrialize AI.

Source: Capgemini (2018): Focus Interviews with Industry Experts & Capgemini (2017): Turning AI into Concrete Value

What can your organization do to close the gap?

From our experience and history of supporting numerous clients on the journey of closing the gap towards scaling AI across their organization, we recommend activities along six dimensions:

· AI Strategy and innovation

· Ways of working

· People and change

· Insights communication

· Governance

· AI solution integration

We explain each of these dimensions in the following.

AI strategy and innovation

A solid AI strategy aligned with the overall corporate strategy and a structured AI innovation process ensure business impact and continuous AI evolution.

First of all, it is crucial to assess and shape AI initiatives from a business value perspective while staying in line with the overall corporate strategy. Taking the example of one of our clients in the retail sector, the main goal was to optimize revenue of ice cream sales in the points of sale (PoS) that our client caters through their field force. The first step for this was to operationalize an AI model that provides any field force agent with data-driven answers to two questions: “Which PoS should I visit next?” and “What should I recommend them to do or buy next?”. Moreover, a systematic innovation process helps to establish further AI use cases continuously to keep up with the speed of the market.

Ways of working

AI-enabled business processes along with agile methods and continuous monitoring lead to more efficient value generation.

Our consumer goods client was quite successful in developing an AI model prototype that recommends “next best actions” for their field force based on historical data. To really operationalize those insights, it was necessary to implement the AI model into the daily business routing of the field force agents. To do so, our client needed to analyze the as-is process first, then identify the relevant points and decisions to update with the AI components and eventually, integrate this into their sales force system as well.

People and change

A systematic change management approach and tested knowledge transfer tools empower your employees to work with AI results in their daily business.

To really make the updated business process stay in the organization and, more importantly, make the field force agents of our client comfortable with it, we communicated transparently and early and took the time for training events as well as workshops. This way, we could transfer the knowledge to them and take their feedback on what to improve. The open feedback approach and active involvement of the end users of the AI-improved business process did not only improve the process itself but also helped to increase the acceptance of the changes by the end users, i.e. the field force agents. If you want to convince the people in your organization of a change, you should make them an active part of the change process early on and foster frequent and open communication.

Insights communication

Advanced tools for visualization and interactive channels to the user help communicate value-adding AI insights.

Presentation and communication of AI insights using the right channels is always important, especially if your goal is to convince people. In the case of our client story, the field force agents were to be convinced of the benefits that the new AI-driven way of making sales decisions brings. To achieve this, there are numerous visualization and dashboarding tools available as well as automatized and interactive communication channels like chatbots and voice bots. Those have potential both for internal use with your employees and external client-facing communication, e.g., as an element of first-level support. In our client story, a big part of the success and acceptance of the AI solution was the way the generated “next best actions” were presented to the field force team. Easy-to-understand though fairly specific “one line” next best actions directly delivered to the IT system the sales force uses in their day-to-day business anyway, was key to this success. Those “next best actions” would sound like “Offer to extend ice cream product range with product XYZ” or “Offer to install a bigger ice cream cabinet with 20 or more slots”.


Governance capabilities in all relevant areas and a defined target operating model allow successful AI scaling across your organization.

To establish AI capabilities in the organization in a way that works smoothly and is prepared to grow with time, it is necessary to define AI-tailored roles and responsibilities, for example strategic roles like “Head of AI” or “Head of Sales” and rather operational roles like “Data Scientist” or “Regional Sales Representative”. In specific, there are several approaches to a target operating model that defines how business units interact with each other and with AI-focused units: you can go all-centralized with all AI roles and responsibilities pooled in a dedicated AI governance or rather decentralized with business unit specific solutions to AI governance. This depends greatly on the importance and extent of AI in your organization. Our client’s national ice cream sales organization went with a centralized AI governance lead by the “Head of AI” role mentioned above.

AI solution integration

Platform-based concepts with suitable tools and the right design principles enable a systematic and smooth integration of AI solutions into existing structures.

For integrating your AI solution efficiently and sustainably, getting equipped with the right tools that enable scaling of AI is crucial. Also, from our experience a systematic proof of concept approach makes a lot of sense in order to detect potential for improvement early in the process before making bigger commitments. In the case of our client, we made sure that the developed AI solution was integrated with the operational sales force system to be able to automatically update results on a regular basis. This provides the field force fast and easy access to the latest and most-adjusted “next best actions”.

How can your organization achieve a scaled AI and realize benefits?

The above overview of solutions along the six dimensions of scaling AI can help your organization to overcome the gap between AI prototypes and AI scaled across your entire organization. The benefits to expect include influencing your sales, boosting your operations, engaging your customers better, and generating valuable insights for your daily business.

To start scaling of AI, we recommend the following general approach:

1. Analyze. Select and finetune a set of solutions from the six dimensions of AI scaling based on your current status quo.

2. Innovate. Focus on value-adding use cases with high feasibility first.

3. Transform. Deploy suitable tools into your IT landscape, involve all stakeholders, find a governance model that fits your organization’s needs by applying the selected dimensions of AI scaling.

4. Deliver. Establish an open and encouraging culture to fuel AI adoption and define metrics to keep track and continuously communicate performance of your AI scaling initiative.

Through our scaling AI approach, we could realize added value for our consumer goods client: once successfully integrated into business, the data-driven sales agent covers PoS across multiple regions nationwide. The implemented self-learning system ensures that “next best actions” for sales agents are continuously improved with feedback from PoS, thus expecting to increase our client’s revenue by a large amount year-on-year. In summary, our client could increase customer satisfaction through personalized AI-driven product offerings, generate higher revenue from premiums with an optimized retention rate, and reach high stakeholder involvement through frequent involvement and transparent communication.

Stay tuned and follow this account for another article on this matter.

About the author:

Gerrit Wiltfang is a management consultant at Capgemini Invent with special focus on connecting business with technology. Find Gerrit Wiltfang on LinkedIn.

Read also:

· #valuefromdata: Episode I

· #valuefromdata: Episode II

· #valuefromdata: Episode III

Would you like to discuss the potential of scaling AI for your business or are interested in a demonstration of our solution portfolio? Please, do not hesitate to reach out to Dr. Katja Tiefenbacher.

Capgemini Invent is the digital innovation, consulting and transformation brand of the Capgemini Group. #designingthenext