.png)
Centralized, Federated, or Democratized Operating Model? - That Is The Question
- The 3rd in a series of 10 articles - How To Successfully Scale Automation and AI in an Enterprise
As many as 70% of digital transformations fall short of their objectives (BCG,2020). Among these digital transformation initiatives, intelligent automation is one of the larger investment areas for companies today and seems to continue to be so in the future (HFS research, 2020). How can one ensure they succeed with their automation & AI strategy and become one of 30% percent that fulfills their digital transformation objectives? Based on my experience as head of automation & AI at a large Enterprise - driving it from scratch to 1000+ deployed automation & AI solutions, giving millions of hours back to the business and transforming hundreds of core business processes. I have seen the complexities in managing scaled-up transformations based on multiple technology platforms hands-on.
This is the third in a series of 10 articles where I want to share what I learned along the way and what I would do differently if I did it again today. This third article focuses on shaping your operating model. As KPMG (2017) writes, once you have set a clear vision and strategy, which was the topic of the last article, the next step is to establish an operating model for automation & AI across the company. This needs to be aligned with the organization and especially your IT department.
Centralized, Federated, or Democratized? - That is the question
There are different views on what type of operating model is best suited for these types of projects, and there are valid pros and cons of each model. All companies are different, so what works at one company might not work at another company based on aspects such as overall centralization level, maturity of automation & AI Initiative, maturity of IT organization, type, size of business, and other cultural matters.
From my own experience, I prefer deploying an operating model that leans towards more centralization, even more so in the earlier phases of the transformation. The reasons behind this are plenty, but one key reason is that I believe a centralized model is superior in terms of scalability, which is crucial if you want to succeed with automation & AI. In addition, without having data to back it up, I believe almost all successful automation & AI transformations have had a solid central organization. In contrast, the majority of all sub-scale initiatives out there have a weak and non-transformational central team. Note, however, that a centralized model does not remove the critical need of having local decentralized champions (read more about champions in article number 2 in this series) and business subject matter experts highly involved in the journey in the specific business or functions. I mean that decentralizing aspects such as platforms, operations or delivery will generally not make you faster and more successful like some local leaders will argue. Instead, it will make you sub-scale and create unnecessary friction with different teams overlapping each other and competing for the same resources.
I believe a centralized model is superior in terms of scalability, which is crucial if you want to succeed with automation & AI.
To succeed with automation & AI, you will need to make significant investments. If those investments are spread too thin following a decentralized model, you will likely not be able to attract the right competence or investments to scale the best platforms. Instead, you will end up with many teams doing more simple solutions, not properly leveraging economics of scale or the technology. For instance, you could end up with multiple teams doing 10-30 simpler RPAs on different platforms instead of one central team doing hundreds of larger, more complex, and value-adding RPAs while making additional investments in scaling Intelligent automation (like AI competence, AI platforms, low code solutions, proper democratization programs, or complementary process improvement products such as process mining).
Over time, a centralized approach connecting the dots will engage the company in more comprehensive end-to-end transformation efforts, not only covering more simple tasks or functional processes but end-to-end enterprise value flows. That is almost impossible to achieve with a sub-scale and underinvested central team.
Furthermore, there is one scenario where I think it makes sense to move towards decentralization. If many or most of the boxes below are checked, I suggest giving it a try.
1. Large enterprise with individual business units having the scale needed to operate end-to-end Intelligent automation in a decentralized model
- Likely at least $3-5 B revenue per business unit
- You are large enough to have a local business unit center of excellence with at least 30-50 full time employee (FTE)
- A large volume of business unit specific data
- Business unit processes and data more or less disconnected from the wider enterprise (limited re-use or value in cross-functional, cross business unit flows)
2. Already matured Intelligent automation program with common touchpoints in place between central teams and local hubs
Even for the case above, evaluating what pieces of the operating model should be centrally vs. locally-driven makes sense. For instance, it might make sense to have Platforms centralized even if Delivery and Operations are done locally.
There are also drawbacks of a centralized model, like speed, lack of transformational leadership, and lack of business competence. However, most arguments for decentralization that I have encountered have little to do with centralization vs. decentralization. Instead, they usually highlight aspects such as competence, trust, financial model, and partner model. For instance, nothing says that you cannot have deep business or functional competence in a centralized team or that you cannot set a financial model complemented with a partner model that enables supply and demand of resources to execute projects to match. The one thing that can make a centralized model fail, however, is weak, non-transformational operational leadership with limited executive sponsorship.
In any case, no matter what model you choose, you will face the democratization theme that is quickly emerging. This was something that I stumbled across when I was in The Large enterprise;
That was many years ago; today, democratization is a fact in many places. Nowadays, it even comes built-in. For instance, everybody with Microsoft Office access will have the ability to build their bot whether the central team likes it or not. Suppose you are a central team or a CIO. In that case, you will need to relate to this and have a strategy around it – to enable democratization, create speed, and remove barriers locally without hampering the overall centralized approach to large-scale transformation. If you get this under control, you will simultaneously manage the democratization and the centralized transformation journey. In the following article, I will continue my reasoning on the operating model, but I will focus more on the six pillars of an operating model that I believe are important to scale Automation & AI.
BCG, (2020). Flipping the Odds of Digital Transformation Success
HFS research, (2020). Spending on automation and AI business operationsworldwide 2016-2023
Kpmg, (2017). Scaling Robotics (RPA) - experience is key
