A version of this article was first published in The Future of Sourcing Magazine, “Sidestep High Failure Rates: Make Your Digital Transformation a Success.” An edited excerpt is provided here.
Everywhere you look in the business press, the topic of digital transformation is at or near the top of the list. It is hard to argue with the notion of transforming an organization to become more adaptive, efficient and customer-friendly. The backbone of this transformation is often automating manual and paper-based processes through AI.
Many real examples of significant returns exist by transforming one process or workflow, so the idea of transforming an entire organization to achieve remarkable results is enticing. Organizations that have adopted digital transformation are seeing significant growth when compared to their lagging peers, according to McKinsey & Company research.
Key Approaches to Any Digital Transformation Project
McKinsey suggests that there are five approaches to plan for and incorporate into any digital transformation (DX) project:
- Ensuring lean process design;
- Digitizing the customer experience;
- Using select business process outsourcing;
- Incorporating advanced analytics to aid with decision-making; and
- Introducing intelligent automation for non-core human tasks.
All of these make sense; however, some speed bumps along the way may amplify the risks of any DX undertaking. The reality is that few organizations are ready to attempt such an endeavor. The obstacles are enormous. Mapping and documenting processes, culture and change management, access to data science skills, access to the data itself and managing many moving parts of an implementation are just a few of the complex tasks that an organization must tackle.
As a result, these capability problems have led to a change of thinking both on the part of enterprises and by the organizations that provide services to them. Let’s examine some of the key challenges along with potential strategies to resolve these challenges.
Addressing Challenges with Proven Strategies
The talent/skills shortage is one key enterprise challenge cited by just about every digital transformation study conducted. According to McKinsey research, 24% cited lack of digital talent as significant; 39% of respondents to the Riverbed research cited the same problem. The talent/skills shortage runs across the spectrum of DX-oriented skill sets. As more processes take advantage of machine learning, a major consideration is the availability of talent who have practical implementation experience with these powerful technologies.
The reality is that few organizations have a data science practice intended to collect and curate beneficial and reliable data sets for use by AI and data analytics. The logical remedy—at least near-term—is use of contractors and consultants as “staff augmentation” to enable organizations to progress with their digital transformation projects. Some of these skill sets are just in too much demand and can be more expensive than many organizations can afford. Utilizing consultants and contractors allows the organization to focus on implementation challenges without the requisite time and cost of hiring and training staff. Contingent workers offer extensive experience that reduces project risk.
Access to Good Data
If a lack of needed skills is a challenge, so too is access to adequate data to power these new cognitive automation tools. For machine learning to work well by providing unattended automation and insights into data, organizations need the ability to curate large data sets. These data sets need to be free from bias and statistically representative. There is initial anecdotal evidence that machine learning will highlight a new problem, namely the data rich versus the data poor. This is not only a problem with the lack of talent/skills necessary to collect data sets, but access to quality data as well.
Data scientists (55% of the respondents) reported that they face the challenge of the quality and /or quantity of training data, according to a 2018 study by Figure Eight. As a result, a cottage industry has sprung up focused on meeting the need for high quality data. These new services, such as Cloud Factor, outsource the data-tagging task and promise quick turnarounds. However, outsourcing the tagging of data does not solve the statistical side of the equation. This is where synthetic data is increasingly discussed. The idea is that machine learning models can create data that realistically resembles an organization’s production data thereby providing high-quality training sets.
Dealing with Budget Constraints
Slightly more than 50% of organizations are constrained by limited budgets when it comes to implementation of DX strategies, according to the Riverbed Technologies study. In a world of competition priorities, it is not unusual to have different projects fighting over the same dollars. Yet, given the relative importance of digital transformation as a whole, it is hard to understand why organizations are not setting aside money specifically for digital transformation projects. Regardless, novel approaches are under consideration to benefit from DX while removing the upfront costs.
One is to outsource the process to a service provider that has significant experience applying automation capabilities to an end-to-end process. Increasingly, these services are no longer based on the traditional “lift and shift” full-time equivalent (FTE) model. Instead, they are based upon output or outcomes. This means that budget dollars become less of a constraint than with traditional large-scale IT projects where the technology must be procured outright. Since many of these contracts allow for reduced costs, savings can be funneled into other DX initiatives.
Other budget-constrained organizations are starting with processes that are the most expensive, but less strategic and more contained, thereby reducing risks and total costs. Once implemented, the net results are that by creating efficiencies that lead to direct savings, the projects become self-funding and can even supply net increases in cash flow to fund larger projects.
Process Complexity and Visibility
Approximately 40% of respondents to the Riverbed study indicated that a key constraint is the complexity of existing processes as well as the inability to “see” the process end-to-end. This is most commonly attributed to cross department workflows. It stands to reason that many organizations developed processes over time and without an overarching strategy. Therefore, the effort to reduce complexity and visibility into each step of the process was sacrificed at the altar of “just get it done” tactics.
That said, processes that are strategic to an organization are not always complex. In addition, an organization can thoughtfully lay out a transformation roadmap that places emphasis on quick wins with high impact, low risk processes. Grading each process on attributes that matter to an organization results in a stacked/ranked set of processes that organizations can tackle one at a time. For instance, an organization can grade processes based upon the level of data available, repeatability of the process and process value. Those processes that rank high on each represent those that can contribute a high level of value but that are less risky than others.
|
Process / Data |
Best |
Maybe |
Little ROI |
Forget |
Next Best |
Risky |
|
Data Accessibility |
High |
High |
High |
Low |
Low |
Low |
|
Process Repeatability |
High |
High |
Low |
Low |
High |
Low |
|
Process Value |
High |
Low |
Low |
Low |
High |
High |
More Substance than Hype
Overall, there is no question that digital transformation is more substance than hype. However, even though many organizations have a concerted effort underway, a lot of ambiguity still exists over the right way to approach DX. While simple rules can be used to create a generalized framework suitable for any organization to help frame projects in a grounded manner, the right path will depend upon the specific needs and “personality” of each organization.