When it comes to data integration and automation, there’s a series of complex decisions that business units and IT departments must tackle to determine the best course of action for both the short and long term. Application Programming Interface (API)-level integration, where systems are integrated beneath the user interfaces (UIs) to ensure a steady flow of data throughout interrelated systems, is one viable option.

One other major option to consider includes the popular emerging robotic process automation (RPA) software, where “robots” can be configured to capture and interpret existing applications for processing a transaction, manipulating data, triggering responses and communicating with other digital systems.

We’re often asked where RPA fits relative to APIs. Our view is that this is an “and” proposition, where RPA can help achieve desired process outcomes in situations where manual workarounds can be eliminated and costly development specialization can be avoided. This is in line with our view of integrated automation, where a “toolbox” approach is best to obtain the functionality needed to enable a digital business transformation. Find your balance, blend the right mix.

MuleSoft bridges the old world of legacy with the new world of open-source API integration

The 2018 acquisition of MuleSoft (see POV) was Salesforce’s largest (by purchase price) in history until the Tableau acquisition in June 2019. You can find HFS’ view on the Tableau acquisition here and our view on the MuleSoft acquisition here.  Both acquisitions are fundamentally about data. While Tableau makes it easier to present and visualize data in a useful and meaningful way, MuleSoft was the previous foundational step. It addresses integration challenges, laying the groundwork in accessing and surfacing data from multiple silos—internal and external. Under Salesforce’s ownership, the intention is that both products continue to serve existing and new customers irrespective of whether other Salesforce products are in play.

MuleSoft’s Anypoint Platform is a hybrid API-management platform with design tooling, unified management capabilities, and a single runtime environment. It connects applications, data, and devices on-premises and in the cloud with an API-led approach. Major competitors include Dell Boomi and Informatica.

The MuleSoft customer base is rich with long-established, traditional companies seeking to reinvent and modernize their application landscape, including recognizable brands such as Coca-Cola, Unilever, Barclays, and Verizon. Companies that built their IT and business operations product by product now need to reorient around customer-centricity. They strive to move forward in a controlled way to transform IT landscapes around heightened customer expectations.

In contrast, there are younger digital-native companies in the process of scaling up that have opted for MuleSoft’s integration technology, too: Netflix, Airbnb, Square, Splunk, and Atom Bank. Atom Bank halved the number of APIs in its mobile banking experience on adopting MuleSoft, while Splunk integrates Salesforce and Netsuite with MuleSoft for real-time order processing. Scale and real-time processing represent two of the most difficult requirements that systems architects and developers work toward. Young companies, opting for this method of integration squarely places APIs in a modern systems architecture, especially as they are unburdened by a legacy of core systems written decades ago and often have the luxury of piecing together greenfield architectures.

A single view of the customer, despite many promises and much effort along the way, still eludes many businesses. In the past 20 years, the number of shadow IT, applications, and business processes exploded while attempting to smooth over these shortcomings—this is how “workarounds” are born. Humans are lumbered with manual work to compensate for the lack of data flow through systems of record and systems of engagement, giving rise to far-from-ideal day-to-day realities in the largest of enterprises, such as swivel chairing and transfers of data from one system to another via spreadsheets.

Exhibit 1: RPA versus APIs, complementary toolbox brethren—not a celebrity deathmatch

^{Source: HFS Research, 2019}

Use API integration when system complexity increases and to achieve scale

In order to merit the time, skillsets, and cost of integrating at the API level, the volumes of applications, interactions, transactions, connectors, and devices must be sufficient. Scaling requirements matter, too. In short, the higher the numbers, the more investment is merited.

But, API integration is not always an option. It’s off the table if you do not have ownership of, or the right to alter, applications’ overall connectivity. For example, options are limited when executing processes in a shared service or BPO scenario. Also, many core legacy applications do not have an API option, and integrating old legacy systems of record with modern systems of engagement is not straightforward; bespoke point-to-point integrations are sometimes unavoidable. Similarly, APIs can’t help if data is unstructured, and this is where intelligent automation has stepped in to fill the void using elements of artificial intelligence (AI) to ingest pdf file formats and voice or chat conversations.

APIs or RPA alone cannot often accomplish your desired outcomes – you need to figure out how to deploy them with an integrated approach

Standard business processes often need many applications (often in the double digits) to work. As the number of systems increases, tightly integrating systems on a point-to-point basis ceases to be a viable option, especially as business models based on ecosystems come into play and require interactions with external systems. APIs or RPA alone cannot accomplish the job. If APIs are available or appropriate for integration needs, they may be the right solution. RPA offers surface-level integration, logging into applications through the user interface, and it can fill the gaps left at a process level. It is business-unit friendly as it requires less technical expertise (scripting, not development), is relatively quick to implement, and carries lower costs. As with APIs, RPA is also increasingly being stacked with elements of AI such as computer vision or natural language processing to give it cognitive capabilities and thus extend its functionality. Please refer to Exhibit 1 for a comparison of APIs to RPA.

The Bottom Line: There is no one-size-fits-all solution for data integration needs. Weigh your alternatives and needs to achieve desired results and minimize technical debt. 

Integrated automation, as opposed to piecemeal automation, is what businesses need to be aiming toward. HFS has advised taking a “toolbox” approach to better, faster, and more cost-effective processes, where software decisions can be made based on achieving the desired outcomes for both short and long-term needs.

Enterprises need to be sure they are not trading results now for a mess of future technical debt. Pick the right tool for the job, not the most convenient.  For example, an RPA tool may be excellent for fixing a legacy workaround activity and improving the data flow, but how will that be impacted in the future when a significant process reconfiguration is implemented?

This, by the way, is exactly why business operations and IT need to work together—to ensure extensibility. The primary aim here is making information flow steadily across and through a network of connective tissue between applications, tackling data silos, surfacing data, and, ultimately, getting value from data.