Be Realistic

Evaluating software can be tedious and time-consuming. It is also tough to get through the inevitable hype. What is real vs. the fluff? It can be tough to tell.

But the decision requires much care and diligence. And it is not just about the expenses.

“A relationship with an RPA vendor could be akin to a marriage that doesn’t end well,” said Thomas Phelps, who is the CIO at Laserfiche. “Unfortunately, it’s relatively exciting to get into, but very difficult and painful to extricate yourself. Vendor lock-in is inevitable. You can’t click a button to download all the process automations you built, and repurpose it on another vendor’s platform. Switching costs will be high if you change to another vendor.”¹

What’s more, the capabilities of an RPA system can vary widely and many of them will likely not have much relevance to your business. The fact is that there is no perfect solution. In other words, there will be certain areas where you’ll need to make compromises.

Check Out Third Parties

A good way to start with creating a short list is to refer to the surveys and research from analyst organizations like Forrester and Gartner. They have the resources and expertise to do deep dives on RPA software.

“Look at Gartner Peer Insights, which are validated user-sourced reviews, to get perspectives from real users—from CIO-level to the enterprise architect,” said Phelps. “Peer Insights provides a wealth of high-level information on the overall customer satisfaction with a vendor, vendor evaluation and negotiation process, implementation, technical support and other areas.”

Another good source is to look at software review sites. They will have their own analysis and user-generated ratings and comments. True, some of the content may be skewed but you should still get a good sense of the main themes.

So for the rest of the chapter, we’ll take a look at the criteria for making an RPA selection.

Minimum Capabilities

Because of the fervent interest in RPA, some software companies have used the term to describe themselves – even if there is little connection! There is nothing new about this. But of course, it could mean terrible results for the customer.

This is why you want to make sure the software has core RPA functions. At a minimum, this means that there must be the creation and deployment of bots, which carry out the automation. While a key part of this is screen scraping, it is also important to have built-in integrations for common applications. And there should be a robust management console for the bots (such as allowing for tracking, orchestration, and analytics) as well as ways to handle business exceptions (this is when a person needs to intervene because the bot does not work).

The first stage of the evaluation should also include a Proof of Concept (POC). After all, RPA vendors often have trial periods for their software. So why not start a small project to get a sense of the features and capabilities?

This often gets confused with a pilot program. But there are clear differences. A POC usually happens in the early stages and is for a development environment. The pilot, on the other hand, is for a real-world application that will go into production.

Who Is the User?

Yes, this question seems kind of simplistic. But the reality is that – when making the decision to purchase RPA software – there could easily not be enough attention paid to the needs of the end user. All in all, this means there will be a risk of a lack of adoption.

“In evaluating RPA software, make sure you select the right tool for the right audience,” said Niraj Patel, who is the managing director of artificial intelligence at DMI. “For example, for front-end RPA, which connects consumers to bots, you will need to select conversational technologies, as opposed to backend RPA which is core system software-oriented. Also, for front-end RPA, you may want to make your call center agents bot owners. Platforms can allow them to measure and monitor bot performance. With backend management, technical personnel should manage and evaluate the quality of the work that is performed.”²

For the most part, the end user will not be a technical person – instead, he or she will be an administrative or a business person. “There is a trick to evaluating RPA vendors – and it’s a major one, because this deviates so much from, say, ERP vendor selection,” said David Lee, who is a senior research analyst of applications – Data and Analytics at Info-Tech Research Group. “Active business participation is so critical, and in fact, IT won’t be the one testing out the bot creation – the business must be the ones building the bots. This gets missed a lot, because too often businesses view this as another technology solution, and expect IT to take it away and come back with a recommendation. This does not work.”³

Funding

For the first half of 2019, venture funding remained on a torrid pace, with $62 billion in investments in US start-ups.⁴ Assuming the pace continues, it will be an all-time record for the year. Foreign markets have also been strong at about $104 billion.

Yet the funding markets tend to focus on a relatively small number of deals because the investment amounts have increased significantly. In other words, even though a category like RPA may be red hot, it could still be tough for many of the players to get sufficient funding.

This is certainly something to note when it comes to evaluating a vendor’s software. Will the company have the resources to invest in R&D, infrastructure, and support? Also, can a small start-up compete when it comes to hiring talent? And will there be enough resources to pull off acquisitions?

Ecosystem

Even the large RPA players do not have the resources to build a true end-to-end solution. In most cases, a customer will likely need other software to fill some of the gaps (and some of them can be quite glaring).

Because of this, you want an RPA platform that has a thriving ecosystem of partners and integrations. A clear indication of this is when there is a bot store (i.e., an app store to download apps). This will greatly cut down on the time for development and also should improve overall performance of the bot. According to Automation Anywhere, its Bot Store has realized cost savings of 50% and 70% from faster automation of processes.⁵

Costs

Of course , this is usually the most top-of-mind consideration when it comes to selecting an RPA system. But if there is too much focus on this, you may ultimately buy the wrong platform, which could severely hamper your project. Keep in mind that the pricing among RPA vendors is quite competitive. So ultimately, the main part of the analysis is whether the software can solve your problems. It’s really that simple.

But when it comes to the costs, there are notable minefields. Some platforms require a large amount of consulting for the setup. These costs can be two to three times the amount for the software. For large enterprises, you can easily spend multimillions on consulting.

This is not to imply that the money is wasted. In certain circumstances, there may need to be much reengineering. And this is fine. But the key is that the service providers must be clear about what to expect.

OK then, so is there any way to get a general sense of how pricing works? Well, things can vary, but let’s take an example that should give enough context on how things work. With a standard RPA vendor, there could be annual payments for the studio/bot designer and the orchestration system, which could easily be over $20,000. But the main cost will be for the bots. Note that most RPA vendors charge on a per-bot basis and the unattended ones usually are much higher. For example, an attended bot could be $1,000 to $2,000 each, whereas the unattended bot could range from $5,000 to $10,000 each (all these costs are payable each year).

The per-bot pricing can also be far from straightforward. The fact is that bots among the different software platforms are different, which makes it difficult to make true comparisons. As a result, you should ask questions about the pricing – and see what you are really paying for. If anything, this could lead to a discussion for lower pricing, especially if you mention features of bots from rival firms!

But some RPA vendors – especially those that focus more on the business process management (PBM) market – may have other pricing models. One would be for inputs like the number of full-time equivalent (FTE) employees or time and materials. And yes, there is the output model, which is based on the volume of transactions . Although, in the end, the per-bot pricing model is the main approach in the RPA world.

Training and Education

A good way to evaluate an RPA system is to take some of the courses. Is the software easy to use? Does it have the features you need?

Some RPA vendors also have free downloads of the software or community editions. Such options are definitely worth looking at.

Then check to see if there is an online forum. What are the comments? Is there a thriving community? Or has it been a while since anyone participated? Or are there many complaints about the software?

Support

You want an RPA vendor that provides full-on support. This would include self-service options like videos, forums, courses, white papers, documentation, and blogs. Also what is the level of technical support? Can you call 24/7?

Then there is the SLA or service-level agreement, which is a contract that sets forth the level of services the vendor will deliver.

Best-of-Breed vs. End-to-End

The seemingly never-ending debate with business and enterprise software is the difference between best-of-breed and end-to-end solutions. True, the nice thing about an end-to-end solution is that there is usually less training and the costs are lower. But of course, the big drawback is that some of the features will be subpar. Even the world’s top technology companies have major deficiencies in their platforms. It’s inevitable.

With RPA, it’s not realistic to have a true end-to-end solution. This is definitely the case for global enterprises, which have extremely complex needs. It’s actually not uncommon for such companies to have several RPA solutions. Yet as noted earlier in the chapter, it’s a good idea to look at those platforms that have a rich ecosystem of partners that offer complementary technologies say for OCR, ML, and NLP.

Thought Leadership and Vision

Yes, such things are nebulous. But when it comes to business software, you want to see if the vision is in line with yours. When it is, the results can be very powerful.

To get a sense of the vision, you can check out a company’s Twitter page, blog, and YouTube channel (viewing the company conference videos can be particularly helpful).

Interestingly enough, you may realize that – with this process – a company’s vision is muddled!

And in some cases, a company’s vision may verge on hype. It’s as if the technology is too good to be true. If an RPA vendor claims that its software can do just about everything, for a low price, and there is little effort – then be wary! This may just be a way to gin up more PR and sales.

“RPA isn’t a magic bullet to do every menial task an employee takes on,” said Arvind Jagannath, who is the director of product management at AI Foundry. “It’ll take some tweaking and adjusting in the early stages before errors start to disappear. It’ll also require an upfront time investment to get acquainted with the tool.”

Industry Expertise

Even though many RPA vendors claim that their platforms are a fit for many industries, the reality is usually something much different. There is instead probably concentrations in certain verticals.

So you need to do some research on this.

Security, Monitoring, and Deployment

As we’ll go into more detail in Chapter 8, there are several potential security risks with RPA software. This is another reason why IT should be brought into the RPA process early on. By doing so, you can greatly minimize the problems.

A solid RPA system should also have strong monitoring as bots can be brittle. So you want a way to detect the issues.

What’s more, how will the software be deployed? Will it require an on-premise setup? Or can you host the software from the cloud?

What Type of RPA?

Do the processes you want to automate require human intervention? Or do they work behind the scenes? Such questions are critical when evaluating RPA software. For example, your needs may require only one type of automation, such as attended or unattended (we discuss these in Chapter 1).

Granted, you might want both attended and unattended bots (which, by the way, is common). Then again, the key is that you understand this early in the RPA implementation process so you will make a better choice on the software.

The Design

One of the keys to Steve Jobs’ huge success was an obsession with cutting-edge design. He once said, “Design is a funny word. Some people think design means how it looks. But of course, if you dig deeper, it’s really how it works.”⁶

It’s a crucial distinction – and something to think about when selecting an RPA solution. Let’s face it, all of them will boast that their design is intuitive and easy. Almost magical! But if you try it out, you’ll often have a much different experience. The sad fact is that much of enterprise software falls well short when it comes to design.

Because of this, there needs to be some time spent with evaluating the workflow, UI (user interface) and UX (user experience). Is it something that workers can learn quickly? And does the software require technical skills – say, some coding?

Next-Generation Technologies

An RPA company will not provide a detailed look at the product road map. Part of this is due to competitive reasons. But a company also does not want to commit itself to tough deadlines . If there are continual delays, customers will eventually stop listening and could go elsewhere.

Despite all this, there are still certain areas that an RPA company should be focused on. Perhaps the most important is AI or ML. These technologies hold tremendous promise in terms of boosting the impact of automation, such as with handling unstructured data.

“Bots as we know them today are best suited for highly repeatable, structured data,” said Bill Galusha, who leads Product Strategy for RPA and Content Intelligence at ABBYY. “But 80%+ of enterprise content is unstructured. What enterprises are requiring for their digital transformation projects is AI technology that complements the bots, making them smarter and able to transform unstructured documents into structured content. This removes process friction and accesses more value from business content. Any type of use case involving content is a prime candidate and many of these processes have customer touchpoints, whether its onboarding a customer, filing an insurance claim, or requesting a loan or line of credit. The concept of a ‘skill’ is something that resonates with business users . This is where an enterprise using an RPA tool or other automation tool could easily add a skill to a digital worker to give that bot more advanced training and understanding for processing content. We see this growing in popularity as marketplaces become more established and offer hundreds of these potential skills that increasingly make the digital bot smarter.”

Conclusion

Making the decision for the purchase of RPA software is definitely a big one. It can be scary. As a reminder, it is a good idea to seek help, say, from a service provider, consultant, or implementation partner.

After all, as seen earlier in the chapter, there are lock-in issues. So when making your evaluation, it’s essential to take a long-term approach. Will the vendor have the resources and vision to go where you want?

There’s no perfect answer. But at least you will have done some due diligence to lower the risks.

In the next chapter, we’ll take a look at another critical part of the RPA process: assembling the CoE (Center of Excellence).

Key Takeaways