Source: Knowledge Capital Partners’ survey of UiPath customers

RPA implementation is pretty straightforward if you have a clear and detailed roadmap. As you can see above, most RPA deployments take less than 2 months. This includes time required to configure, test and launch RPA bots into production.

Below, you can find a detailed roadmap for RPA implementation. If you want to know more about RPA first, here is the most comprehensive article for that.

Identify processes that will yield the greatest benefits

To maximize impact of RPA, identify most impactful processes. These processes tend to be

  • Impacting both cost and revenues: Most impactful processes are expensive and touch customers. For example, quote-to-cash can be expensive if pricing rules are not clear. Speed and effectiveness of quote-to-cash process can definitely make or break a sale. Such processes are good candidates for RPA if they can be automated.
  • High volume: One of the key benefits of RPA is reduction of human effort. You should start automating your highest volume processes first.
  • Fault tolerant: If a process can not handle any errors, then its automation should either be deprioritized or there should be a quality control process to ensure that automation errors get caught. RPA bots rely on user interface (UI) to carry out their tasks. They can have errors due to UI changes or process changes. For example, it makes sense to automate invoice-to-pay process for most companies. However, payments above a certain value would need to be approved by humans.
  • Error prone: The more manual errors in a process, the more benefits your company can get by automating such a process. Manual mistakes can cause significant customer experience or regulatory problems especially in customer facing processes.
  • Speed-sensitive: Any processes that can delay delivery of services to customers are good candidates for automation as automation can make processes instantaneous.
  • Requiring irregular labor: Since finding temporary labor is difficult. processes with irregular labor demands force companies to employ for peak demand which is inefficient. RPA bots can easily scale up or down, easily managing peak demand.

It is almost impossible to know every process in a large organization. Another great source for automation ideas is commonly automated processes. We have prepared a list of common RPA use cases and case studies. The most popular process to automate is invoice automation, feel free to read how RPA, complemented by AI vendors, can automate invoice capture and processing.

Process mining is another alternative to identify the most suitable processes to automate. By utilizing the performance metrics, process mining tools can show automation rates and potential for businesses. To learn more about how to use process mining for RPA implementation, feel free to read our in-depth guide.

Select processes that can be easily automated with RPA

Such processes tend to be:

  • Rules based: Ideal processes can be described by specific rules. RPA bots need to be programmed and if the rules of the process can not be programmed, then that process is not a great candidate for RPA. AI can be trained with complex rules and even uncover rules that are not apparent to human operators. However, automation of such processes requires careful observation of RPA results since there may be cases where AI incorrectly identifies rules.
  • With few exceptions: This is similar to the “rules based” criteria above. However, some processes have so many undocumented rules that even if they are rules based, it is time consuming to identify all rules via interviews with domain experts. Such processes are not good candidates for automation.
  • Company-specific: Is this a process that all companies undertake in the same way or is it unique to your company? For example expense auditing takes place in a similar fashion in most companies of similar sizes. Building an RPA system for expense auditing would be costlier and less effective then just using a solution built for such a process. And in case you are wondering if there’s a custom tool for a process you want to automate, ask us, we probably know about it.
  • Mature: Automating a process that is changing every day is a waste of time because developers will spend a lot of time on maintenance. Stable processes are good candidates for automation.
  • Not on the roadmap for new systems: Replacing legacy systems can automate processes even more effectively than RPA. RPA bots need to rely on screen scraping and may introduce errors. Additionally, installing two automation methods for a process does not make sense.

Finally, even if a process is not a good candidate for automation as a whole, it could possibly be broken into automatable sub-processes that yield large benefits when automated.

Instead of our 2 step process for identifying the right process, you can also use WorkFusion’s approach. Their approach involves a few more steps than ours. Both approaches aim to identify and prioritize processes for automation.

Major differences between the 2 approaches is that we recommend

  •  designing process improvements after you get management alignment on which process to work on. Since process improvement is time-intensive, it only makes sense to do it when the whole organization is on the same page.
  • you check your list against commonly automated processes. They are a great source for automation ideas
Source: WorkFusion’s Automation Quickstart Guide
Source: WorkFusion’s Automation Quickstart Guide

Convince the organization

Even in automation, it’s really about the people.

Get management and team buy-in

Get management buy-in

Though “move fast and break things” mantra is making itself heard across the enterprise, management buy-in is still a thing if you work for a large company. Knowing the process to be automated and possible benefits from automation, you can build a compelling case for company leadership. The keywords are RPA, AI and Return on Investment (ROI). The first 2 are topics that executives feel compelled to invest in these days. As for ROI, an executive should demand to know the ROI in any investment.

Get team buy-in

Are you automating an in-house process or an out-sourced process? If it is an outsourced process, you are essentially just bringing in savings and the team managing the outsourced process will likely be happy to manage the automated process.

However, it is a completely different situation with an in-house team. No one wants to wake up one day and discover that their job is redundant. It is necessary to have an open and honest discussion with the team about what automation will bring. Some points to cover are:

  • Convince the team of the change. It is easy to say and hard to do. Unless you have strong management support, it can be impossible to convince teams about the need for automation.
  • Design the new, leaner team
  • Formulate post-automation steps like how redundant team members will be upskilled and which teams they can join.

Convince other business units

  • IT/technology is the most important stakeholder as IT should be approving major tech purchases from a technical perspective. The important questions to ask are?
    • Did we buy any RPA technology before? For a list to check against, you can refer procurement to our list of RPA companies. If there were previous RPA projects, then it is best to learn who the vendor was, outcome of the RPA roll-out and learn from previous experience.
    • What are the major planned changes in the tech stack? For example, if the organization will work with Citrix in the next year, it is best to complete the RPA Proof of Concept (PoC) on Citrix.
  • RPA Center of Excellence (CoE): Most large organizations are setting up centers of excellence for RPA. CoEs would need to be involved in RPA projects.
  • Data/Analytics: Bots leave traces for every action they complete. Working with data/analytics unit to ensure that bots create valuable and easy to use data can save RPA teams from a lot of headaches down the road.
  • HR: If RPA is going to play an important role in your organization, it is important to train the staff on RPA. This enables company’s own staff can be empowered to maintain existing RPA programs and roll-out new programs.

Prepare for implementation

Understand the process

Processes can be understood by interviews with the operators that currently run the process but relying only on this approach is

  • costly – Interviews take time
  • error-prone – People have imperfect memories and are prone to numerous cognitive biases

An alternative is to combine interviews with analysis derived from process mining. Process mining software enables companies to analyze their logs to understand real-life process flows. By using real-time data and event logs, process mining software shows the actual conditions of processes to identify bottlenecks, unnecessary steps, and provide factual insights.

Feel free to read our research on process mining and the top companies in that industry to learn more.

Improve the process

Processes evolve due to regulatory pressures and market pressures. Though they are sometimes improved with top-down lean or 6 sigma projects, these are few and expensive. Therefore, most processes have significant potential for improvement.

Just consider the use of fax machines in the US healthcare system. Numerous media including Vox report how US healthcare system relies on hospitals sharing records with faxes or hand delivered documents because digital healthcare records were not built in a compatible manner across different institutions.

So before proceeding with the RPA implementation, it is worthwhile to look for improvements in the process as process improvements can

  • simplify the process
  • make it more understandable therefore reducing the necessary programming and auditing effort
  • improve customer experience

Implement the solution

Choose your partners

There are 4 types of partners that can help roll out an RPA solution. While it may seem like a quick & cheap approach to use just an RPA solution for RPA deployment, case studies indicate that companies save significant time and money by using a best-of-breed approach(i.e. using process mining and machine learning tools in combination with RPA).

Choose your process mining solution

While it is not mandatory to use a process mining solution for an RPA implementation, having a process mining solution helps companies

  • prioritize automation opportunities
  • understand processes in detail which facilitates RPA development
  •  track process changes after RPA implementation to ensure that the deployment is successful.

Choose your RPA solution

We have a detailed guide on how to choose your RPA technology provider. Since RPA is an evolving field with new solutions such as no code RPA, it is helpful to spend a bit of time to understand the latest list of things to pay attention while buying an RPA solution.

Choose the AI/ML providers to facilitate your RPA deployment

While RPA is great for automating rules based tasks, it is hard to automate more complex tasks such as getting data from documents with RPA. RPA providers may provide ways to program such functionality but since this is not their focus area, the solutions do not tend to be the best performing ones. It makes sense to research AI/ML providers that solve the specific problems you are trying to solve. For example, invoice automation is easy to automate with almost complete no-touch processing by integrating an API provided by a deep learning solution provider like Hypatos.

If you have access to unique data that could be used in process automation, you can also consider working with an AI consultant. We can help you find the right company to build a custom AI solution:

Identify partners to build custom AI solutions

If you want to learn more about custom AI solutions, feel free to read our whitepaper on the topic:

Download our in-Depth Whitepaper on Custom AI Solutions

Choose your RPA implementation partners

If your company is low on manpower, you can partner with a consulting company or BPO provider to help your company automate its process. We have a comprehensive list of RPA implementation partners.

Develop your solution

Initially, a detailed process map needs to be prepared identifying which parts of the process will be automated. Here’s an example from WorkFusion:

WorkFusion’s Automation Quickstart Guide

Contribution of subject matter experts from your organization is critical while preparing the process map. This is especially relevant if the process is not well documented. In our experience with large companies, most processes are not well documented.

After the role of RPA bots in the process are clarified, RPA bots can be programmed. Trade-offs such as quicker deployment vs more flexibility need to be weighted carefully while developing the solution. Following well established lean software development and quality assurance processes will ensure that business and technical teams are aligned and progressing.

A recent development is the launch of RPA marketplaces which provide reusable plugins/bots to facilitate RPA development. Implementation teams would be well advised to check out their RPA platform’s marketplace for readily available code and not re-invent the wheel. Feel free to read more from our guides on RPA marketplaces and reusable RPA bots.

Test your solution

Importance of testing can not be over-emphasized. We explained 3 different types of RPA. For example, in attended automation, minor differences in users’ systems such as some users using MacBooks or even different screen resolutions can lead to unexpected bugs. All major scenarios need to be thoroughly tested before the pilot. Using historic data enables more realistic tests

Run a pilot

  • Set targets for the pilot: These could be about accuracy (e.g. share of successfully processed invoices) or automation (e.g. cases completed without human intervention).
  • Run a live pilot: Each day, team in charge of the process reviews a random selection of bot output.
  • Evaluate pilot results: Run a detailed evaluation considering rare cases and difficult inputs. Only finalize the pilot when previously agreed targets are met.

Go live

  • Design the governance of new, bot driven process with support from the current team. For example put in place mechanism for maintenance to keep the bots functioning as the process changes.
  • Clarify roles and responsibilities
  • Build a fallback plan: A fallback plan will be helpful if the RPA solution requires rework after roll-out. Though such a plan would not be used most of the time, it is quite beneficial to be prepared when fallback is needed.
  • Go live: Communicate new process to all relevant stakeholders and go live!
  • Analyze results:
    • Monitor results: During implementation, process mining tools can track bot performance and measure the level of automation to see if the project achieved its goals
    • Record savings and analyze results to inform future RPA projects.

Maintain the RPA installation

In line with changes in the market and regulation, you will need to change your processes. Putting in a place a capable team in charge of the installation is critical for the future success of your RPA installation. Process mining tools can help teams monitor changes in processes and as a result identify when RPA bots need to be maintained/modified.

Now that you are ready to implement RPA solutions, feel free to ask us appropriate vendors for your company’s size and industry so you can proceed effectively:

Let us find the right vendor for your business


Blue Prism’s customer survey

Deloitte survey of RPA experience of 400 business users

How useful was this post?

Click on a star to rate it!

As you found this post useful...

Follow us on social media!

How can we do better?

Your feedback is valuable. We will do our best to improve our work based on it.


  1. Thank You for sharing a complete guide on best practices to be followed for implementing RPA. it’s Very Informative and Useful For us

  2. Great article. FYI, at the bottom you have a typo or wrong link in the following. The link is to a BluePrism study, not UiPath:

    We leveraged UiPath

  3. You put together a very thorough article. The simple truth is that sometimes projects fail, for a very large number of reasons. According to IBM Systems Magazine, up to 25% of technological projects fail downright, while up to half of them require extensive revisions by the time they are set to go. I’d like to add some basic pitfalls that I stumbled upon: not choosing the right processes to automate in the beginning, trying to implement robotic process automation on your own, not setting clear objectives for your automation strategy and not ensuring the scalability potential of your software robots.

Leave a Reply

Your email address will not be published. Required fields are marked *