Blog: Setting Boundaries in Hyperautomation Projects

Working in the automation space is both exciting and challenging. In a recent hyperautomation project with a client, I was provided invaluable insights into the complexities and nuances of automating business processes using MuleSoft RPA.

Here's a detailed look at my experience, the challenges faced and the lessons learned, along with a project glance on automating quotes in Excel and PDF formats.

How do we assess data limitations?

One of the initial hurdles was assessing the limitations of the input data. The client had vast expectations, yet the initial data provided was minimal. The resulting gap made it difficult to establish clear boundaries for what could be automated.

The solution

• Set concrete parameters: Establish clear criteria for the initial assessment to determine the feasibility of automation. This helps manage client expectations and ensures realistic commitments.
• Define scope early: Clearly outline what can and cannot be automated based on the data provided in order to reduce the risk of over-promising and under-delivering.
• Request diverse sample data: Ensure that the client provides a wide variety of sample data at the beginning. More data helps in accurately assessing the complexity and variations, leading to a better time estimate and more robust development.

Example

In our project, the client initially provided a few sample quotes in both Excel and PDF formats. We handled the exceptions present in these samples and developed the bot accordingly. However, after development, the client provided a number of additional quotes for testing, each with significant variations and limitations. 

By having more comprehensive sample data initially, we could have better assessed the variations and complexities, leading to a more accurate time estimate and development process. Changes made after the initial development phase took far more time than if those variations had been considered from the beginning. The diversity of sample data is crucial to ensure the bot can handle the full range of real-world scenarios it will encounter.

Can bots think like humans?

When it comes to Robotic Process Automation (RPA), a common misconception is that bots can think and make decisions like humans. However, MuleSoft RPA bots can only automate repetitive human tasks based on predefined instructions. They do not possess the cognitive abilities to think independently or make decisions.

What can we do?

• Communicate limitations: Clearly explain to clients that MuleSoft RPA can automate simple tasks by doing what a human would do on a regular basis, but it cannot think like a human and make decisions of its own unless Cognitive Process Automation comes into picture.

Creating an ecosystem where Salesforce flows, APIs and RPA bots are combined can automate many human tasks. For instance, a MuleSoft Composer might trigger an RPA bot, which then triggers an API to perform actions in Salesforce. However, this ecosystem can only handle predefined tasks and processes. The bot will follow the instructions given without the ability to think independently or handle unexpected variations the way a human would. This is where the limitation lies in automating processes beyond a certain point.

Example

Looking back to our project, the bots were designed to extract data from quotes. Without cognitive automation, handling unstructured or highly variable data in the PDF quotes was challenging. We communicated this limitation and worked on standardising the format of the data to improve automation accuracy.

Expectations vs. reality

Balancing client expectations with the reality of what the bot can achieve is crucial. We often face the dilemma of wanting to gain the client’s trust by promising extensive automation, but the bot’s capabilities are limited.

Key steps

• Establish clear communication: Regularly communicate progress and limitations to the client.
• Set realistic timelines: Based on a detailed initial assessment, set realistic timelines that reflect the true complexity of the tasks. Clearly outline what can be automated and what cannot to avoid any mismatch between expectations and reality.

Example

Initially, we assumed that Excel quotes would be simpler to automate than PDF quotes. However, the complexity of the Excel formats turned out to be equally as challenging as the PDFs. We ensured that future timelines reflected the actual complexities involved, making it clear to the client which aspects could be automated and which would require manual intervention. Our approach helped align client expectations with the bot’s capabilities, providing a more accurate project scope and timeline.

Should we standardise processes?

For specific processes like quote-to-cash, it's essential to have a predefined set of prerequisites that the customer must meet during the initial assessment. Having concrete requirements ensures that the project will work as expected and provides clarity and consistency.

Approach

• Predefine requirements: Ask clients for specific details and set defined limitations for different processes to ensure we can accurately assess the feasibility of automation, and set realistic expectations.
• Checklist of prerequisites: Before starting automation, ensure all prerequisites are met. It supports the ability to guarantee that the bot can automate up to a certain percentage, and the rest can be taken care of with little human support. The bot can handle more scenarios as the database grows.
• Use standard templates and accelerators: Convince clients to use standard templates for Excel and PDF files. Standardisation allows us to leverage an accelerator (a pre-developed chunk of the RPA process) to automate tasks more efficiently. Further, it saves significant effort and cost as we are able to replicate success for multiple clients. With minor modifications, we can achieve substantial results. The approach is particularly effective for automating processes involving future data that the client will receive, as opposed to existing data which may already be in varied formats.
• Sell automation packages: Develop accelerators that can be sold as packages, ensuring straightforward automation for future customers.

Example

We encountered substantial challenges during our project due to the lack of structured templates and the varied formats of the quotes provided. The absence of standardisation led to significant difficulties in automation and created unrealistic expectations.

How should we value our projects?

Valuing projects based solely on time can be misleading. Instead, it's important to consider additional factors that reflect the true value delivered.

New pricing models

• Value-based pricing: Charge based on the value provided or the time saved for the customer, rather than simply the time spent on development, aligning the cost with the benefits received.
• Flexible pricing models: Use different pricing models for support work, such as hourly rates, while ensuring the main project cost reflects the overall value delivered.
  
  

Tool limitation

Currently, MuleSoft RPA lacks features that enable multiple developers to work simultaneously on a project, which can hinder collaboration and efficiency. To address this, we need to advocate for product enhancements that support simultaneous development and code merging on platforms like GitHub. In the absence of these features, clear communication and manual synchronisation are crucial to keeping the development process aligned.

Areas of improvement

To ensure we meet project timelines without overburdening the team, we can implement the following improvements:

• Client specification document: Create a detailed client specification at the project’s start to clarify goals and expectations.
• Manage change requests: Only agree to change requests if the client accepts potential deployment delays, or address changes during support work, not during development.
• Maintain work-life balance: Prioritise developers' work-life balance by setting realistic deadlines and avoiding last-minute rushes.
  
  

Conclusion

Navigating the challenges of hyperautomation projects requires a balanced approach, clear communication and a focus on realistic expectations. By working together and continuously improving our processes, we are able to both deliver value to our clients and maintain a healthy work environment for our team.