Humans and computers seem to have different roles. Humans can make complex, abstract decisions that computers are only now learning how to make, and even then in very narrow domains (think chess). And computers can do things much better than humans, namely doing anything related to computation. So when it comes to doing tasks in an organization, the question is: do I get a person to do this task, or do I automate it?
This guide aims to answer this question, about where it’s appropriate to automate tasks and not, and where you can find the benefits.
What is business process automation?
Business process automation is taking activities that would otherwise be done by a person and then getting a computer to do them instead. However, that doesn’t mean everything needs to get automated all the time. Automation can be built for part of the process you’re trying to set up and then having a human take over at some stages where it’s required.
The benefits of automating business processes
Automating business processes give a return on investment that increases over time
When a task is successfully automated, it saves time that would otherwise have been spent doing the task. Those savings get accumulated every time the automation is used, which can scale infinitely. You can put a number on the savings with this formula
Money saved from automation = time saved xX hourly rate of the person who was doing the task
Automating business processes have a big impact on employee leverage
When time is freed up from automations, it also allows employees to do higher-productivity tasks better suited for a human such as creative work, time spent planning, thinking, and developing others.
Automating business processes reduce the error rates on repetitive tasks
When people are assigned to do repetitive tasks over long periods of time, errors from human input start to arise. Because even repetitive tasks require you to pay attention, and any person’s attention will eventually slip.
Another problem is that people who are assigned repetitive tasks are typically assigned a lot of similar repetitive tasks that might be slightly different. So the lack of variety over time reduces a person’s attention span. Crucially, it becomes much easier to mix up tasks between each other because your mind has a harder time distinguishing similar tasks versus unique ones.
And this is where automation can shine. Computers don’t lose attention, instead they just read and carry out instructions given to them.
Get notified on new posts
How do I know if a business process can be automated?
The easiest way to discover if a business process can be automated is to ask yourself the following two questions:
- Is my process done in the same way, with the same steps, every time?
- Does my process require a person’s judgement at all?
If you answered yes to 1 and no to 2, then the process can likely be automated.
If these questions are not immediately answerable, follow these steps:
- List out the steps done for a task or activity that would be done by a person manually
- Identify which parts of the process are done in the exact same way every time, and which ones are done differently
- Understand the risks of automating your process
- If it’s not done in the exact same way every time, are there predictable inputs that can change how the automation can lead to the same outcome?
- Is there a machine learning mathematical model that can approximate the outcome if none of these criteria are fulfilled?
Should you get a person to do the work manually?
Getting a person to do the work 100% of the time (0% business automation)
Sometimes it’s better to get a person to do 100% of the work and not automate a process. For many tasks that require judgement, machines are not well-suited to do them at all.
Tasks that are well-defined and repetitive can be automated.
Using technology to do 100% of the activities (100% business process automation)
While getting to 100% of activities automated is usually the goal, it’s not possible in most cases. The reason being that often people will get to 95% or higher in automating their tasks, but the last 5% are a hurdle much more difficult to overcome. They are usually left last for a reason, meaning they were the hardest processes to automate.
Combining technology with human input to automate activities
Probably the most underrated and under-used form of automation, using a hybrid approach to automating processes is a great way to get a lot of processes semi-automated.
A great example of this is how the engineers at Paypal built automated systems to flag potentially fraudulent transactions, but then had a person on the back-end make the final judgement on whether the transaction was fraudulent.
Scaling product-market fit when it’s really a service
Say you’re a business that sells marketing software. One product sells reviews software that helps small businesses collect five-star reviews.
You realize that your current customers also need to get ranked on Google. So you decide to sell them Search Engine Optimization services, which requires you to hire a dedicated manager for every account. Clients are happy with the service and pay money, so you decide to expand your portfolio.
Everything is working great until you need to start growing quickly. You realize that different customers wants different things, so the amount of manual work increases depending on the customer. And then you realize that you need to hire more people to fulfil the workload required for your service, and it’s hard to consistently hire people of the same skill. So you end up hiring people with varying skills, which means some clients don’t always get great service.
In other words, services don’t scale well and should be productized as soon as possible. Meaning that if you are trying to scale a service, you run into the following problems:
- You will get an uneven quality of service, due to hiring people that will provide an uneven level of quality based on their skills and character
- Staffing talent is difficult because you may need to hire really quickly or fire really quickly (if you lose a lot of revenue where clients do not need the service)
A great real-life example of this is how Parker Conrad learned from his previous company Zenefits, where he ran into a similar problem. With his newer company Rippling, he decided to spend almost two years just engineering the new product to reduce all of the manual work upfront:
“But when asked what he’d learned from Zenefits, Conrad looked past those troubles and instead recalled that “one of the mistakes that we made was that we did a lot stuff manually behind the scenes. When you scale up, there are these manual processes, and it’s really hard to come back later when it’s a big hard complicated thing and replace it with technology. You get upside down on margins. If you start at the beginning and never let the manual processes creep in . . . it sort of works.”
– Parker Conrad
Assessing the risks with business process automation
Assess the consequences for getting an business process automation wrong
Before even starting to think about automation, think about what happens if the automation breaks or does something wrong. Sometimes, it can make automations not worth it right off the bat.
A good way to assess the scale of consequences is to understand if they are reversible. Meaning if something goes wrong, you can reverse the action or not.
For instance, if you send an email to an unhappy client automatically asking to upgrade their service, you can’t unsend that email, which in turn might have made the client more unhappy.
Assess the potential impact of business process automation on other systems
Most businesses build a web of automated processes unintentionally. It’s a gradual process where a business might add one new automated flow per week. And without a dedicated person to organize all of these automated systems together, it’s usually too late where a business unveils the risk it creates.
The main risk is that automations will start to produce unintended consequences.
Beware of trying to reduce automated client-facing tasks and communication
As a rule of thumb, it’s usually best to continue communications with clients manually. The main reason is that clients will be able to tell if you’re doing something automatically (not a human) very easily. Automated templated emails don’t usually work as a result, unless you have a template that you can customize before sending it out.
How to automate your business processes
- Map out the steps you want to automate: write a list of all of the tasks down in a process and see which tasks can be automated
- Choose the appropriate technologies to automate your processes
- If you’re changing data, create a backup: creating a backup ensures that if something goes wrong, you can revert back to how the automation was before
- Build the automations in a test environment if possible: A test environment is a safe place to test your automation before it goes live internally or externally. Which means that you’re not interfering with live data that can affect your business.
- Extensive testing under safe conditions: try to break your automation and use it in different ways. So that when it’s being used in a live environment, the errors in your system are picked up before they are brought over to a lot of people.
- Run unit tests: if your automation applies to a lot of people, or a large part of a process, then test the automation on a smaller set first
- Build automatic kill switches into your automations: sometimes things go very wrong in automations, often when something you didn’t anticipate happens. So that’s why it’s important to develop automatic kill switches which reduce the damage done, by turning off your automation
- Build redundancies into your automation at different stages of your processes: which means that you want to build systems to check to see if the automation is working properly throughout the process. So that if something isn’t working properly, it gets identified early.
When to get the engineers involved
The key to understanding when to get engineers involved mainly has to do with scalability. Often, no-code or low-code automated systems don’t scale well across thousands of internal users. So when you think a process needs to scale rapidly, it’s usually time to get engineering work done to automate processes.
The top tools to automate business processes
Use a programming language like Python to automate business processes
The reason programming languages are good for business processes is that they are flexible and you get more control over your automation. The flexibility comes from being able to interface with other software very easily through packages and libraries. Packages and libraries are collections of code other people wrote that allow you to do things. For instance, if you want to automate in excel and move data to Salesforce, there are libraries to directly interface with excel and Salesforce and you can easily bridge the gap.
Python is an easy-to-learn programming language that can automate common processes. The top resources to get started with Python are:
- Learning python for beginners
- Anaconda (the best way to manage packages in Python)
- Automate the boring stuff (tutorials to automate common things in Python)
Zapier
Zapier is one of the most popular automation tools as it can bring pretty much any two systems together without using code. It relies on a trigger and action model.
Use workflow tools in your customer relationship management software
Customer relationship management software like Salesforce and Hubspot have extensive workflow automation capabilities, meaning they can automate common human tasks.
Conclusion
The truth is that humans and computers are meant to cooperate. And this is true of automations in particular, where both must work together to get the most out of their strengths.
One of the most important conclusions from this guide is that automations can provide a lot of benefits, but also with potential huge drawbacks if not done properly. So be careful, get some small automation wins in, then start to scale.