Machine learning, artificial intelligence, robot workers. These are buzzwords we hear often. Yet we find that mostly, people throughout government are unsure of the actual impact these tools will have.
Over the last decade another phrase has joined the lexicon: Robotic process automation (RPA), promising to bridge the gap between outdated, legacy processes and true artificial intelligence for automated workers.
Recently, as part of a bold transformation strategy, a council we’re working with asked us to research how RPA could be applied to local government organisations. I’d like to share what we’ve learned and things to consider if you’re wondering if this is a good fit for your organisation.
What is RPA good for?
Most demos we’ve seen for RPA are riding the low code platform wave. At the heart of these low code platforms are the idea that your staff can, without any technical knowledge, build sophisticated processes and integrations between legacy systems. RPA takes this concept a step further by providing a method of using existing technology to automate the tasks a human can do when interacting with systems; such as creating new records, saving them and using simple logic to make decisions. It often uses screen scraping and simple logic to automate tasks that are onerous.
The example most tools on the market highlight is invoicing. In this video, the bot goes through a set of predefined processes for drafting an invoice, validating information against an email and submitting. The upside is, this doesn’t require investment in new systems (other than the RPA tool itself). The downside is that you have to be comfortable with the bot completing tasks without intervention. Service managers have to ask themselves:
- Can we design to flag potentially fraudulent transactions? How will knowledgeable humans check it’s working?
- What are the edge cases that the bot can’t deal with? What percentage does this cover?
What’s the catch?
RPA tools promise ‘quick wins’ for service managers: use what you’ve got, replace manual tasks and do more in a day. But after analysis and research, we’ve found limitations. Mainly in the pricing model.
Most RPA tools charge an upfront cost for access, followed by supplementary charges for each bot ‘worker’ used. To justify the cost, it seems to only make sense if your team has a lot of simple tasks that can be completed by a bot (high volumes of cases: over 500 times a day per ‘worker’) without too much human intervention. Without that volume of work, the cost of implementation and supplementary charges per bot to complete work seems high to us.
This isn’t to say that there aren’t opportunities to use this tool across services, but having dedicated ‘workers’ for each admin job gets expensive quickly. As an implementer of RPA, you’ll have to think carefully about how you prioritise a bot’s time. There are also outstanding questions around licensing. Legacy suppliers fight against automation, requiring each bot to have a full user license. This can easily eat into savings targets if you’re not careful or have legacy systems fighting against automation in their business models.
Building the right foundation
If you see an opportunity to bring in RPA into your service, it’s important to start by laying the right foundations for bringing the tool into your work. At FutureGov, we believe in the Power of service design to deliver end to end services that meet user needs and reduce costs.
To successfully bring RPA into your service, you must have clearly defined processes. You need to be comfortable defining an end to end experience for your service. This includes all diverging paths for the people using your service and knowing how to navigate through the options.
You’ll also need to understand how this process works alongside your team. In many complex services, you may be able to automate parts of the journey, but not all. Often, there are points where a human will need to get involved. A person in your team will need to check that the automated actions are appropriate, are not exploited and are safe for the citizen.
What does this mean for you?
For RPA to make sense, you need a lot of transactional, manual processes which you can confidently pass to an automated worker. You need to ensure that the bot can deal with enough work to justify its cost over another human worker, that has the right licenses and that these licenses don’t deplete your savings. And most importantly: you need to lay the right foundations that fulfil the needs of citizens and can support automation. That means well-defined processes, mapped out and well understood, clear boundaries and a clear way of detecting fraudulent claims, such as fake invoices.
Robotic process automation can be a tool to use to deal with high levels of demand or manual processes in the short term. But we don’t believe it will replace the need for well designed, user-centred services. RPA isn’t something we should invest in without first having a plan on building proper integrations between systems and getting rid of outdated tech.