80% of organisations fail to progress RPA beyond pilot. Here’s why, and how, to avoid becoming a statistic

Tim Olsen, Intelligent Automation Director at Hays Technology

Getting started with Robotic Process Automation (RPA) isn’t rocket science really. Buy a few licences, find some low hanging fruit, and get a developer to deliver them. Maybe you’re just starting out on this journey, or maybe you’re someway down the line, but if you don’t have the benefit of hindsight yet, here’s a tip:

It’s about to get a whole lot harder.

I spoke to senior people within the RPA vendors a year or so back, and those who were candid admitted that most organisations failed to scale beyond a handful of processes.

Before I go into further details, I have to caveat this – there is no one journey that everyone passes through with RPA. I accept that I have a broad brush, but if I were to look at CoE growth across multiple industries, and regress the morale/energy/performance vs time, I am confident we would see a clear pattern emerge.

Start up

In the first stage, an innovator, probably someone who is tech savvy or who has seen RPA before, inevitably sets up RPA to improve the productivity, or reduce the drudgery in their department. They already have a list of key processes which will be high return, and chances are no-one is scrutinising the ROI too closely anyway at this stage. The outcome is impressive, the objectives are realised, and the business’s interest is piqued. Morale is high.

Uncoordinated expansion

Someone, probably Director level, sees the benefits and wants them scaled across the organisation, and directs the RPA team to replicate the activity more broadly. This is when the problems start. If there is a lack of clear communications and a senior level objective stated, there will be immediate resistance. Employees will fear for their jobs, managers will fear a loss of control and status. As the number of processes automated grow, the quick wins will be exhausted, and it will become harder to find positive ROIs.

InfoSec will start to take an interest in proceedings and inevitably bog things down. HR and unions may become involved. Bots may need to meet regulatory standards. What’s more, as the number of bots grows into the 20s or more, developers may be spending almost as much time maintaining the bots as developing new ones. Infrastructure may be unwieldy and become a bottleneck.

At this point, many organisations will decide RPA is just too hard and throw the towel in.

The good news is that the majority of the blockers can be overcome.


The centre of excellence

The organisations that learn and adapt will position the CoE (we’ll come back to that) so that it spans the organisation rather than being siloed. It will receive clear and unequivocal backing from the CEO and an objective will be communicated and embedded into the culture. The platform will ideally be cloud based, or at least not tin-dependent. Specialist skills will need to be brought in and/or developed, and a clear RACI so that the team and stakeholders are clear about their accountabilities.

These, along with a defined framework and process, defines the Centre of Excellence. The CoE ensures quality standards, and economies of scale and is fundamental to the success of the programme (note, we are now talking about an ongoing activity, not a discrete one). The maintenance of the bots can’t be left to luck and a scalable solution with monitoring, alerting and continuous improvement will need to be embedded into BAU.

With a bit of luck and a lot of hard graft, the programme will now start its upward trendline on the change curve.


The key factor influencing acceleration will be the engagement of stakeholders at all levels. A mature CoE will no longer need to hunt for opportunities; they will be brought to them by the workforce once trust is established and benefits are recognised. In larger, more diverse organisations it may be opportunistic to federate the CoE with small satellite teams located in departments, adhering to the CoE standards.

As the number of processes grow, the CoE will target increasingly complex scenarios and bring in more technical solutions, such as Chatbots, NLP and Machine Learning to enable end to end automation.

The most common automation journey is shown below with the change curve overlaid.

For full disclosure at this point I have to credit Mathew Southan who was the first I have seen to overlay the change curve on the common experience – my independent research has bought me to the same conclusions, albeit a little later.

The question for those initiating automation for the first time is: can the curve be flattened, can we avoid the trough of disillusionment and just get steady improvement? Well, yes, to some extent. By recognising the need early to apply structure and scalability and having the foresight and discipline to deploy it, the trough can be avoided. It is inevitable that some lessons will be learned the hard way, but firm foundations will give the programme strength as it grows.

Get more insights into how automation is shaping the world of work here.



Tim Olsen
Intelligent Automation Director, Hays Technology

Tim worked in digital transformation for 20 years developing solutions to improve user journeys and experience for blue chip clients. More recently he grew the UK’s largest RPA CoE and went on to specialise in helping organisations overcome their barriers to scaling automation. He is a thought leader and evangelist for Intelligent Automation, and leads the IA Consulting specialism for Hays.