Massive thank you to Rob King of Wzard Innovation for reviewing another of our events!
RPA AND AI MEETUP: ITS ALL ABOUT SCALE
December 16, 2018
Last month was the RPA and AI Meetup in London. Hosted by Bloom Search this was the fifth of their quarterly events. Chaired on this occasion by Danny Major from Thoughtonomy, the panel of experts consisted of:
Alec Sutherland, Automation Technical Lead, John Lewis
Emma Kirby-Kidd, Process Automation Lead for Ageas UK
Graham Lee, Group Head of RPA, Thomas Cook
Delivery models (waterfall or agile) don't work for RPA. John Lewis were moving towards a more agile approach, but we didn't find the sprint a good fit. We normally experience 6 week cycles of development in a sprint, but the small quick deliveries didn't work.
It's not all about robotics! It's important to clean up the processes first, otherwise its rubbish in, rubbish out. A strong process ethos is essential. It takes time to build knowledge, so don't be afraid to make mistakes, people don't come in with the skill so there needs to be time available to learn. This also requires a high level of engagement and empowerment with everyone involved in the programme. Although it may sound strange, we humanise robotics.
RPA won't fix everything. Smaller process are better so you have the opportunity to lean, fail fast and recover quickly. Agile teams who are pulling the work themselves works well. It is important to share your knowledge.
I started with just myself and 1 developer, which eventually grew to 2 developers. We then added an After Care team who looked after scheduling, maximising the utilisation of the robots and managing the continuous improvement of robots in production. We operated for 4 years with this model, and see ourselves as the Robotics Team, not a CoE.
Operational areas achieved the greatest benefits. We use Blue Prism, but found their support limited so used online information and forums to find answers to the problems we faced.
When starting out there was just myself and a BA. There was a lack of IT and infrastructure engagement, and no funding from IT.
We use the NICE platform, and now have an RPA Platform Team rather than a CoE. NICE provide a high level of support including professional services and Proof of Concepts free of charge. We target both front office and back office processes.
When selecting processes identify the benefits up front so that you don't start chasing the wrong priorities. It can still be a challenge, for example an 8 FTE benefit suddenly falls to 4 FTE benefit following implementation, but it's up to the business to realise the benefits themselves.
Next quarter our budget is £200K, and we expect to deliver £600K in benefits.
We started with centralised process improvement, analyst and delivery lead roles.
We are now moving to a federated model. Within the business there will be automation specialists who are responsible for feeding the pipeline with new processes for automation and provide on-the-ground support to local teams.
Developers remain centrally controlled. We are targeting back office and avoiding any front office or processes involving customer data at the moment. Next year we anticipate this will change as we move into the customer contact centre.
We have worked with many back office processes but are also working front office processes. For example we have consultants navigating 11 different screens to process a cancellation, it was an awful customer journey and an awful consultant journey!
Part of my role is to sell RPA, getting the company buy-in and starting people on their RPA journey. I always ask them to tell me their problem first. I have 150 requests for automation on my list.
We recently completed a retrospective and it was clear that environments and data cause the greatest delays (waiting time) when getting ready for for testing and regression testing.
I'm starting again in new functions, building up the skill set in the new areas and educating them in RPA. Inexperienced people choose their most complex processes and stakeholders are looking for 100% perfect coverage. Whereas we work on the 80/20 rule, our After Care team take responsibility for continued work on the remaining 20% of work to see if a greater number of transactions can be handled by the robot.
Only humans make errors, the robots will repeat the process correctly every time.
We currently have 30 robots, and occasionally our core platform will fail causing problems for automation.
I face the same problems and issues as the other panelists. The key to this is don't tell me what to automate, rather tell me the problem.
We have a static core team. The business analyst role is critical, they are not just a process analyst, but someone who understands the business. The product owner or sponsor works side-by-side with the team to deliver the automation opportunities. We don't focus on the whole end-to-end journey, we focus just on what can be done.
There is always rumours of job losses, but this hasn't been our experience.
My background is in the military, and it takes 6 weeks to train a soldier, yet it takes 16 weeks to train a customer representative to book a holiday. Someone needs to revisit those processes!!!
RPA will move into API's, with capabilities more similar to the API console. Bolt-ons for machine learning will be available. This will make the automations more manageable.
There are no standards for RPA and I don't see this changing.
RPA capabilities have advanced significantly and will continue to evolve to meet the business demands.
Graham, Emma and Alec all recounted the same experiences, that their respective RPA programmes had not resulted in any direct job losses from the business. This did not extend to 3rd parties to whom the activities may have been previously outsourced, or natural attrition which may reduce total headcount.
While benefits are measured in terms of FTE we use the term Hours Back to the Business to make departments responsible themselves for the realisation of benefits.
We have 3 types of automation: (1) Lifetime Automation; (2) Band Aid (Temporary Fix) automation; and (3) Incident management.
I have a paid After Care Team. We have robust testing and deliver to a 6-week release cycle. Changes are moved into UAT and added to the release if successful. The costs are low, they are just the cost of the team.
We are never down for more than 2 hours and constantly monitor the release cycle.
We started with a pilot, and its still a pilot 3 years later. We prove operational (business) acceptance as we're replicating what the person does (and we are already accepting that risk).
Systems and Architecture need change processes, but this risk is different. Robots won't make a mistake.
Service implementation teams now have confidence in our process, we still keep them informed but the meetings are short.
Start Small, Think Big, Scale Fast
Show them the bang for the buck, once they see the money they will buy in
Sponsorship is key: winning hearts and minds, engage often
Another fantastic event, held in a wonderful location with a panel of interesting speakers who provided a good range of experience and approaches to make every answer insightful. Thank you to all the panellists for sharing their knowledge and thoughts.
If you are looking to start or scale your RPA and AI journey then please contact us at Wzard Innovation to learn more about working with a partner can accelerate the pace of your journey.
Rob is co-founder of Wzard Innovation and author of Digital Workforce, an executive guide to the introduction of Robotic Process Automation.