Can artificial intelligence solve the ERP complexity conundrum?

  • Posted on May 1, 2018
  • Estimated reading time 5 minutes

This article was originally written by Avanade alumn Ismael Quteifan.

The Fourth Industrial Revolution (or industry 4.0) is the new wave engulfing the business world. Whilst technologies in the space of 3D printing and artificial intelligence are gaining traction, a question hovers over the enterprise resource planning (ERP) domain. Historically, ERP’s had re-invent themselves many times over to stay in the game of advancing technology solutions. This time, the challenge facing ERPs is different: two problems that need to be solved simultaneously. The conundrum is that the solution of one problem works against resolving the other. Could AI provide an "intelligent" answer to this state of affairs?

The need for more
In a changing business environment, competitiveness depends on insight and agility. Here lays the issue, ERP should be used for decision-making and not as transactional engines.

This called for an uplift in ERP functionality to include insights, metrics and KPIs a modern solution needs to cater for. However, by enriching ERPs with these bells and whistles, we increased their data content, and hence made reaching the end goal overwhelming. There should be a way not only to crunch data, but also to provide intelligent insights into trends and relationships that simplify this process for the busy executives. Especially when most configurations allow for one or two tiers of routing logic to be configured.

This is where artificial intelligence (AI) comes into play. AI can help to provide insights and ultimately improve the user experience. The best example of this application is the upsell (buy better) and cross-sell (buy extra) in the retail industry. There another two applications for AI in the service and maintenance industry: predictive maintenance and condition monitoring. This is a funny example though, as condition monitoring can be categorised under insight, whereas predictive maintenance is some sort of a decision-making engine.

Lots of ERPs started providing these features. Microsoft Dynamics D365 for Field Service (D365FS) uses Azure AI and cognitive services in predictive maintenance. This has been applied to two prominent case studies in both Rolls Royce and Michelin. Additionally, some ISVs for Dynamics 365 for Finance and Operations (D365FO) in the Enterprise Asset Management (EAM) space links to Azure Machine Learning (ML) for the same purpose. Applied to inventory, demand forecasting models in D365FO can also dip into ML to suggest inventory levels.

The examples above are what can be referred to as out-of-the-box Cognitive Services. More AI applications can be developed through custom learning algorithms and complex ML models. The connectors to these services are available, but an AI engineer is needed for these developments.

AI can also help in decision-making by suggesting next best action. If you want to email a contact to follow up on an opportunity in Dynamics 365 for Sales (D365FS), or defaulting financial dimensions based on multi-tier rules (D365FO). The possibilities are endless: transport route optimisation, inventory picking routes or alternate production schedules when a machine breaks down.

Less is more
The other problem is the growing requirement for end users to have fewer manual, repetitive and mundane processes. In that regard, traditional ERP design works against itself. To be everything for everyone, an ERP is like a blank canvas that is highly configurable. This makes its tasks, by nature, incremental and hence more laborious. It is pretty much like having a space-shuttle dashboard of dials and switches when what you're trying to do is to process 100 vendor invoices before month end. Productivity tools are needed more than ever now, and with demand on ERPs to become more granular in their functionality, the task is even harder.

RPA (Robotic Process Automation) is a form of AI where virtual bots can log in as a normal user and automate certain processes, or specific steps in these processes, to boost productivity. This can be beneficial in high-volume repetitive services like invoice processing, payment journals, credit and collection, PO approval, split order picking between warehouses based on available inventory or firm planned orders according to specific criteria. Productivity does not have to be hard-coded in ERP.

But RPA tools are not limited to the mundane activities. In past years, tasks involving free text, handwritten documents, or audio/video inputs were not good process automation candidates. Nowadays, businesses can automate processes that were off limits before by combining base RPA tools with cognitive services. It is also possible for human and bots work together (aka "assisted automation") where human input is only required for certain steps in a process.

Many RPA frameworks that can be leveraged to achieve that. Take Blue Prism for example, the frameworks are generally easy to use and do not require significant technical expertise to configure. After spending enough time to "teach the bot;" the process can flow with little ongoing effort. This is another exciting feature because it empowers the users by reducing dependency on external or specialised resources to achieve automation.

ERPs are becoming more complex, and to keep up with the times, they also need to be more intuitive. Gone are the days where hard-coding intelligence is the only way to streamline the flow of business processes. Machine learning and RPA are two AI tools that, when combined, can solve the conundrum of achieving more with doing less. We are now watching these three acronyms combined to deliver the boost required in the new era.

Learn about Avanade's expertise and approach to artificial intelligence

Kent Johnson

I love it.  I see where this is headed and I have no doubt this will be the direction we go in the near future (who knows what long term will look like).  My one concern is that a number of organizations may have trouble digesting this quantum leap and so will need very astute change management as an integral part of the implementation.  I also agree that businesses need a nimbleness that hard coding business rules simply cannot deliver - things are changing too fast and the number of business rules keeps growing too quickly.  We have this issue at a client in Minneapolis who is trying to hard code everything in their universe.  It is not working.

May 16, 2018

Lee Woodward

Hey, Ismael.  Thank you for these keen insights.  Recently I had the pleasure of participating in Avanade's Southeast Innovation Summit in Atlanta.  The team that I worked with came up with a brief case study dealing with precisely this topic.  Our scenario involved the processing of customer product returns, traditionally a cumbersome process.  By introducing AI at certain pain points in that process, we proposed to streamline the returns management work stream and, more importantly I believe, to derive information that would drive continuous improvement in product design, manufacturing, marketing, and user education.  Reducing the repetitive workload and driving new strategic insights will free up people to do what they do best, innovate.

May 15, 2018

Avanade Insights Newsletter

Stay up to date with our latest news.

Contact Avanade

Next steps

Talk to us about how we can bring the power of digital innovation to your business.

Share this page