Category Archives: Solution Architecture

Artificial Intelligence is transforming ERP solutions

binary-1536623_1920

If you don’t innovate fast, disrupt your industry, disrupt yourself, you will be left behind – John Chambers

Without a doubt, artificial intelligence (AI) will have a profound impact on the footprint of ERP solutions in the foreseeable future. AI will enable organizations to further optimize their operating model made up of business processes, software applications, governance structures and technology infrastructure.

To orchestrate this transformation, organizations must revamp their IT strategies and roadmaps and ingest the value of AI and ERP integration. These technologies go hand in hand because they cover the same spectrum.

AI enabled ERP solutions will by default impact the heart and soul of day-to-day operations. The mix of people, process and technology is going to change. AI solutions will take over routine tasks  in the end-to-end business process model that are currently performed by humans. This particular change is driven by an ongoing need to reduce operational cost. It is an irreversible process: “You either disrupt or get disrupted.”

At the same time AI can augment people’s capabilities and the effectiveness of the organization as a whole. This can be realized by shifting the focus to non-routine, analytical and creative tasks. Such a shift can only happen when AI and ERP are being addressed at the same time.

With the evolution of ERP solutions over the last 2 decades, organizations have gained access to a wealth of structured data. Nonetheless, they continue to struggle with transforming data in meaningful information, decisions and actions. The last 5 years, this situation got further challenged with the explosive growth of unstructured data that organizations capture without a clear approach on how to properly use it.

The ingredients to solve this situation are available and becoming more and more mature every day:

  • ERP solutions that enable organizations to run cost-efficient and effective operations
  • Big data solutions that can manage structured and unstructured data
  • Business analytics solutions that can provide information with a user-friendly experience
  • Cloud infrastructure that can make enterprise solutions widely accessible
  • Computing power that allow enterprise solutions to manage large data sets and complex algorithms
  • In-memory database technologies to explore large data sets in real-time
  • AI solutions that can learn, speak, read, respond, predict and execute transactions
  • Internet of Things (IoT) technology to capture real-time performance information

Customer service is an example of a functional area where AI technology is emerging. Very recently, KLM deployed an AI solution that has  the intent to improve the response to customer inquiries received through social media channels.

The deep learning and natural language processing solution for KLM was developed by DigitalGenius. It masters more than 60,000 customer questions and answers. That number keeps growing, because the ‘smart solution’ continues to learn and improve overtime. When a customer service agent receives a question, the solution gives a proposed answer. It’s up to the discretion of the service agent to follow up on the advice or not. The human interaction with the customer is still there, because customers still find that important. KLM is the first airline that deployed a customer service AI solution in the industry.

An AI enabled ERP solution for customer service integrates the customer interaction with the work order management process. The AI solution understands and learns from historical inspection reports and work orders. Depending on the nature of the customer inquiry it gives a proposed answer to the service agent. The AI solution assists with the planning and scheduling of the work by finding the earliest possible date to dispatch a service technician. That’s possible because it understands the required skill set and availability of required service parts.

This scenario is relevant for example for Cities. They render multiple services to commercial and residential customers. A customer may have more than one service issue at any point in time. An AI enabled ERP solution would in this case have the ability to provide insight into the status of all services by accessing and interpreting data from many systems. There may be many work orders in different stages of completion managed by different operational units. The AI enabled ERP solution would assist the agent with adequate communication to the customer, and effective coordination of the work with the departments.

Maintenance is another functional area where AI will be integrated with ERP solutions. A digital assistant  (DA) can help the service technician with the root cause analysis for corrective maintenance issues. The DA has deep understanding of the technical structure, performance and maintenance history of the troubled equipment. It also knows how  the equipment performs compared to similar units at other sites. The service technician is asking questions to the DA and gets evidence-based recommendations back. The DA obtained knowledge from the core ERP system and sources from the OEM. Inspection reports and work orders are important process documents to maintain for that purpose.

AI solutions are starting to appear in the area of predictive maintenance, which is different than preventative maintenance. The latter is triggered by time, events, or meter readings and results in planned, scheduled work. Predictive maintenance is much more based on real-time information about the actual performance of the equipment. Oftentimes, sensors and other Internet of Things (IoT) technologies play a critical role in capturing that information and relaying it back to the AI enabled ERP solution. Predictive maintenance has the objective to reduce maintenance cost. Where preventative maintenance indicates that a part has to be replaced, predictive maintenance may recommend to replace it later based on the actual condition.

SAP is gradually moving into the AI space. SAP has a cloud-based predictive maintenance solution that is based on prediction models and machine learning algorithms. The service technician has access to dashboards with key performance-indicators and ad-hoc reporting capabilities. It helps the service technician to understand the real-time performance of the technical infrastructure and intervene when required.

Earlier this month, SAP announced ‘SAP CoPilot, the chatbot for enterprise users. SAP CoPilot is based on natural language processing and machine learning technologies. The main purpose of SAP CoPilot is to explore data and business situations, and assist users with evidence-based decision making with the help of a DA. SAP CoPilot will gradually learn from empirical data, which means that its recommendations become more reliable overtime.

The potential of AI enabled ERP solutions is immense. There are a number of steps to take before you can embark on a project implementation. The formulation of a digital strategy is one of them. I will elaborate on these steps in the next article.

Bas de Baat

Functional areas where machine learning is applied first

eye-491625

“The future of business process improvement is on making them intelligent. Machine learning is the driving force”

Machine learning is on a steep adoption curve and making its inroads in our daily lives and work. The application of the technology won’t be an issue at all. There’s an abundance of meaningful value propositions for many functional areas, business processes and roles across multiple industries.

Software vendors of enterprise business solutions are focusing their product development on machine learning and other related artificial intelligence technologies. CEO Bill McDermott of SAP said that intelligent applications will fundamentally change the way you do work in the enterprise in the next decade. He mentioned that we need the system to tell us what to do.

In many publications about machine learning we read about IBM Watson beating humans at Jeopardy, or Google’s AlphaGo beating a Go world champion. There are predictions, for example from Nick Bostrom,  that indicate that singularity, or the moment in time when a computer will be as intelligent as a human, is going to happen around 2040.

We also know that machine learning solutions work with structured and unstructured data. Data is growing at a fast pace and doubles every 2 years. 80% of the data is unstructured and 20% is structured. The technological advancements of the last 5 – 10 years removed barriers that artificial intelligence has been struggling with for many decades. Computing power being the most important one. As an indication, IBM Watson can read 200 million pages in 3 seconds and understand the content.

Machine learning solutions are coming our way. A fundamental principle is that they predict based on past behaviour.

Think about weather predictions. IBM acquired the digital assets of The Weather Company in 2016 and is leveraging IBM Watson platform to provide meaningful services to businesses and consumers. The weather notifications you can receive on your smart phone are coming from IBM Watson, a machine learning solution.

Machine learning solutions aren’t always perfect and there are ways to go. Facebook is using machine learning to recognize and understand photos, video and audio that are posted by its users. Content that is not meeting Facebook’s standards is removed. An example of this is the recent removal of a 1973 world press winning photo of the Vietnam war. After a public outcry Facebook reversed its decision.

A few weeks ago I had a personal experience with Facebook’s machine learning solutions and how it influenced a post. I was using an iPhone app to splice a few videos and thought it would be great to add audio to it. I selected the tune from the app library. When I posted the video, Facebook rejected it, because it believed that I did not have the rights to use the audio. Facebook’s interpretation was wrong, because the tune was general available to the app users. It’s an indication that machine learning solutions can learn on their own, but need human intervention to train them.

Machine learning is going to change the way we design and optimize business processes and functional roles. Automation will shift the role of humans more to exception based interactions and real-time, evidence-based decision making. The level of people, process, technology and data integration will further increase. Standardization of end-to-end processes in the supply chain will further manifest.

My expectation is that the software vendors of enterprise business solutions like SAP, Oracle and Salesforce will put their focus on functional areas where there is a high volume of routine transactions first. Think about the order-to-cash process where recurring orders flow through the order entry, fulfillment and delivery processes with limited human intervention. Think about the customer service process where the scheduling of work orders is further automated. Through internet of things technologies, real-time data becomes available that enable machine learning solutions to schedule service orders at the right time with the right spare parts ordered and skilled technician assigned.

There will be niche solutions too where machine learning is augmenting the human capability in a specific area. Think about travel and expense management solutions, where the processing of entries is for the greater part done without human intervention. Another example is recruitment. Machine learning solutions will take over the steps of identifying and screening candidates. Recruiters will receive a short list of qualified candidates and can focus more on the softer aspects that do require human interaction, for example determining if the candidate is a good fit with the organization.

Interaction with customers through call centres and other channels like email, apps or internet is another example. Machine learning solutions can understand text and speech and process simple transactions. Audible from Amazon does that to process refunds for audio books that the customer does not like.

Machine learning and other artificial intelligence solutions are at the top of Gartner’s Hype Cycle for Emerging Technologies, 2016. The evolution of the technology in the next decades will be fascinating, because it is coming so close to our existence as human beings. The potential to apply it in a meaningful way to our live and work will be enormous.

Bas de Baat

How cognitive computing will change project management

26984515732_4cf0d08643_z

You aren’t going to stop it. The trend is going to keep moving – Ginni Rometty

Cognitive computing is rapidly making inroads into the professional services workspace. The emerging technology will have a lasting impact on all jobs across all industries, with health care, finance and legal as early adopters. As much as it changes current jobs, it will also create new jobs.

For many years, IBM is the leader in the field of cognitive computing with IBM Watson.  According to IBM, the purpose of cognitive computing  is to enhance and scale human expertise, rather than an attempt to replicate human intelligence. IBM prefers to call it augmented intelligence (AI) instead of artificial intelligence.

The underlying thought is that cognitive computing functions in an assisting, sub-ordinate relationship to humans. This is an interesting point of view and positioning of the technology, because many experts believe that cognitive computing has the potential to advance in a superior relationship to humans.

There are a number of technologies that are related to cognitive computing like machine learning, text to speech recognition, natural language processing, image detection, sentiment analysis, and others. All of these technologies have the intention to improve human productivity and decision making.

Cognitive computing will have a material impact on the way we manage technology-driven-change projects. It is a fantastic opportunity to bring the role of the project manager to the next level. The emerging technology shall operate as an assistant and expert in many PM disciplines. It will change the execution of tasks and shift the focus of the PM to more creative and analytical activities. It will provide better information to make decisions.

Here are 5 examples of where I think cognitive computing will have a material impact on project management in the future:

Methods, Tools and Best Practices – The AI assistant is knowledgeable of all the relevant methods, tools and best practices for the project, because it can read and understand speech. The PM can ask specific questions and gets accurate feedback from the AI assistant in real-time. The information can be used for any PM task. As the project progresses and the AI assistant learns about the project deliverables, it can give recommendations to the PM, based on what could be versus what’s actually being delivered. It’s basically a quality check on deliverables that helps the PM to manage expectations. At the conclusion of the project, the AI assistant conducts lessons-learned sessions with the project team and updates the knowledgebase for use in other engagements

Scope management – The challenge with managing scope is not only the change management aspect. It is also the verification of the scope that is being delivered. Something we are not necessarily good at once we are getting close to the finish line. The AI assistant is capable of understanding the planned scope, based on the statement and detailed definitions in design documents. With that knowledge it can verify the scope based on data from status reports and test systems. The AI assistant can make a recommendation to the PM where the project is at risk close to a go live

Time management – Project scheduling can be a daunting task, because of its complexity. The AI assistant can not only provide a baseline schedule that the PM can adjust and refine, it can also make predications based on historical and empirical data. This improves the productivity of the PM and the entire project team. The AI assistant can plan and forecast the required resources based on an estimation model that it maintains with data from the project itself and other projects. The AI assistant can determine if the project is on track and if there are tasks at risk that are on the critical path. A prerequisite to many of the functions that the AI assistant can provide is the access to data. For example, project team members must record time at the task and deliverable level

Cost management – Based on the scope definition, baseline schedule, resource plan, approach and risk tolerance, the AI assistant can calculate a cost baseline that the PM can adjust and refine. As the project progresses, the AI assistant can make an ETC and EAC forecasts based on earned value parameters that the PM has set. Based on the approach the AI assistant can calculate the cost impact of alternative delivery scenarios. As example, it can determine the cost and schedule impact of using more off-shore resources

Organizational Change Management – This is an area where the PM can provide more value with the arrival of the AI assistant. When a majority of the routine tasks have been delegated to the AI assistant, the PM can apply his creative and social skills on driving organizational change. The AI assistant and PM can work collaboratively in this field. As example, the AI assistant can provide a baseline of questions to conduct change impact assessments and training needs analysis. Based on the analytical outcome of the response, the PM can optimize the change management plan and properly engage with the key stakeholders. Furthermore, it can determine what course are required to train the project team and end-users. Another example is stakeholder management. Based on text analysis, the AI assistant is capable of understanding the key characteristics of the main stakeholders and provide recommendations on how to best engage with them. The analysis is also benefiting the PM in aligning and committing the stakeholders to the project goals

The evolution of augmented intelligence or cognitive computing in the professional services workspace is fascinating and should be welcomed with open arms. I strongly believe that an AI assistant can further strengthen the role of the PM and increase the value of services to the client. The majority of the examples that I have used have yet to be developed as applications for practical use. The technology is there. It is a matter of when, not if.

Bas de Baat

3 steps to make project portfolio management a business process

mark-516277_1920

If you can’t describe what you are doing as a process, you don’t know what you are doing – W.E. Deming

Organizations have a hard time to plan and execute the right initiatives, because project portfolio management (PPM) is not deployed as a business process.

With an increasing need and desire to innovate and change the ways we are doing things, one would expect that organizations are keen on project portfolio management. In an ideal state, PPM is managed as a business process, equivalent to the more traditional process like finance, marketing and sales, procurement and human resources management. Yet we are not doing that, or at best we are trying, but agree that there is room for improvement.

The PPM process should be cross-functional of nature and serve internal customers from all business areas. To make that happen it should reside in a business function that by default is set up as such. A Project Management Office or Information Technology Department are then quickly becoming the logical candidates.

What are the steps to deploy PPM a business process?

Align Leadership

Ideas, become successful when we all buy into it, make it happen, and live up to it once it is in operation. Implementing PPM as a business process is a game changer and requires adequate change leadership. Part of that is executive alignment. A key element of the alignment process is visualization. Senior leaders must be able to envision what the future-state looks and how that improves their business area and the organization as a whole. An introduction to the high level process design, a demonstration of the PPM application, and a walkthrough of a few use cases, are instruments to get them all on the same page. Once the alignment is there, a change leadership committee should be established, tasked with delivering the PPM solution.

Implement and Deploy

PPM is an enterprise application, which means that the implementation and deployment must be managed as such. The project team is a balanced representation of the organization with functional and technical resources. If these two principles are violated, the probability that the end-users do not adopt the PPM solution as intended, is high. The focus of the implementation must be on business process, analytics, application and governance. These four components make up the integrated PPM solution, and all need to come into play at the same time. Examples of PPM applications are: Innotas, Workfront, Clarizen, ChangePoint, and others

PPM projects tend to fail when the focus is primarily on the application. Organizations rush through the software product capabilities, make design decisions on-the-go and forget the importance of the business process, governance and analytical requirements. Mobilize a team with internal and external resources. It is imperative that the vendor can provide the expertise in all the four areas of the PPM solution, and can assist the change leadership committee with manifesting the future-state

Execute, learn and adjust

When the PPM solution goes live, it’s the start of a new beginning. The primary focus of the project team and business must be on user adoption and tying the experience back to the original business case. It is a good idea to have super user representation in all of the business areas. The super user is a functional expert in the PPM solution and an evangelist pur sang. It is the first line of support for all the end-users. The PPM business process has a natural cadence where at set times and gates, certain activities must be completed. It is not uncommon that this is a one year cycle. As a consequence, the learn and adjust cycle is at least equal to that period. The organization must go through all the hoops and loops, complete lessons-learned sessions and optimization steps, before the project can be declared a success and closed.

Project portfolio management (PPM) must be perceived as a business critical process for organizations who have the intention to grow, accelerate and improve. Those organizations who want to be an outlier and exception in their marketplace, out serious effort in implementing and deploying a robust PPM solution. It is part of innovation and getting better than your competition.

Bas de Baat

Is the Price Right? “No surprise” cost estimates

Estimation

“Well, maybe up-teen zillion was too general a cost estimate” – Chris Wildt

Besides defining the project scope and attracting qualified people, estimating project cost can be quite a complex task. What are some of the measures that you can implement to be more certain about the accuracy of cost estimates such that little to no surprises come on your way?

The first measure is to make sure that there is a well-documented specification of the deliverable, whatever that deliverable maybe. In order to get there you must have a breakdown of the product or service that the project will deliver. And because that breakdown is subjective to change, you have to maintain it. At times that can result in change requests from the vendor. Document and document. Although it may not be an attractive tasks to do, overtime it certainly pays off (think for example of business process management and solution sustainment).

The second measure is about process. You need to define the delivery process as well as the scope, cost estimation and configuration management processes. The delivery process requires involvement from the subject matter experts, as they are the most knowledgeable and experienced of how a deliverable will be created. The project leader acts in this situation as a facilitator and involves all key players, from designers to developers to testers up to staff who sustain the solution. They all must come to a mutually agreed to delivery process, that becomes the standard. Periodically review the delivery process and make improvements. When the input is right (specification) and the delivery process is right, the output will meet your expectations. The scope management process is critical for a number of reasons. It tells the sponsor and other business stakeholders what the planned project outcome is in a ‘tangible manner’. Furthermore it forms the baseline of project plan, budget, and the contractual agreement with the vendor. The cost estimation process must be standardized such that all parties understand their roles and responsibilities, and what is required to come to an agreeable price. Full transparency of how the team calculated the estimate is crucial (for simplicity, I am making the assumption that cost and price of a deliverable are the same). And lastly, the configuration management process is important as it sets you up for long term success. If you keep track of the cost associated with the build up or configuration of the product or service, it becomes a reliable benchmark for future requirements. Let’s say for example that a dashboard with operational key performance indictors is a deliverable. Keeping track of the estimated and actual cost, helps you in the future when there is a similar business need. It all sounds very straightforward, but there aren’t many organizations who do it.

The third measure is to include calculation models for cost estimates of deliverables in the contract with the vendor or in an agreed to project standard. Here is a good example: vendors (system integrators) who design, build, test and implement ERP solutions, use standardized models that calculate the level of effort for certain development objects like workflow, reports, interfaces, conversion programs, enhancement programs and forms. Depending on the level of complexity the estimator says that for a medium complex report, it takes 3 days to specify, 5 days to develop, and 2 days to test, so 10 days of effort in total. It is quite common that the client only hears 10 days of effort and a certain price from the vendor, instead of the full breakdown. Depending on the negotiated, overall price of the project, change requests can become the ‘bread and butter’ for the vendor. Especially in saturated, highly competitive markets, where vendors may come in with a relatively low price and try to ‘recover’ along the way. Depending on the type of contract (see a previous post about fixed price), the status of the project and other factors, vendors respond differently to change request, effort estimation and pricing and at times present an unreasonable cost estimate in the eyes of the client. If it is not managed well, it can lead to major conflicts. Why would we let that happen? Why would we not be fully transparent about calculation models instead? When two parties embark on a journey to implement a technology that in many cases is the enabler to transform the business, it is wise to spend energy on those value driven activities instead of debates, conflicts or disputes about project cost. At the start of the project, get concurrence upon a standard that works for both and that, together with the other measures discussed above, will set the project up for success. Consider using industry benchmarks to validate the accuracy of the calculation model.

The fourth measure is about effort contingency. Oftentimes, the initial cost estimate of a new or changed deliverable is nothing more than an order of magnitude, SWAG or guesstimate. The vendor has to go with the client through the specification process first to understand the detailed requirements, before the estimate can be firmed up. Make sure that you incorporate a mark up for effort contingency in the cost estimate (some extra days). This kind of contingency differs from cost contingency. That’s more of a percentage based on a number of conditions and added as mark up on the total project cost. It’s more a safety net for unforeseen and in principle be ‘untouched’.

So in closing, to better manage surprises on the project cost side, strive for full transparency. Meaning: clarity in defining the business requirements, standardization of the delivery process, scope management process, cost estimation processes and calculation models, and finally using an adequate mark-up for effort contingency in the cost estimate.

Bas de Baat

Program Manager Enterprise Applications, PMP© | Solution Architect