Skip to Content

Level 1 – Easy ; 30 minute read

Audience: Project managers, business analysts, subject matter experts

Author: Mark Muir, SAP BTS, S/4HANA RIG Americas

 

Welcome back!  In this blog I want to be mindful when pairing machine learning and solving unique challenges in your business.  Challenges present themselves in various forms e.g. business problem or latent need in the marketplace each one may have their own criteria for success (a.) complex business process ‘tip of the iceberg’ scenario where unseen circumstances expose a larger degree of effort, skills or (b.) new business prediction requires the right technology e.g. NLP (Natural Language Processing).

Contextualizing our challenges (a +b) let’s walk through each use case and identify useful tips.

Use Case (a)

  • Evaluate your challenge, qualify the business need to justify proceeding with machine learning.
  • Business Requirement: Numerous use cases are well documented by SAP and on the internet, below are use cases you may find helpful.
    • SAP Predictive Analytics: Automated Analytics User Guide and Scenarios – Section: Modeling here
    • Bring Your Own Model (BYOM): Functional Services provides readily consumable pre-trained models that can be used as a web service by calling simple REST APIs; examples only.
      • Document Feature Extraction API is capable of extracting feature vectors for any given document which can be used for comparison, informational retrieval, clustering or further processing.
      • Image Classification API calculates and returns a list of classifications along with their probabilities for a given image
      • Product Image Classification API classifies images into a fixed set of categories of products that are common in eCommerce.
      • Machine Translation API translates multiple translation units from a source language into multiple target languages.
      • More available at the SAP API Hub
    • Train Your Own Model (TYOM): Machine Learning models for Ready to use Services (also known as Functional Services) are developed and maintained by SAP. Examples only
      • SAP Cloud – Cash Application Integration (1MV)
      • SAP Cloud – Integration to SAP Real Spend and SAP Financial Statement Insight (1KU)
      • SAP Cloud – Propose Resolution for Invoice Payment Block (2XX)
      • SAP Leonardo ML Foundation link
      • SAP Cash Application link
      • SAP Resume Matching link
      • More available at SAP/Products
  • The following questions are relevant for ‘a+b’, uniqueness of the problem will trigger additional discovery questions.
    • “Can the business clearly define the problem”, if ‘no’ this is a showstopper, invest time getting it right.
    • “Do you have stakeholder buy in”, if ‘no’ or the response is ‘tepid’ go back to the business case and operations managers to bolster the case for machine learning.
    • “Do you need automation”, if ‘no’ consider a manual work-around.
    • Can you write rules”, if ‘yes’ the problem may well be resolved by customizing standard code or using a rules-based engine.
    • “Can you formulate the problem clearly”, if ‘no’ go back to the business and refine to express a precise form.
    • “Do you have good clean data”, if ‘no’ go back to the business to source data sets that will return good model output.
    • “Do you have a regular pattern”, if ‘no’ go back to the business if needed enhance your data with other data sources.
    • “Do you have meaningful features i.e. does the data have existing labels to help a machine make sense of it?”, if ‘no’ work with the business to enhance your data with other data sources if necessary.
    • “How do you measure success and can you afford for some allowance of error”, the business owner should be involved throughout the entire process, the end result should be a cumulative effort of passing through quality gates before handing over the model to the business.  There should be no surprises, define KPI’s and performance indicators to measure quality and success during the process, they may also change after use in a production environment.

Machine Learning: Types of Scenarios

  • A graphical depiction of applications under the SAP Leonardo Machine Learning Foundation umbrella can be seen in table 1, this will be covered in more detail in Where’s the beef?
  • Machine learning core capabilities and types of scenario can be evaluated in table 2, what is important to highlight here is BYOM and TYOM represent two major areas where SAP innovation and consumption of use cases are at its. You can learn more about SAP innovation by visiting the SAP roadmap/machine learning and Innovation Discovery.
  • More use case examples can be located at com/Products, SAP Best Practices Explorer (key words: machine learning, predictive) and Predictive Analytics suite Predictive Analytics integrator (PAi).

Table 1

 

Core Capabilities

Types of Scenario Service Type Use Case Examples
Inference Ready to use Services · Pre-trained services solving a generic use case e.g. classifying images or detecting topics in documents · Ref: Functional Services
Bring your own Model [BYOM]

· Bring Your Own Model (BYOM) services help user to deploy their own Machine Learning models created using various technologies to store and deploy them in SAP Leonardo ML Foundation. There are 2 main components to manage models and model servers in BYOM scenario.

–        Model Repository API

–        Model Deployment API

· Deploy, publish and run your own ML Model as a service

· Manage your model’s status, monitor resource consumption and validate quality periodically

· Leverage features such as authentication and scalability.

· Deploy your own Machine Learning models.

· Ref: BYOM overview

Training Customize Model

· SAP Leonardo Machine Learning foundation provides readily consumable pre-trained models that can be used as a web service by calling simple REST APIs. Explore the functional services such as image classification, product image classification, topic detection, time series changepoint detection.

· Re-train and tailor image classification services based on your own data

· Simple APIs for retraining ML models – no extensive machine learning knowledge required

Search SAP API Hub for published ML API’s here.

 

· Ref: Retrainable Services

Create Training [TYOM]

· Ready to use Services is a collection of ML services that customers can use it out-of-the-box to enable certain ML capabilities such as text classification, image classification in their applications. Machine Learning models for Ready to use Services (also known as Functional Services) are developed and maintained by SAP.

· Leverage pre-trained ML services via simple Web APIs allowing immediate usage

· Explore a rich set of ML services such as image classification, topic detection or time series changepoint detection as building blocks for creating your own intelligent app

Search SAP Best Practice Explorer for ML scope items here.  Switch between latest S/4HANA versions e.g. *cloud edition and on-premise.

 

Predictive Analytics suite Predictive Analytics integrator (PAi).

 

· Ref: Training Service

 

Table 2

Use Case (b) – Characteristics: Planned, Innovative Technology (leading)

  • SAP CoPilot is a good example how customers can plan their own innovation roadmap with SAP, detailed product roadmaps highlight innovations and release cycles driven by a cloud first strategy (quarterly innovations e.g. 1808, 1811) ; S/4HANA on-premise (annual innovations e.g. 1709, 1809).
  • Business Requirement: The business (center of excellence) wants to use artificial intelligence to automate and simplify how end users create and manage incidents within the software application.

 BIS Bot

  • Digital assistant based on the Natural Language Processing (NLP) using free text and voice.
  • Uses new technologies such as Machine Learning to determine the user intent, translate the user request into an executable call to a service and provide the user with the result of the query or action as provided by the service.
  • Besides incident creation and feature request creation (available), SAP support tools such as Schedule an Expert, Expert Chat etc. will be integrated to the BIS Bot to realize real time support channels.

Flow Chart

The flow describes creating an Incident via the Bot.  Bot collects the information from the context of the conversation and creates an incident in the backend ticketing system.

Bot integration hub: Tasks to create an incident

 

  • Please refer use case (a) for qualification tips, the questions apply for all scenarios only in this example we assume our use case is well thought through.
  • Additional Chatbot qualification questions to consider.
    • How do you measure ROI, what are you looking for in a chatbot implementation (i.e: improve customer satisfaction, automatise answers, reduce costs, etc.)?
    • What are the KPIs to monitor the ROI of a chatbot?
    • What is the volume in number of chats and/or calls you have within customer support?
    • Do you have meaniningful data (high volume) that can be used to train your bot e.g. chat/call logs?

Paths to Transformation

This is a good point to introduce SAP Leonardo – Paths to Transformation a design thinking approach to maximize your machine learning discovery but we will cover a lot more of SAP Leonardo, technologies and accelerators in Where’s the Beef?

·       Optimize initiatives build intelligence into an already existing process and business model i.e. improving productivity and process reliability with embedded intelligence

  • Focus on the problem on hand, least amount of risk, fastest ROI

·       Extend initiatives rethink existing processes to arrive at new outcomes i.e. capture new sources of value to extend the business process e.g. IoT existing processes with industry innovation kits

  • Rethink business process’, arrive at new outcomes, fast ROI

·       Transform introduce new business models into the value chain i.e. reimagine manufacturing at a customer’s site with our open innovation service

  • Most disruptive, higher effort, opportunity for new revenue streams

Paths to Transformation e.g. Optimize

Wrapping up

I intentionally tucked the conversation of Digital Transformation into the end of the read simply because it can be the elephant in the room, machine learning after all is the purpose of the blog.  That said, there can be no denying to develop and benefit from an intelligent enterprise it is prudent to understand SAP’s product strategy and roadmap.

Having your own well-defined strategy helps deliver optimal value concerning a dimensions quality, service, and time.  The value of the process is generated through the implementation of IT-systems and organizational structures in an optimized model of business reality1.

  • Strategy Management 1
    • Impact of transformation drivers
      • What are the transformation drivers and what impact are they going to have on the business model?
    • Vision and Positioning
      • What are the business options for the future and how does our vision for the future business model look like? What is the strategic positioning?
    • Value chain
      • How is the value chain transformed? which parts have to be optimized
    • What is the future competitive advantage?
      • What is the competitive advantage and what are the value potentials that can be realized through the transformation?
    • Organization & ECO system
      • How should look like a suitable (partner) organization to ensure the achievement of strategic goals?
    • Technology
      • Which technological concepts are necessary – what kind of technologies are available?
    • Risk
      • What kind of risks do we have to consider? How can we address them on a strategic level?

Key Take-Away’s

Promising use cases are selected and prioritized based on two dimensions.

  • Technical feasibility e.g. data quality and availability (in a co-innovation scenario with SAP, access to real data is an imperative)
  • Business potential e.g. saved effort or generated revenue

Conclusion

The prevalence of AI in the market has displaced the hype making AI more of an opportunity and less of a threat, in all certainty it has become a significant differentiator to a customer’s future business strategy.

More Information

  • SAP Leonardo Machine Learning road maps and Innovation Discovery
  • SAP Leonardo Design-Led Engagements Basics is an openSAP course, with the help of a case study you will gain an understanding of an SAP Leonardo engagement.
    • Introduction to the Course
    • SAP Leonardo Engagement Tracks
    • Jobs to be Done
    • The Explore phase
    • The Discover phase
    • The Design & Prototype phase
    • The Deliver phase

References

  • 1 SAP Business Transformation Services (BTS) has created a methodology for successful implementation of transformation projects: the Business Transformation Management Methodology (BTM2)

Further reading in this Machine Learning blog series

  1. An enablement guide for managers
  2. Solving unique challenges in your business
  3. Adoption in industry
  4. Where’s the beef?
  5. Introduction to Modelling
  6. Approaching your Project
  7. Machine learning and IoT
  8. Co-Innovation with SAP
To report this post you need to login first.

Be the first to leave a comment

You must be Logged on to comment or reply to a post.

Leave a Reply