Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Cognitive Enterprise Services

1,503 views

Published on

Last week, I visited University of New South Wales, in Australia, and gave a talk providing an overview on research work over the past couple of years to establish a framework for Cognitive Enterprise services, specifically on how the lifecycle of enterprise services from sales to delivery and operation is transformed in the cognitive era. I provided an overview of the research assets we have produced, some of which is in industry or the company in various ways.

Published in: Data & Analytics
  • Be the first to comment

  • Be the first to like this

Cognitive Enterprise Services

  1. 1. © 2015 IBM Corporation Hamid R. Motahari-Nezhad IBM Almaden Research Center San Jose, CA The Journey to Cognitive Enterprise IT Services: A Framework for Cognitive Services and Business Processes Talk at University of New South Wales, Sydney,Australia. Nov. 29, 2016
  2. 2. © 2013 IBM Corporation Major Technology Trends Impacting Enterprise Business 2 Mobile Social Cloud Internet of Things 20162000
  3. 3. © 2013 IBM Corporation We are here 44 zettabytes unstructured data 2010 2020 structured data Data is the world’s new natural resource! (Ginni Rometti, IBM Shareholders Report, 2014) We are here Sensors & Devices VoIP Enterprise Data Social Media 5
  4. 4. © 2013 IBM Corporation Mega Trends: Data, Cloud, and Mobile 4 80% of the world’s data today is unstructured 90% of the world’s data was created in the last two years 1 Trillion connected devices generate 2.5 quintillion bytes data / day 3M+ Apps on leading App stores By 2017 The collective computing and storage capacity of smartphones will surpass all worldwide servers 48% of enterprises are moving to the cloud to replace on-premise, legacy technology today 72% of enterprises have at least one application running in the cloud, growing from 57% in 2012 The average enterprise uses 738 cloud services.
  5. 5. © 2013 IBM Corporation A new computing paradigm is emerging Tabulating Systems Era Programmable Systems Era Cognitive Systems Era
  6. 6. © 2013 IBM Corporation Intelligent Assistance and Machine Learning - Landscape 6 IPSoft’s Amelia
  7. 7. © 2013 IBM Corporation Cognitive Era 7 Discovery & Recommendation Probabilistic Big Data Natural Language as the Interface Intelligent Options
  8. 8. © 2013 IBM Corporation Towards Computing-At-Scale as the Shared Characteristic of Recent Advances 8 Scalable Computing over MassiveCommodity Hardware Building Stronger Super Computers Cloud Computing Crowd Computing Advanced individual algorithms Mass computing applied to AI Complex array of algorithms applied to make sense of data, and offer cognitive assistance Big Data Individual MLAlgorithm Cognitive Computing
  9. 9. © 2013 IBM Corporation Understands natural language and human communication Adapts and learns from user selections and responses Generates and evaluates evidence-based hypothesis Cognitive System 1 2 3 Cognitive Systems do actively discover, learn and act A Cognitive System offers computational capabilities typically based on Natural Language Processing (NLP), Machine Learning (ML), and reasoning chains, on large amount of data, which provides cognition powers that augment and scale human knowledge and expertise Watson
  10. 10. © 2013 IBM Corporation ENTERPRISE SERVICES 10
  11. 11. © 2013 IBM Corporation Enterprise Services 11 A. Service Provider • Individual • Institution • Public or Private C. Service Target: The reality to be transformedor operated on by A, for the sake of B • Individuals or people,dimensions of • Institutions or business and societal organizations, organizational (role configuration) dimensions of • Infrastructure/Product/Technology/Environment, physical dimensions of • Information or Knowledge,symbolic dimensions B. Service Customer • Individual • Institution • Public or Private Forms of Ownership Relationship (B on C) Forms of Service Relationship (A & B co-create value) Forms of Responsibility Relationship (A on C) Forms of Service Interventions (A on C, B on C) Spohrer, J., Maglio, P. P., Bailey, J. & Gruhl, D. (2007). Steps toward a science of service systems. Computer, 40, 71-77. From… Gadrey (2002), Pine & Gilmore (1998), Hill (1977) A B C Vargo, S. L. & Lusch, R. F. (2004). Evolving to a new dominant logic for marketing. Journal of Marketing, 68, 1 – 17. “Service is the application of competence for the benefit of another entity.” Major Types of Service (provider perspective): • Computational/technology services • Business/Enterprise services • People Services Service Offerings Definition & Design Service Sales Pursuit Transition and Transformation Service Delivery & Operation Lifecycle of Enterprise (IT) Services
  12. 12. © 2013 IBM Corporation Information Technology Service Models Client Managed Procure, Own, Install & Manage [CAPEX] Vendor Managed in the Cloud On-Demand as a Pay as You Go (PAYG) price [OPEX] Applications Data Runtime Middleware O/S Virtualization Servers Storage Networking Traditional IT Applications Data Runtime Middleware O/S Virtualization Servers Storage Networking IaaS Infrastructure as a Service Applications Data Runtime Middleware O/S Virtualization Servers Storage Networking Applications Data Runtime Middleware O/S Virtualization Servers Storage Networking Managed IaaS Managed Infrastructure as a Service Applications Data Runtime Middleware O/S Virtualization Servers Storage Networking PaaS Platform as a Service Applications Data Runtime Middleware O/S Virtualization Servers Storage Networking SaaS Software as a Service Customization, higher costs, slower time to value Standardization, lower costs, faster time to valueStandardization, lower costs, faster time to value ClientManaged VendorManagedintheCloud Local, Dedicated Public Workforce Perspective Staff Body x Price x Utilization Outsource Body x Price x Utilization Digital Workforce (Bots + Body) x Price x Utilization ClientManaged…….…VendorManaged
  13. 13. © 2013 IBM Corporation Managed Information Services: From RFP to Transition and Delivery 13 Opportunity Deal Deal Deal Checkpoints/ Contract T&T Steady-State Renewal Identification Validation Qualification Pursuit QA/Risk Analysis Delivery Engagement Transition & Transformation Renewal Steady-State Delivery Business Development RFP Receipt Week 1 • Team Formation, and assignment • Control Matrix Preparation • Window of opportunity to ask questions from client Week 2-x RFP Response Deadline Solution & Approvals in Place • Proposal Writing • Client Presentation Preparation • RFP Response Items … • Detailed SOW Analysis • Baselines • SRM • Solutioning • Reviews • Approvals Control Matrix SRM FRM Baselines SOW Solutioning • Proposal • Client Presentation • Attachments/schedules Reviews and approvals CSE PM Transition and Transformation Plan • Contract Writing • Contract Analysis Service Pursuit Demystified: From RFP to Contract
  14. 14. © 2013 IBM Corporation Cognitive Enterprise IT Services Framework 14 Prior Deals Service Offerings Guidelines, methodologies People Profiles Lessons Learned Service Delivery Data Opportunity Deal Deal Deal Checkpoints/ Contract T&T Steady-State Renewal Identification Validation Qualification Pursuit QA/Risk Analysis Delivery Engagement Transition & Transformation Renewal Steady-State Delivery Business Development Current Deals Pipeline Revenue & Finance Information Integrate and Make the Data Available Using Interfaces (APIs) Deal Information Management Enable Reusing Deal Artifacts and Sharing Knowledge Deal Team Analytics Find Expertise and Recommend Them Deal Competitive Assessment Analyze Competitiveness based on Cost/Price Deal Win Prediction Analytics to provide deal win prediction, and pipeline ranking Sales Pipeline Revenue Prediction Cognitive RFP, Proposal and Contract Analyzing RFPs to extract requirements, and author RFP Response, and Contract Drafts Cognitive Solutioning Compose the set of service offerings that meets clients requirements
  15. 15. © 2013 IBM Corporation COGNITIVE RFP, RESPONSE AND CONTRACT 15 Hamid R. Motahari Nezhad, Juan M. Cappi, Taiga Nakamura, Mu Qiao: RFPCog: Linguistic-Based Identification and Mapping of Service Requirements in Request for Proposals (RFPs) to IT Service Solutions. HICSS 2016: 1691-1700
  16. 16. © 2010 IBM Corporation© 2016 IBM Corporation Input and problem statement § RFP Documents are textual documents sent by service requesters describing the requirements for IT services – The requirements are stated in natural language, with a varied format in general § RFP package contains 10s or 100s of document, each with 100s of pages describing various aspects of existing IT environment (detail baseline), and future state requirements § There are hundreds of requirements stated for each IT service in each RFP that need to be identified and analyzed, including who’s responsibility (service provider or customer) is to perform each § Different clients organize the documents and content differently, and use different vocabulary and terminology to refer to IT services and requirements § Identification of what constitute a requirement is very challenging – The structure (organization) of the document, the language construct of sentences and also client vocabulary differs – Natural language by definition can be ambiguous, documents have incomplete information, and expertise needed in interpreting and understanding requirement
  17. 17. © 2010 IBM Corporation© 2016 IBM Corporation Example IT Service Requirements
  18. 18. © 2010 IBM Corporation© 2016 IBM Corporation IT Service Requirements Analysis: the need for a meta-model 18 “Service provider shall provide onsite Desktop Services dispatching resources on 24 hour a day, 7 day a week basis, for Supported Equipment and Supported Devices at all Client’s Service Locations, which locations may be modified from time to time by Client in accordance with the applicable Change Control Procedure”. Responsible Party: Service Provider Verb phrase: shall provide Topic/Service: OnsiteDesktopServices SLAneeds: 24 hour a day, 7 day a week Services for: Supported Equipment and Devices Locations:All Client’s Service Locations Duration of service: <Contract term>
  19. 19. © 2010 IBM Corporation© 2016 IBM Corporation Requirements expressed in different form and structures A Subsection Sub-requirements SP’s Requirement Indicators SP Requirements (Extract these!) A Requirement Title of the table, potentially Service Topic [Customer]
  20. 20. © 2010 IBM Corporation© 2016 IBM Corporation Research Problems § Requirements identification – What statements constitute a requirement in RFP documents? – Requirements vs sub-requirements? § Requirements topic identification (IT services) – Which IT services they are talking about? § Service Offering Mapping - Solutioning – Which IT Service Offerings meet the client requirements? § Continues learning through Human feedback – How to manage human interactions, feedback and adaptive learning? 20
  21. 21. © 2010 IBM Corporation© 2016 IBM Corporation From RFP (Request for Proposal) to Proposal: Methodology Overview 21 RFP Documents Processing Requirements Extraction Provider Offering Matching Solution Composition Proposal Response AutomationPast RFP Response Matching Extracting requirement statements from an RFP Matching past RFP Responses for Reuse
  22. 22. © 2010 IBM Corporation© 2016 IBM Corporation RFPCog for Cognitive RFP Analysis: Overview 22 RFP Documents Contract Documents Requirements Identification Service Catalogs ITIL Requirements- Driven Offerings Composition Requirements-driven Technical Solutions Composition Solution Patterns Customer Service Vocabulary Solutions Taxonomy Provider Offering Taxonomy What are client requirement statements? What services offerings/solutions these requirements map to? Requirements Topic Identification and Grouping What are in-scope and out-of-scope service?
  23. 23. © 2010 IBM Corporation© 2016 IBM Corporation RFP Docs Structure Analysis Pattern- based Requirement Candidate Identification NLP-based Deep Learning for Requirement Identification Machine Learning- Based Topic Identification Document Table Section Paragraph Sentence Cell In what sectionof what document is the requirement from? Boundary Identification Requirement Patterns How does clients state requirements? Patterns: •( [Subject] + (shall | must | is required to | … ) ) + Action Verb + … •[Subject] is/are responsible for … • Where does a requirement start and end? è What is a requirement span? è Req., and Sub- req. identification Recognize noun (phrases), verb (phrases), … Requirement Features Apply NLP techniques for recognition of Who does what? Word Dependencies and Implicit Feature Identification Topic/Service What is the requirement about? • Linguistic-based Requirement Focus Identification • Topic-related Feature Extraction Use Domain Knowledge • Provide Service Taxonomy • Information Technology Infrastructure Library (ITIL) • Customer Vocabulary extracted from Documents Apply Supervised Learning using • Support Vector Machine • Logistic Regression RFPCog: Method Steps for Requirements and Topic Identification
  24. 24. © 2010 IBM Corporation© 2016 IBM Corporation Cognitive Solutioning - Requirements to Service Offerings Mapping § For a given requirement (or requirement group), the focus is to identify service elements (at multiple level of hierarchies) that map to the requirements, and their sub-requirements – IT Service Catalog-aware Phrase Matching – Considering the body text, concept hierarchy through a statistically-built semantic model to identify matching § Novel Method for matching noun phrases in requirements and offerings: a modified Longest Common Sequence (LCS) term matcher. – One main difference with other similarity metrics such Cosine and Jaccard is that the LCS preserves the order of tokens in matching, while other don’t. – Missing keywords in the two phrase are penalized based on the importance of the keyword 24 Based_Similarity_Score= #LCS / Weighted_ Denominator, where Weighted_ Denominator is defined as the weighted sum of the number of missing words in the E_Seq. Final_Similarity_Score = Based_Similarity_Score * (1 – net_distance/C), in which C is a constant for the maximum length of noun phrases in the population, and NetDistance is the absolute difference in tokens order difference of the LCS in NP_Seq and E_Seq (caters for additional terms in between) “Storage management solution” and “management solution”, keywords: storage, missing words
  25. 25. © 2010 IBM Corporation© 2016 IBM Corporation IT Requirements to Catalog Mapping – Interactive and Explorative Visualization 25
  26. 26. © 2010 IBM Corporation© 2016 IBM Corporation Experimental Results – Requirements Topic Identification 26 ML-based Topic Classification Performance (TP Rate) 0.9518 0.8733 0.7587 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 SVM Logistic Regression Naïve Bayes TPRate Support Vector Machine (SVM) Performance Details TP Rate FP Rate Precision Recall FMeasure ROC Area Class 0.986 0.232 0.958 0.986 0.972 0.877 F 0.768 0.014 0.908 0.768 0.832 0.877 T Weighted Avg. 0.952 0.198 0.951 0.952 0.950 0.877
  27. 27. © 2010 IBM Corporation© 2016 IBM Corporation Related Work § Templated Information extraction from text – Steven Bird, Ewan Klein, and Edward Loper, Natural Language Processing with Python, http://www.nltk.org/book/, visited July 2015. – Ana-Maria Popescu, Information Extraction from Unstructured Web Text, PhD Thesis, Uni. Washington, 2007. § Extraction of requirements from textual software descriptions (Concepts, and Models according to SVBR - Semantic Business Vocabulary and Rules- , and OPM - Object-Process Methodology, or LTL - linear-time temporal logic) – Ashfa Umber, Imran Sarwar Bajwa, M. Asif Naeem, NL-Based Automated Software Requirements Elicitation and Specification, Advances in Computing and Communications. Communications in Computer and Information Science Volume 191, Springer. 2011, pp 30-39. – Dov Dori, Nahum Korda, Avi Soffer, Shalom Cohen, SMART: System Model Acquisition from Requirements Text, Business Process Management (BPM). LNCS. Vol. 3080, 2004, pp 179-194. – Shalini Ghosh, Daniel Elenius, Wenchao Li, Patrick Lincoln, Natarajan Shankar, Wilfried Steiner, ARSENAL: Automatic Requirements Specification Extraction from Natural Language, SRI INTERNATIONAL, 14 July 2014. § This work is the first to investigate the problem of requirement extraction from natural text in RFP documents, and specifically those from services domain – Evidence-based topic identification – Novel concept-based, and cognitive similarity measure for requirements-offerings 27
  28. 28. © 2013 IBM Corporation PREDICTIVE ANALYTICS FOR IT SERVICES DEALS 28 Hamid R. Motahari Nezhad, Daniel B. Greenia, Taiga Nakamura, Rama Akkiraju: Health Identification and Outcome Prediction for Outsourcing Services Based on Textual Comments. IEEE SCC 2014: 155-162 Daniel B. Greenia, Mu Qiao, Rama Akkiraju (and Hamid R. Motahari Nezhad): A Win Prediction Model for IT Outsourcing Bids. SRII Global Conference 2014: 39-42 Peifeng Yin, Hamid R. Motahari Nezhad, Aly Megahed, Taiga Nakamura:AProgressAdvisor for IT Service Engagements. SCC 2015: 592-599 Aly Megahed, Peifeng Yin, Hamid Reza Motahari Nezhad:An Optimization Approach to Services Sales Forecasting in a Multi-staged Sales Pipeline. SCC 2016: 713-719
  29. 29. © 2013 IBM Corporation Outsourcing Service Opportunities - Pipeline Management §Service providers maintain and manage a pipeline of service opportunities to pursue. §Service pursuit management is a very elaborative, time-consuming and resource- demanding process (for large deals, $10M+) § Effective pipeline management (pipeline prioritization) and maintaining a pipeline of healthy opportunities are key for service providers –Opportunity win prediction –Opportunity health analysis 29 Objective: Build a predictive model for estimating the probability of winning strategic IT service deals, and ranking deals in the pipeline
  30. 30. © 2013 IBM Corporation Sales Opportunity Data §Quantitative information about the deal (categorical, and numerical) –Hundreds of numerical and categorical information about deals including client name, deal size (contract value), sales stage , sector, deal complexity, market analysis, quality and risk assessment, etc. §Deal comments made by the sales team and also by technical solutioning team –Comments are made at time intervals (often weekly) –Comments are short, sometimes cryptic, with specific jargons –Often do not include full English sentences, sentences are connected (no punctuation), etc. 13
  31. 31. © 2013 IBM Corporation Business and Technical Problems §Predicting the outcome of an engagement by devising a predictive model that uses both quantitative and textual comments, and analyzing them to find predictive features. –Predicting the outcome of the engagement based on quantitative and comments –How early we can predict and with what accuracy –Pipeline ranking §Identifying the health of an engagement by looking at the textual comments that made by the sales team –Engagement health: understanding the current status of the engagement by looking at the comments 14
  32. 32. © 2013 IBM Corporation Win Prediction Model: Combined Quantitative and Qualitative Model Historical Quantitative Data Score each deal and produce a prioritized list of deals Sales executives receive prioritized list 1) Deal 1 2) Deal 2 3) Deal 3 … n) Deal N Current pipeline data Logistic Regression & Bayesian Model Historical Deal comments Comment-based Prediction Model Cmment -based scores Quanti tative- based scores Combine Predictions Extensive feature engineering with defining derived features 15
  33. 33. © 2013 IBM Corporation Prioritization Performance Evaluation 33 The Win Prediction ranked list is frontloaded with deals that are likely to win: 70% of wins are in top 40% of the list. 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20 0.22 0.24 0.26 0.28 0.30 0.32 0.34 0.36 0.38 0.40 0.42 0.44 0.46 0.48 0.50 0.52 0.54 0.56 0.58 0.60 0.62 0.64 0.66 0.68 0.70 0.72 0.74 0.75 0.77 0.79 0.81 0.83 0.85 0.87 0.89 0.91 0.93 0.95 0.97 0.99 Cum.FractionofWins Cum. Fraction of Data Randomly Prioritized Win Probability Prioritized TCV prioritized Expected Revenue Prioritized
  34. 34. © 2013 IBM Corporation Deal Win Prediction using Comments 34 Textual comment Pre-processing, and Key n-gram Selection Sentiment-based Tag extractions Correlation Analysis of Extracted tags With outcomes Sentiment-based Tags tags-based Outcome Prediction Model Textual Features (key n-grams) Weighted Combined Outcome Prediction Text-based Prediction Model Builder Tag-based Prediction Model Builder Textual Feature (n-gram) Selection TermExtractor Sentiment-based Tag Extractor Feature Preparation and Selection Module Text-based Outcome Prediction Domain Vocabulary and Types project Comments New (open) project comments project Comments (Training) project Comments (Training) Combined Predicted Outcome Sentiment-based Outcome Prediction Hamid R. Motahari Nezhad, Daniel B. Greenia, Taiga Nakamura, Rama Akkiraju: Health Identification and Outcome Prediction for Outsourcing Services Based on Textual Comments. IEEE SCC 2014: 155-162
  35. 35. © 2013 IBM Corporation Illustration of the approach Sentiment-based Tag Extraction Comment Text Vocabulary SP Internal BU Partner Competition Customer New tag computation, and tag- based Outcome Prediction The set of terms identified as frequently appearing terms in from Loss Reason fields: Proposal, Price, Solution, Cost, … . Phrase-Entity Relationship <subject, phrase: sentiment, object>: new sentiment C1 C2 … … Cn Text pre-processing, comment subset selection, text feature selection C1 C2 … … Cn … Prediction (Weighted) Tag-based Predictor Sentiment- based features Project Entities Text-based Predictor Text features (n-gram) Final Predicted Outcome Comments score = ∑ s(i)* w_c(i), i is phrase with a sentiment in the update s(i): sentiment score of I, w_c(i): class memebership to indicative terms 18
  36. 36. © 2013 IBM Corporation Experiments § 4,105 historical engagement data over 3 years as the training set § Close to 500 in-flight engagement deals as the testing set 36 Experiment Overall Accuracy Win Prediction Accuracy Win Prediction Recall Loss Prediction Accuracy Loss Prediction Recall Free-form text 61.5% 72% 60% 51% 76% Text with Concept- based Features 70% 85% 61% 55% 81% Text with Concept- based and Sentiment- based Features 72.5% 87% 62.5% 58% 84%
  37. 37. © 2013 IBM Corporation37 0 5 10 15 20 25 30 35 1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 73 77 81 85 89 93 97 101 105 109 113 117 121 125 129 133 137 141 145 149 153 157 161 165 169 173 177 181 Predictive Value of the Number of comments - Win Outcome Total Comments Predictive Comment Evaluating how early (# of comments, here) the prediction matches the final outcome: between 1/3 and half of the comments A follow up analysis shows that only in 11% of cases the prediction may change as new comments become available
  38. 38. © 2013 IBM Corporation Combining Quantitative and Qualitative Analyses 38 Quantitative Model Relies on historical attributes for historical deals Comment-based Model Leverages deal team “local” insights to gauge the trajectory of the current deal (micro view). Prob. Of Winning = Weight1 x Quant Score + Weight2 x Qual Score Quant Model Qual Model Historica l Sales Data Current Deal logs Model output is combined using weights (logistic regression).
  39. 39. © 2013 IBM Corporation Sentiment-based Deal Health Analysis Historical comments Break the comment text into sentences Week 1 Week 2 Week 3 … Week n S1 S2 … Sm Sentence-level Annotation Comment-level Annotation Comment-level Annotation Comment-level Annotation … Deal-level Health Status Win, Promising, Progress, Neutral Warning, Troubled, Loss Weighted aggregation of scores Mapping each labels to a score between -1 .. 0 .. +1 23
  40. 40. © 2013 IBM Corporation Opportunity Health Analysis based on comments § Mapping each opportunity comment to a health status – “Promising”, “Progressing”, “Neutral”, “Warning”, “In-Jeopardy” § Examples – Price needs to be approved by WW – Customer has asked for some changes to the proposal – Client requirements are to be confirmed [early stages] – Agreement to proceed w/ Provider1 & Provider2 – ABB accepted the proposal from Competitor – The issues with Partner has been resolved 40 Deal Health Analytics Tool Offers functions for Monitoring the status of Deals as Sellers Comments arrive during the quarter.
  41. 41. © 2013 IBM Corporation DEAL PROGRESS MONITORING 41 P. Yin, H.R. Motahari-Nezhad, A. Megahed, T. Nakamurra, A Progress Advisor for IT Service Engagements. IEEE SCC 2015 (to appear).
  42. 42. © 2013 IBM Corporation Problem Definition and Objective § Limitations of the Win Prediction Model: Prediction is for eventualwin or loss, not for the event of the deal being rolled over to the next quarter – Prediction is for eventualwin or loss, not for the event of the deal being rolled over to the next quarter – There is no prediction capability for the outcome and timeline of milestones (key to deal success) – There is no idea on when key events (such as win or loss) would happen § Objective: – Building a model that gives analytical insights about the key events and milestones as well as the timeframe within which they happen 42
  43. 43. © 2013 IBM Corporation Analysis § Analysis shows that distribution of time intervals for the occurrence of key events and milestones decays exponentially § Longer time interval of no activity (event progression) leads to a higher chance of losing 43 Time Unit EmpiricalProbability (c) Probability of Loss w.r.t. time unit Geometric distribution
  44. 44. © 2013 IBM Corporation Methodology § Devise a Bernoulli based deal-specific process for the prediction of event time intervals – It identifies the probability of the occurrence of events and thus helps in understanding how fast or slow a deal is moving forward – This model is used to learn the weights of deal attributes to compute the parameter of a geometric distribution for the next event occurrence time interval § Bernoulli-Dirichlet Generative Process: models the type of occurred events: win, loss, Milestone update – It is trained to learn the weights of deal attributes to compute the parameters of a stochastic process that models the type of next occurring event § Prediction – The model estimates the probability of different event types given the deal attributes X, and time interval T, i.e., probability that the given event may happen within the time interval T 44
  45. 45. © 2013 IBM Corporation DEAL COMPETITIVENESS ASSESSMENT 45
  46. 46. © 2013 IBM Corporation The basic premise to be used throughout the Deal Competitive Assessment is to be able to compare a given ‘Compare From’ source to available “Compare To” data sources through a standard method of peer selection, and to present the output in a standard way globally Tower/Service Scope Peer Criteria Peer Selection Criteria Compare To Sources Bid Data Market Data Delivery Data DiminishingNumberofDataSamples X1 X2 X3 X4 X5 Local Sources Compare From Sources Deal Metric Standard Global Representation Standard Model Offering Standard Models Offerings Contract Prices Pricing Deal Competitiveness Assessment 30
  47. 47. © 2013 IBM Corporation Approach to Assessing Competitiveness §Mine ‘similar’ prior deals and market benchmark data §Determine the upper and lower bounds on unit costs and unit prices for each of the service involved in an IT service solution. §Add things up to get upper and lower bounds, and assess the percentile of the given case. §Create a case management solution, where: –Users can edit/add/remove services involved. –Users can see/change/add peer deals §The key challenge is in determining ‘similarity’ among complex IT service solutions. We present an approach to derive close comparables in this effort 47
  48. 48. © 2013 IBM Corporation Peer-Selection Filtering § Boolean: Has global resources or not § Geographical: Where it was § Categorical: Won, lost, or either § Numerical: Quantity of services § Unstructured text: Attributes with long text descriptions, images, etc. § Timing: recent enough. 48 Tuple: {service, # of units requested, $unit cost, $ unit price, geo deliver-from, geo deliver-to} D1 s1, 200, $44 s2, 300, $2.88 s3, 2000, $555 s4, 1000, $674 cs1, N/A, 10% cs2, N/A, 20% D2 s2, 200, $3.50 s4, 3000, $500 cs1, N/A, 12% cs2, N/A, 18% D3 s1, 500, $40 s3, 1,500, $450 cs1, N/A, 15% cs2, N/A, 22% D4 s2, 200, $3.50 s4, 1500, $620 cs1, N/A, 12% cs2, N/A, 18%
  49. 49. © 2013 IBM Corporation The System View of IT Service Solution Price Competitiveness Analysis 33
  50. 50. © 2013 IBM Corporation Sales Pipeline Revenue Prediction Methodology Overview 50 Historical Win Conversion & Growth Data What future opportunities would come into the pipeline that will be won by the end of the period (Growth)? Wouldwewinthese opportunities(Conversion)? Non-Linear Optimization Model Linear Optimization Model Optimal Weights Optimal No. of Historical Periods to Use (N) Current Pipeline Revenue Prediction (Conversion & Growth Apply Weights on N Historical Conversion and Growth Rates Apply Rates to Current Pipeline Objective: Predicting the revenue of sales pipeline for different sales stages Aly Megahed, Peifeng Yin, Hamid Reza Motahari Nezhad:An Optimization Approach to Services Sales Forecasting in a Multi-staged Sales Pipeline. SCC 2016: 713-719
  51. 51. © 2013 IBM Corporation FROM SERVICES TO COGS, AND TO COGNITIVE BPM What advances in AI and Machine Learning mean for Service Computing and BPM? 51
  52. 52. © 2013 IBM Corporation Service Computing: From API to CCL § The End of using API for Programming Business Logic – APIs will be used to initiate Cogs (Intelligent Bots) – The Business Transaction to be performed in Conversations with Cogs § Cogs representing Providers/Consumers,spanning over a spectrum: – From Cogs taking over the interface of existing Apps – To Cogs codifying and understanding the business logic and engaging in conversations to transact § Cog Conversation Language (CCL) – CCL should provide support for defining a rich natural language conversations for a Cog to deliver business functionalities to the users (other Cogs, and Humans) • The Language to Program Cogs • An initial example is Watson Dialog Services Template Language 52 Source: blog.cloudsecurityalliance.org
  53. 53. © 2013 IBM Corporation The notion of Service/People Composition to be Re-Defined § In current Hybrid composition/mashup (People, Services) methods: – Services are represented with API calls – People are integrated with Human Tasks (GUI is the interaction paradigm) – Composition methods are finding deterministic models of interactions, defined apriori § We are moving towards dynamic composition of cogs and human in which – Cogs are participating in NL conversations – Human are approached through messaging and natural language – Composition are performed dynamically during the conversation,require non-deterministic models, defined in online and on-demand model 53 Weather Cog Health Agent Personality Insight Cog. Provider Cogs Travel Cog 1 Travel Cog 2 Planning a Vacation Trip Considering preferences, experience, conditions, cost, Availability, etc. Mediated and facilitated by Cogs Human-Cog interaction Cog-Cog interaction Natural Language Natural Language, CCL, (ACL, KQML, etc.)? ACL: Agent Communication Language, KQML, etc.
  54. 54. © 2013 IBM Corporation The App Composition (Mashup) is already moving away from explicit API calls § Implicit Data Sharing with the notion of Central Shared Context on Mobile Platforms – Events – Notifications – Metadata descriptions § Google Now on Tap (implicit integration) – Central Shared Context § Apple Proactive 54
  55. 55. © 2013 IBM Corporation Process Automation Stages in Enterprise & in IT Services Humans (Manual) Program/ Workflow Robotics (RPA) Cognitive 55 Issues Current Enterprises facing • High volume of manual processes • With high variability • Involving unstructured data “85% of a typical firm’s 900+ processes can be automated.” High Cost of Automation using Traditional Approaches (to go from 50% to 85%)
  56. 56. © 2013 IBM Corporation Historical and Future Perspectives on BPM 56 Databases BackendSystems Layer Self-Generating Integration SAP using java API Web Service API Excel using com API MSMQ using com or java API Databases using jdbc API Business Rules Layer Production Business Level Objects Business Level Objects Inv oices Business Lev el Obj ects AFE’s Business Level Objects Anything Business Level Objects Process Layer Any Process General Workflow System and UserInteractionsCalculation Interface Layer Web Service Presentation Presentation XML API BackendSystems Layer Self-Generating Integration SAP using java API SAP using java API Web Service API Web Service API Excel using com API Excel using com API MSMQ using com or java API MSMQ using com or java API Databases using jdbc API Databases using jdbc API Business Rules Layer Production Business Level Objects Business Level Objects Inv oices Business Lev el Obj ects AFE’s Business Level Objects Anything Business Level Objects Process Layer Any Process General Workflow System and UserInteractionsCalculation Interface Layer Web Service PresentationPresentation PresentationPresentation XML API XML API BPMS TQM General Workflow BPR BPM time ERP WFM EAI ‘85 ‘90 ‘95 ‘05‘00‘98 IT Innovations Management Concepts DatabasesDatabases BackendSystems Layer Self-Generating Integration SAP using java API Web Service API Excel using com API MSMQ using com or java API Databases using jdbc API Business Rules Layer Production Business Level Objects Business Level Objects Inv oices Business Lev el Obj ects AFE’s Business Level Objects Anything Business Level Objects Process Layer Any Process General Workflow System and UserInteractionsCalculation Interface Layer Web Service Presentation Presentation XML API BackendSystems Layer Self-Generating Integration SAP using java API SAP using java API Web Service API Web Service API Excel using com API Excel using com API MSMQ using com or java API MSMQ using com or java API Databases using jdbc API Databases using jdbc API Business Rules Layer Production Business Level Objects Business Level Objects Inv oices Business Lev el Obj ects AFE’s Business Level Objects Anything Business Level Objects Process Layer Any Process General Workflow System and UserInteractionsCalculation Interface Layer Web Service PresentationPresentation PresentationPresentation XML API XML API BPMS BackendSystems Layer Self-Generating Integration SAP using java API Web Service API Excel using com API MSMQ using com or java API Databases using jdbc API Business Rules Layer Production Business Level Objects Business Level Objects Inv oices Business Lev el Obj ects AFE’s Business Level Objects Anything Business Level Objects Process Layer Any Process General Workflow System and UserInteractionsCalculation Interface Layer Web Service Presentation Presentation XML API BackendSystems Layer Self-Generating Integration SAP using java API SAP using java API Web Service API Web Service API Excel using com API Excel using com API MSMQ using com or java API MSMQ using com or java API Databases using jdbc API Databases using jdbc API Business Rules Layer Production Business Level Objects Business Level Objects Inv oices Business Lev el Obj ects AFE’s Business Level Objects Anything Business Level Objects Process Layer Any Process General Workflow System and UserInteractionsCalculation Interface Layer Web Service PresentationPresentation PresentationPresentation XML API XML API BPMS TQMTQM General Workflow BPRGeneral Workflow BPR BPMBPMBPM time ERPERP WFMWFM EAIEAI ‘85 ‘90 ‘95 ‘05‘00‘98 IT Innovations Management Concepts Ref: Ravesteyn, 2007 ‘16 Social BPM iBPMS: Business Process Analytics ‘2021 The Future of BPM is also Cognitive Dark Data Cognitive BPM Cognitive Analytics Cognitive Processes Interact LearnEnact Cognitive Capabilities
  57. 57. © 2013 IBM Corporation Dark Data: digital footprint of people, systems, apps and IoT devices § Handling and managing work (processes) involves interaction among employees, systems and devices § Interactions are happing over email, chat, messaging apps, and § There are descriptions of processes, procedures, policies, laws, rules, regulations, plans, external entities such as customers, partners and government agenies, surrounding world, news, social networks, etc. § The need for activities over interactions of people, systems, and IoT devices to be coordinate 57 Citizens Assistant Business Employees/ agents Plans Rules Policies Regulations TemplatesInstructions/ Procedures ApplicationsSchedules Communications such as email, chat, social media, etc. Organization Dark Data: Unstructured Linked Information IoT Devices and Sensors
  58. 58. © 2013 IBM Corporation Spectrum of Work: Processes and Cognitive 58 Structured Processes Unstructured Processes Knowledge-based Routine Existing Technology Dark Data: Mobile, Social, Communication (email, voice, video), Documents, Notes, Sensors BPM Engines Workflow Engines Case Management Groupware Knowledge-Intensive Processes Email, Chat, Messaging Ad-hoc, unstructured Processes Cognitive Process Management Conversational Interface for Processes Cognitive Process Learning Cognitive Process Analytics Cognitive Enactment
  59. 59. © 2013 IBM Corporation Cognitive BPM Systems § A Cognitive BPM system is a cognitive system that provides cognitive support in all phases of a process lifecycle over structured and unstructured information sources, and is able to continuously discover, learn and proactively act to support achieving a desired outcome – It offers cognitive interaction and analytics support over structured processes – For unstructured processes, it offers intelligent and integrated process (model) definition, reasoning and adaptation • Process is not assumed apriori defined; but is discovered, learned and customized based on accumulated knowledge and experience –It continually learns to improve the process 59
  60. 60. © 2013 IBM Corporation Cognitive BPM Lifecycle 60 Cognitive BPMS Define Enact Monitor Analyze Next Steps, Adapt Interact Sense Learn, Discover To Traditional BPM Cognitive BPM
  61. 61. © 2013 IBM Corporation Cognitively-Enabled Processes: Shifting process lifecycle from Define-Execute-Analyze-Improve to Plan-Act-Learn § For each enactment of the overall process, many iterations around this loop § At a given time, multiple goals & sub-goals may be active – Numerous threads of activity – Each thread modeled essentially as a “case” as in Case Mgmt – Cf. [Vaculin et al, 2013] § As new information arrives the cycle might re-start for some or all threads – Planning based on new info • New goal formulation • Planning to achieve those goals § “Cognitive Agent” helps by – Perform the planning – Learn from large volumes of structured/unstructured data – Over time, learn best practices and incorporate into planning Plan / Decide Act <<World Effect>> Learn Richard Hull, Hamid R. Motahari Nezhad: Rethinking BPM in a Cognitive World: Transforming How We Learn and Perform Business Processes. BPM 2016: 3-19
  62. 62. © 2013 IBM Corporation Towards Cognitive BPM: Example Scenarios 62 Example (1): Integrate IBM BPM with IBM Watson http://www.ibm.com/developerworks/bpm/library/techarticles/1501_mehra-bluemix/1501_mehra.html#N1009D Email, Chat, and Calendaring apps are the most used channels for doing work in the enterprise Addressing the work organization and management for Knowledge workers: monitoring communication channels (email, chat), and: - Capturing, prioritizing and organizing work of a worker - Identifying actionable statements (requests, commitments, questions) and track them over the course of conversations Example (2): eAssistant for Knowledge Workers
  63. 63. © 2013 IBM Corporation63 Inbox - Verse Highlighting actionable statements Recommending fulfilment actions IBM Insight 2015 – The session on “Given your collaboration tools a brain”
  64. 64. © 2013 IBM Corporation64 IBM Insight 2015 – The session on “Given your collaboration tools a brain” Send File Action Archetype Send File Action Archetype Send File Action Archetype
  65. 65. © 2013 IBM Corporation65 IBM Insight 2015 – The session on “Given your collaboration tools a brain” Invite/Calendar Action Archetype Automated Invite Parameters Extraction Calendar Entry Creation
  66. 66. © 2013 IBM Corporation eAssistant App and APIs 66 Watson (& BigInsight NLP) Apps and Services on BlueMix CollaborationTools Enterprise Repositories, Applications and Data Sources Feeds Repositories Document collections … eAssistant Apps Personal Knowledge Graph Builder Conversation Analytics, Auto-Response, Prioritization Calendar and Scheduling Assistant Cognitive Process Learning To-do, Task and Process Assistant Cognitive Work Assistant APIs Semantic Role Labeling POS tagging Dependency Analysis Co-reference resolution Named Entity Recognition Knowledge Graph Builder H. R. M. Nezhad. Cognitive assistance at work. In AAAI Fall Symposium Series. AAAI Publications, November, 2015.
  67. 67. © 2013 IBM Corporation Cognitive BPM: Selected research challenges § Cognitive process learning: 4Knowledge acquisition methods from unstructured information (text, image, etc.) 4Combine with traditional process mining on logs 4Building actionable knowledge graphs & executable code § Cognitively enabled processes: Plan-Act-Learn 4Blending of “model” and “instance” 4Recognizing goals from digital exhaust and process history 4Advances in planning research – incremental, multi-threaded activity, richer goal languages, prioritized and soft goals, … 4Enough uniformity to support reporting, identification of best practices § Cognitive Assistants for business processes 4Assist workers across numerous tasks, including process management & optimization 4Interactive learning where cognitive agents ask process questions 4Gradual learning through experience, and process improvement
  68. 68. © 2013 IBM Corporation Summary § The Future of Computing is …. § The Future of Work is …. § The Future of Services is …. § The Future of BPM is …. § A huge, unprecedented opportunity for the research community to advance our understanding,methods and technology underpinning these transformations and disruptions! 68 Cognitive Cognitive Computing Cognitive Assistance Cognitive Services Cognitive BPM
  69. 69. © 2013 IBM Corporation QUESTIONS? Thank You! 69
  70. 70. © 2013 IBM Corporation Model of Human AdministrativeAssistants: conceptual framework 70 T. Erickson,etc.: Assistance:The Work Practice of Human Administrative Assistants and their Implications for IT and Organizations,CSCW’08. Blocking, Doing, Redirecting Key to the performance of Assistants
  71. 71. © 2013 IBM Corporation Cognitive BPM in Cognitive Assistants/Agents § Goals – Increasing worker’s productivity, efficiency, and creativity (serendipity) § Current cognitive assistants are focused on personal space or virtual conversationalagents § Cognitive Work Agent – Is process and work aware – Monitors worker’s input channels and interactions (emails, chats, social connections,external and internal environment, knows rules, policies and processes) – Proactively acts on worker’s behalf and reacts to requests: becomes a copy of you in work environment • Commands/requests - Responds to simple requests intelligently • Situational awareness – monitors the environments to overcome information overloading (selective). • Deep QA: process questions, how-tos, previous successfulprocess experience – Organizes and assists your work • Extract tasks/commitments, promises, commitments • Managed to-dos: status updates, over-dues, plans • Manages calendar, schedules,social contacts • Finds and present prior related interactions to a particular conversation – Learns how work gets done, and can take care of them for their human subject 71
  72. 72. © 2013 IBM Corporation Cognitive Assistant § A software agent (cog) that – “augments human intelligence” (Engelbart’s definition1 in 1962) – Performs tasks and offer services (assists human in decision making and taking actions) – Complements human by offering capabilities that is beyond the ordinary power and reach of human (intelligence amplification) § A more technical definition – Cognitive Assistant offers computational capabilities typically based on Natural Language Processing (NLP), Machine Learning (ML), and reasoning chains, on large amount of data, which provides cognition powers that augment and scale human intelligence § Getting us closer to the vision painted for human-machine partnership in 1960: – “The hope is that, in not too many years, human brains and computing machines will be coupled together very tightly, and that the resulting partnership will think as no human brain has ever thought and process data in a way not approached by the information handling machines we know today” “Man-Computer Symbiosis , J. C. R. Licklider IRE Transactions on Human Factors in Electronics, volume HFE-1, pages 4-11, March 1960 72 1 Augmenting Human Intellect: A Conceptual Framework, by Douglas C. Engelbart, October 1962

×