Artificial intelligence is a wave that is already here and transforming enterprises, promising a future vastly different from our experiences, environment, and work today. If your agile team isn't already touching backlog items related to AI, it soon will be - or a team better prepared to handle AI work may intercept what will become an increasingly larger share of enterprise workloads.

Agile teams have the imperative to start working with AI technology and tools, incorporate an AI focus into their current and future backlogs, and begin delivering more than just working code: Working models, experiments, and cognitive services that augment and expand the capabilities of their applications and enterprises.

This new world includes new team roles, like Data Scientist; new development pipeline tools: Automated data ingest and scrubbers, data lakes, algorithm libraries, test and training data repositories, model registries, AI labs; and new processes and tasks, such as cognitive service integration, data labeling, model training, testing, and deployment, and algorithm performance evaluation.

This talk describes the tools, infrastructure, roles, and agile process changes required for taking on AI epics so that your team can can get started on its own AI journey.

 
 

Outline/Structure of the Talk

I. The AI imperative

A. The wave is here - current adoption trends for cognitive capabilities and services

B. Enterprise readiness

i. AI adoption is not a product purchase - the ubiquitous nature of AI

ii. Defining what is possible given where you are, and how to plan for growth

II. Nuts and bolts of building an intelligent system

A. AI components

i. Cognitive services: Natural language processing, visual recognition, translation, speech-to-text, text-to-speech

ii. Machine learning: Unsupervised vs. supervised learning, model ensembles, model QA

iii. Data management: Data lakes, data cleaning, labeling, test sets, training sets, model registries, deployment patterns

B. Identifying AI opportunities

i. Process decomposition

ii. Data analysis

iii. Interviews/surveys - "day in the life"

C. Cognitive solution approaches

i. Third-party cognitive service integration

ii. Machine learning model pipeline

iii. Hybrids

III. Put the AI in Agile

A. AI-specific agile team resources

i. Data intake management

ii. Data scientist

iii. Enterprise system-and-data coordination

iv. AI QA

v. AI security

B. Practices for discovery, elaboration, and construction

i. AI as a portfolio investment theme

ii. Discovery buckets: Process, data, and artificial intelligence

iii. Constructing field-able experiments

C. "Deliver working AI" through DevOps and development pipeline enhancements

i. AI infrastructure

ii. Devops processes with AI enhancements

iii. Feedback loops and measuring percentages

Learning Outcome

An understanding of the Agile practices, team roles, processes changes, and devops/pipeline changes need to incorporate AI development efforts into an Agile team structure.

Target Audience

Current or new Agile practitioners ready to incorporate strategies to attack the work of the future.

Prerequisites for Attendees

Basic Agile principles, execution experience is helpful but not required.

schedule Submitted 2 years ago

Public Feedback


    • Lisa Schlecht
      keyboard_arrow_down

      Lisa Schlecht - Why you’re not getting value out of scaled agile (and how to fix it)

      45 Mins
      Workshop
      Intermediate

      Your organization has ventured into agile and you have one or more agile teams. You may have experimented with additional coordination layers like programs, or portfolios.

      But you’re not seeing the business outcomes everyone had hoped for. And stakeholders are getting restless.

      You want to add value at scale while avoiding excessive overhead. But what you have in place isn’t quite working. And you’re wondering what to try next.

      This discussion will cover some typical issues that organizations face when shifting to a scaled agile model for creating value in complex environments.