Program Management, PMO, and AI Best Practices

Summary: PM / PMO Role in AI Best-Practices in an Organization.

AI Is Emerging from Point-Solutions to Enterprise Function.

As stated by Google in their Gemini Enterprise – AI in Enterprise Workflow announcement: "For the past few years, AI at work has been about deploying point solutions for individual tasks. And while helpful, these are only incremental improvements. They don’t deliver fundamental transformation because they can’t work across the silos and fragmented apps where work actually gets done.

"The reality is, you simply can't transform a business through individual prompts, questions, or tasks. The real breakthrough comes when you move beyond the prompt to automating entire workflows."

Every individual, every function in an organization, including Program Management, will discover and adopt AI features. AI is truly a disruptive technology with massive upside and unpredictable consequences. AI already has massive and broad early investment, likely followed by productivity enhancement, labor and resource enhancement and displacement, and eventual winners and losers. Every aspect of these will have challenges.

So, going a level above "which tools to choose now" (which will continually evolve), how can organizations adopt AI tools and product content coherently, for best possible outcomes in productivity, labor, resources, and business result?

PMO has a stake in this: PMO is already responsible to the organization for effective operation of a major process key to its business success. Part of the answer for AI will be that Program Management should play a pivotal role in capturing and implementing Enterprise AI Best-Practices in Programs and Organizations as AI evolves.

Published concurrently on svprojectmanagement.

Published concurrently on www.softtoyssoftware.com.

A recent Survey conducted by Refactoring illustrates current Enterprise use of AI, and highlights opportunities as well, that can be addressed by PMOs.

The Survey, entitled The State of AI Adoption in Engineering Teams by Refactoring covers Demographics, Personal and Team AI adoption, Skills and Jobs, and Adoption Path of AI.

The Survey Identifies Currently Common Use Cases for AI. In Order of Frequency:

• Coding
• Automation and Repetitive Tasks
• Research and Learning
• Problem Solving and Brainstorming
• Documentation
• Testing and QA
• Operations

The Survey Reveals Organizational Effects of AI

• "The balance between operation and strategic hours in our workdays will concentrate more on the strategic side, which for me is where we generate value. — Director of Engineering
• "It’s taking away the toil work and leaving more space for the challenging things I can run with. — VP of Engineering
• "I spend less time with repetitive and boring work, and more on study and what really matters. — Tech Lead
• "I’m delighted when an agent completes a task for me without me having to lift a finger. I see AI doing boring, simplistic tasks for me while I focus only on the tasks that require human thinking. — Software Engineer
• "I sometimes feel like it's a superpower that help me skip the bs and focus on what interests me (building great products). — Staff Engineer
• "For 73.2% of respondents, AI has changed the skills they look for, with the #1 effect being a stronger focus on high-level engineering chops (e.g. system design) vs expertise on specific languages and frameworks."

The survey identifies some use cases, including this example Best-Practice Adoption

Documentation

• "Using AI for documentation creates a great feedback loop, because docs not only benefit humans — they benefit AI as well, which then is able to 1) understand code better, and 2) provide better answers about the codebase, which is another widespread use case."
• That is, Use AI to help create documentation. Then put the documentation into the repo to be used by AI to help explain the code as needed.

A Common AI Adoption Issue is discussed

Team Adoption

• “The lack of top down – or just shared – direction not only hampers the amount of benefits the team as a whole can get from AI; it also fails to engage the minority of skeptics that exist in almost every team:
• “[Adoption process is] a mix between top down and participated. There’re some ICs really engaged but it’s required a lot of top-down to get traction. Still, many ICs don’t use even the most proven use cases
• “The main reason for this is that, simply put, we are still at a stage in which no one knows what they are doing.
• “When asked for the biggest challenges in team AI adoption, the #1 factor mentioned by engineers is the lack of best practices and volatility of tools:
• “Most teams realize the quality of outcomes vastly changes based on context design and workflows, but very few are actually investing in learning these skills through dedicated resource allocation. Most engineers are just expected to learn on the job.
• “Lack of understanding / training of how to best leverage tools and how to provide correct context. Lack of experience. Not enough sharing of best practices.”

The Survey analysis introduces an organizational model for Exploration and Adoption of AI

• Individuals explore (tools, use cases, practices) -> Team embraces -> Organization empowers

Explore

• Create knowledge-sharing avenues through ceremonies where people can demo wins and spread emerging best practices.
• Small automations; Small feature AI implementation; small refactoring/migration tasks; Documentation and Test creation by AI

Embrace

Following when initial exploration has created a basic level of proficiency across the team, it's time to embrace adoption and graduate it into team practices that go beyond individual usage.

• AI code reviews; AI testing standards; AI documentation standards; Documents into repos for AI context
• AI meeting summaries
• Feedback loops, Quantification

Empower

• Expand scope of individuals
• Full-stack developers; specification via prototyping
• Cost reduction, velocity improvement, professional growth, organization scaling
• Leadership Vision

This model reveals an opportunity for Program Management to take action

Coordinating and propagating AI Best Practices is a key opportunity for Program Management, led by an organizational PMO.

To re-state Google’s problem statement:

• AI use is expanding beyond individual task query, to Enterprise relating cross-datasource collaboration including proprietary internal and external data sources, and cross-domain process automation.
• Infrastructure elements are evolving to support and propagate Agent-based solutions, leading to autonomous Agentic solutions.
• A rich new environment, and immediate exploration, is evolving to support AI-enabled Enterprise functionality.
• The universe to be explored includes existing tools and data structures to be used for AI, as well as new constructs to host newly AI-enabled Enterprise data and workflow. Following examples illustrate the depth of exploration currently available and expanding.

PMO focus would be on Enterprise-level AI best practices, more than on individual-level practices. Focus would be on Team and Organizational practices and automations rather than individual Prompt engineering.

• Specific platforms cited
• Anthropic MCP protocol
• CData MCP Agent Platform
• Client-Server based platform hosting Agent Connectivity, Security and Semantic functions
• Foundations
• Google Gemini A2A
• Any-to-Any based platform enabling autonomous Agent functionality
• Gemini Enterprise – AI in Enterprise Workflow
• Many more platforms are emerging…
• Salesforce Vibe (link), Grok Enterprise (link), …
• Reference: www.softtoyssoftware AI-Resources – Many use cases, technologies, tools, courses, references.

AI Agent Platform Principles

• Early AI broad focus has been on Prompt engineering. Now, as enterprise agentic applications gain focus, structure of AI applications is also evolving quickly and demanding focus, while engineering of Prompts created in the applications also remains important.
• Emerging AI-enabling frameworks provide development leverage, providing connectivity to on-prem and external data sources for context, and also connectivity to and among agents leading to automation and autonomous AI operation.
• New structures, tools, and methods are being adopted that use these constructs. Here’s some context.

These are the principal barriers to Enterprise deployment beyond pilot

Data Fragmentation

• As more data sources become essential as AI context, APIs to a broadening set of data types (e.g., database, file, un-structured) are numerous and expanding.
• As more data sources are incorporated for context, schema alignment capability becomes a paramount requirement.
• "You may know your Schema, but AI doesn't" – CData (verbal, during Foundations conference)

Security

• Data to be used by an organization may be proprietary, well-understood and In-House; or Public, from both external or internal sources, with secured visibility.
• Access Security must be mapped from application through Agents to Data Sources, maintaining security provisions at every level.
• Internal Agents within an organization may provide shared access to both internal and external data. External agents may provide access to organizational data, from both external and internal organizations. Access security controls use of data at every level through such paths.

Agent Platforms Provide Connectivity

• Application to Agent
  • MCP: Anthropic Client-Server Agent API.
  • A2A: Google Any-To-Any Agent API.
• Agent to Data Sources and Context
  • Application APIs, JSON
• Agent to Agent
  • A2A: Google Any-To-Any API.

Agent Platforms Provide Semantics

• Smart Data Dictionary
  • Interpret data Schemas and translate for JOINs among data sources.

Agent Platforms Propagate Security

• • Client through agents to internal or external data, mapping and controlling Role capabilities through access paths.

Agent Platforms Provide Framework for Business Content

• Reporting, smartly connected and smartly created
  • Supporting business management needs with broadening AI-controlled data
  • Database sources, un-Structured content including JSON, Excel, Text, Application-Specific data.
  • Integrating research, vision and decision-making among internal and external data sources.
• Process Automation
  • Automation APIs and Applications (e.g. n8n, Zapier, Latenode, Power Automate)
  • Agent – to – Agent
  • Autonomous operation
  • Collaborative integration with and among Partner processes.

Autonomous Agent Processes

• • Agents contain Process logic, Workflow Automation, etc..

Now, let’s translate the Potential Energy of Vision to the Kinetic Energy of Implementation (a job familiar to Program Managers!): how Program Management can participate in exploration of AI, and then in AI best practice implementation into organizational processes and into business benefit.

Exploration of AI tools, methods and workflows will occur throughout the organization. Some explorations will be organized; many will be organic. Results, and effort required, will vary. Program Management is well-positioned to capture results, to organize and host evaluation and identify winners, and to facilitate their adoption.

PM participation involves organizational capture, evaluation, and deployment of AI best practices into adoption across an organization. PM is more likely to focus on promoting use of relevant discovered AI in organization product and workflow, rather than to architect AI capabilities themselves.

• Facilitate AI exploration within Programs
• Capture and Collect AI Practices
• Facilitate AI-Process grading and sharing
• Expand grading/sharing process to include organizational processes

Derived Organizational Adoption Process

The following abstracted AI adoption model expands points from the preceding linked Refactoring article, translating to implementation actions.

• Explore AI tools, use cases, implementation ->
  • By individuals, architects, organizational task forces
• Teams embrace ->
  • PM captures best practices via Retrospectives, Survey and other external contact.
• Organization Empowerment
  • PM propagates adoption of best practices via org-level focus, conferencing, dialog, and documentation.

Organization Empowerment: Within Enterprise, Create and host an “AI Best Practice Forum” to Capture, Grade, and Deploy AI Best-Practices

• Within Organization collect a forum of AI practitioners
  • (for example, a large Enterprise organization's similar PM forum was twice per year, and involved several hundred PMs each time)
  • Presentation, questions, summary documentation on internal implementations of AI tools and methods.
• Periodic Presentation Forum
  • AI Forum participants sign-up for 5 – 10-minute chunks to present an AI practice for attention (success, considerations, pitfalls etc.).
  • Each presentation provides tangible documentation of the process, implementation, results, effort, and considerations.
  • Forum meeting includes some Business or Technology presentations relevant to Program Managers and PMO (e.g., Agentic AI Enterprise Platform evolution; AI Ethics; Regulatory; Security handling propagation re on-prem and external data sources)
• Practices presented and documented are collected by PMO, and placed on an organizational “AI Practices” website
  • Practices on the website are tagged to program, and may attribute credit to tech, PMs and key leaders
• Each Practice on the resulting website includes ability to comment, and rate thumbs-up/thumbs-down.
  • This is to gain quantified community grading of effectiveness.
  • Raters can optionally identify themselves or remain anonymous.
  • This might be similar to up-vote/down-vote systems seen on or
• Website provides quantified analysis of ratings of included Practices
  • Perhaps differentiating anonymous vs. identified ratings.
  • Objective of ratings is to promote most successful (best) practices; and can further identify practices needing further development before broader adoption.
• AI Best Practice Forum Focus Can Expand Beyond Adoption by Programs
  • PM will focus on Program AI adoption, but the processes involved can expand to facilitate AI evaluation and adoption in functions beyond in-Program AI adoption.
  • For example: MfgOps, Supply Chain, Finance, Marketing, Sales, Product Management, Support, et al.
• PMO Can Facilitate Surveys of More Program Management Practitioners
  • Sources of AI adoption opportunities may be broader than internal exploration.
  • Identify non-Proprietary practice.
  • Via PMI forum meetings
  • Partners; AI tool providers (participation, or research) Survey organizations
    For example, Refactoring collaborated with

AI Tool and Method Deployment Within the Organization

Dona Sarkar LinkedIn post illustrates a current example within Microsoft

• "This has to be a C-Suite AND grassroots activity.
• "Someone in the C-Suite MUST schedule a weekly demo day where THEY go first and show their AI use cases, prompts and such. This is how you find your data and AI issues and fix them in real time!
• "This keeps people from feeling afraid they will be seen as being replaceable or lazy if they use AI AND it will give them new ideas for their own AI usage".

Deployment Recommendations

• "Start Small ->
  • Within a small subset of programs, incorporate and adapt local processes to use identified AI usage.
  • Reciprocally, adapt and evolve AI processes to fit into local processes, technologies, and organization structure, as AI processes are locally deployed.
  • Organizationally broad Transformations to incorporate AI best practices may be adversely disruptive, where testing of new methods has been focused and limited, where adjacent processes may be affected with unexpected consequences. So finding successful methods in "pilot" projects is advised.
  • This is especially true for cross-functional organizations – for hardware for sure, and even software programs likely involve Design/development, SQA, IT, Release, Support, and maybe Finance organizations and processes.
• -> Expand What Works"
  • As functional adoptions succeed, expand deployment incrementally to more product elements, projects, programs, organizations: HW/SW, DevOps, Ops, Marketing, Product Management, Sales, Support, Manufacturing Ops, Supply Chain…
  • Create a project leader to oversee implementation.
  • With expanded deployment, scale is also likely to require adaptation and evolution of both existing processes and new AI practices.
  • A plan for adoption should include effects on adjacent processes, verification, documentation, metrics and cost.

Program Management organizes, coordinates and ultimately directs a substantial portion of any organization's mission-work. Program Managers and PMO are therefore well-positioned to identify and propagate AI best-practices across an organization. In fact, one could argue that it would be irresponsible not to step up to that challenge!

The data collected and cited as described just above, led by Program Management in the organization as described, and the methods explored to identify, evaluate and propagate AI Best Practices, can surely have a very positive effect on organizational effectiveness, its business prospects, and on realization of personal and organizational growth and aspirations!