Managing AI Projects (In Large, Legacy-Driven Companies)

Today’s edition of the Augmented Advantage is co-authored by Valentin Marguet, project lead in the automotive industry.

There's a group of some 90 million business professionals who often overlook the drastic impact of AI on their profession. A profession that has already been under great pressure during the digitalization of companies. Yes, we're talking about project managers (PMs).

Project management is facing a growing influence of AI from all directions - with many professionals unprepared, especially in large, legacy-driven corporations.

Valentin studied how to leverage AI in his role and shares practical insights. Together, we’ll show you how project managers can effectively handle AI and drive project success.

Let's go!

Project Management + AI = 🫠

Project managers today face AI on multiple fronts:

1. Using AI to boost their own productivity

2. Managing projects with increasing AI components

3. Proving human superiority over AI in project management

While the first point could warrant its own future blog post, points 2 and 3 are closely intertwined and focus of our issue today.

If project management can't prove more effective than algorithms and computers, PMs risk becoming the new secretaries (largely replaced by technology). And managing projects has become harder than ever, especially with AI initiatives emerging everywhere, legacy industries included.

Why Managing AI Projects Is So Hard

AI projects are unique beasts: short, dynamic, hard to assess, risky, and fed by constant technical evolution.

Approaching them with traditional project management methods is a recipe for failure. Make no mistake, AI projects need management, and they can run with known methods. However, these methods must be adapted into something specific to win the race.

We can break down this approach into three parts:

1. Setting the stage for success

2. Turning plans into action

3. Controlling and improving continuously

These are classical methods that shouldn't surprise you. What's novel is how you adapt them to AI.

Today, we're focusing on Part 1 - setting AI projects up for success.

Setting AI Projects Up for Success

When tasked with bringing an AI project to life within a legacy context, anyone in charge should start by answering these crucial questions:

• What is the real problem I want to solve?

• When and how does the project end?

• What resources are needed and available?

Looks easy, right? But AI projects make these questions surprisingly tricky to answer, especially for companies used to a completely different set of projects.

Here's a thought: what if we combined the best of both worlds? A hybrid approach - marrying traditional project management's structure with the flexibility needed for AI - could be the key. This method allows for context-specific problem-solving and adaptable planning, crucial for AI projects in legacy environments.

With this in mind, let's address these three main clusters to set up your project for success:

1. Project Decomposition: Define and break down the problem.

2. Granular Planning: Create a detailed, flexible plan.

3. Team Formation: Allocate resources effectively.

Let's apply this advice and go to the next level. We'll list the main challenges and best practices within these three clusters:

Project Decomposition

Main Challenge: Complexity and dynamic nature of AI technology.

AI projects are inherently complex and dynamic. While the technology is powerful, it's still maturing rapidly. Today's best-in-class tool might be obsolete next week. Besides, legacy companies often face additional challenges such as managing datasets, selecting appropriate algorithms, and integrating AI solutions with existing systems. The risk of scope creep looms large, making it difficult to maintain focus and direction.

Solution: Structured, yet flexible breakdown.

To manage these challenges, equip your project with a structured yet flexible breakdown.

Step-by-Step Approach:

1. Define the Main Problem: Clearly state the overarching problem your AI project aims to solve. Validate this problem from multiple business perspectives. Consider both immediate goals and long-term vision.

2. Identify Major Components: Break the problem into major components like data collection, model development, testing, and deployment. Balance detailed tasks with room for iterative refinement.

3. Decompose Further: Divide each component into smaller, manageable tasks. For each sub-problem, set SMART (Specific, Measurable, Achievable, Relevant, and Time-bound) goals to ensure clear expectations and accountability, but give space to more flexible objectives for exploratory aspects.

4. Visualize the Structure: Create a tiered Work Breakdown Structure (WBS) to organize these tasks logically and sequentially.

   • Top tiers: Fixed phases and major deliverables

   • Middle tiers: Key components with flexible execution

   • Lower tiers: Adaptable tasks for iterative development

   • Pro tip: Color-code elements (e.g., green for fixed, yellow for flexible, red for highly adaptable) to quickly communicate adaptability levels. Here's an example of a WBS chart of a hypothetical AI-powered quality control system:

     

5. Schedule Regular Reviews: Business problems and project goals change over time. Conduct sprint-like reviews to reassess goals and adjust the breakdown as needed, involving both technical and business stakeholders to adapt to new insights or changes in technology.

Granular Planning

Main Challenge: Unstable timelines and timing conflicts.

AI projects often face unstable and unpredictable timelines. Even with proper decomposition, various factors can cause delays, such as underestimating risks or the complexity of moving from prototype to production. Additionally, AI development frequently encounters timing conflicts, where one task might be ready while another becomes a bottleneck.

Solution: Detailed, yet flexible planning..

To address these challenges, create a precise, logical, and flexible plan. A hybrid approach, blending structured and agile methods, often works best for AI projects in legacy environments. Both traditional and adaptive methodologies can be effective, depending on the project scope and complexity.

Step-by-Step Approach:

1. Create a Detailed Timeline:

   • Develop a logical timeline with taks and milestones, detailed for near-term and flexible for long-term.

   • Define tasks dependencies while allowing adaptable task order.

   • Identify inputs and outputs for each phase.

   • Use tools like Gantt charts to visualize the schedule.

2. Assess Risks and bottlenecks:

   • Identify potential risks and bottlenecks at each stage.

   • Develop contingency plans and strategies to mitigate their impact.

   • Implement regular risk reassessments to adapt to AI project evolution.

3. Allocate Buffer Time: 

   • Incorporate buffer times in your schedule to account for unexpected delays and issues, using techniques like the Critical Chain Method to manage buffers effectively across the project timeline.

4. Blend Agile end Waterfall Methods (based on project needs):

   • Waterfall: Suitable for well defined and stable components projects. Divide the project into distinct phases (requirements, design, implementation, verification, maintenance).

   • Agile: Fit for uncertain aspects of AI development. Split development into small, manageable increments called sprints.

5. Regular Updates and Reviews: 

   • Maintain an up-to-date project schedule.

   • Establish regular review sessions to assess progress, identify bottlenecks, and make necessary adjustments.

Bonus - Tools and Techniques: Utilize project management software supporting hybrid approaches (e.g., JIRA with BigPicture, Monday.com, Asana, etc.). These tools should enhance your hybrid method, allowing for both structured planning and agile task management in AI projects within legacy environments. Remember, these tools are just the cherry on top, not the solution itself (yet).

Team Formation

Main Challenge: Expertise and Resource Constraints

Find the right talent mix for AI projects is hard. Business acumen is crucial, but technical AI skills are also non-negotiable. Limited budgets and resources, and legacy existing organization structures further complicate the situation, so you need to show the value before getting more funding.

Solution: Blended teams

Blended teams offer a flexible solution, combining diverse individuals working on varied schedules and employment models to achieve specific business goals. They typically include:

• Full-time and part-time employees

• Freelancers

• Direct contractors

• Consulting firms

• Staffing agencies

The composition may evolve as the project progresses, with roles shifting based on project needs.This approach allows companies to get the right talent mix at the right time while acknowledging both internal and external resource constraints

Step-by-Step Approach:

1. Define Roles and Responsibilities: Clearly outline the roles and responsibilities required for the project. Use a RACI (Responsible, Accountable, Consulted, Informed) matrix to assign roles for each task.

2. Identify Key Skills: Ensure a mix of domain, IT/engineering, and AI expertise on your project team, which typically spans multiple departments (cross-functional teams).

3. Resource Allocation: Prioritize tasks using frameworks like MoSCoW (Must have, Should have, Could have, Won’t have) to so you can focus on critical activities first. Run the framework several times during the project to reassess your needs.

4. Leverage External Expertise: Partner with AI consultants or firms to fill expertise gaps if needed. It should foster knowledge transfer to build in-house AI capabilities over time.

5. Promote Continuous Learning: Support ongoing development to keep your team updated in the quickly evolving AI field. Establish a 'learning budget' for team member upskill programs.

6. Establish Clear Communication: Set up regular updates and meetings to keep everyone aligned and communicate transparently.

7. Monitor and Adjust: Continuously assess team performance and be ready to reallocate roles and resources as needed. Implement regular 'team retrospectives' to assess the effectiveness of the hybrid team structure and make necessary adjustments

Conclusion

AI projects are not just another item on your project portfolio. They represent a new frontier in technology and business, demanding a fresh approach. While the fundamental principles of PM remain relevant, the unique challenges of AI projects require us to adapt and evolve our methods.

Try a balanced approach - blending structured methods with adaptive techniques

Remember, the goal isn't just to 'implement AI,' but to create meaningful, sustainable value for your organization by bridging traditional processes with innovative AI capabilities.

As we've seen, setting the stage for success is crucial, but it's only the beginning.

In on of our next issues, we'll further explore Part 2 of managing AI projects: turning plans into action. We'll explore how to execute your carefully laid plans in the dynamic and often unpredictable world of AI development.

Stay tuned!

Valentin and Tobias