Managing AI Projects Effectively

AI technologies are evolving faster than ever, creating both opportunities and challenges for organizations trying to stay ahead. While this rapid pace offers exciting possibilities, it also comes with some tricky challenges. For example, in our AI Workshops, we strongly advise clients to have a three-month review cycle built into their processes. You need those planned so that you can adjust accordingly.

Some other challenges include managing scope creep, keeping costs in check, or ensuring we stay compliant with new laws. Let’s look at how these changes can impact AI projects, along with some strategies to tackle each one. I’m grouping these by the type of risks your project will face.

Related podcast of this content that was created with NotebookLM: https://open.spotify.com/episode/31vlq07T380TEaeXicf3BD

Risk: Scope Creep

Scope creep is a major problem for AI projects in a rapidly evolving environment. While unrealistic expectations are a typical problem, the immaturity of the technology will lead to many unknown unknowns, leading to unplanned work to address the challenge at hand.

Managing Expectations

The excitement around AI can sometimes lead to unrealistic expectations from leadership or stakeholders. Teams may be asked to deliver more than what’s feasible, or too soon. This problem is so common that we have an AI Strategy & Awareness Accelerator to help stakeholders better understand what’s possible.

Strategies

• Clear communication of limitations. Be upfront about what AI can and cannot do, setting realistic expectations from the start.
• Deliver in small increments. Focus on delivering tangible results incrementally, showing progress along the way without over-promising.
• Manage stakeholder expectations. Regular check-ins with stakeholders to align on goals, timelines, and progress will prevent surprises down the line.

Scope Management

AI projects are particularly vulnerable to scope creep, especially as new technologies emerge mid-project. Many teams develop the “Shiny New Object” syndrome, continuing to spin on new capabilities but failing to deliver what was promised. Also, picking the appropriate use cases for AI tech can help prevent these challenges. Keeping focused on original goals while balancing innovation is crucial. We go deeper into this topic and data management in our AI Use Case & Data Readiness Accelerator .

Strategies

• Set clear goals early. Define what success looks like from the start and stick to these objectives, with a change management process to handle unexpected requests.
• Agile project management. Use agile methodologies to allow for flexibility in technology adoption without veering off-course. Tie success to impactful output rather than checking boxes.
• Educate stakeholders. Help stakeholders understand the importance of sticking to project scope, so new technologies don’t derail the project timeline.

Data Management Challenges

As AI models become more sophisticated, so too do the demands on your data. Ensuring your data is clean, well-structured, and accessible is more important than ever. This may spawn more projects to create the necessary infrastructure and clean up the data. Starting with small scope will help mitigate this to an extent.

Strategies

• Prioritize data governance. Implement strong data governance policies to ensure data quality, privacy, and accessibility.
• Automate data processing. Use AI-driven data cleaning and transformation tools to handle large datasets more efficiently.
• Create a standardized data platform. Establish a standardized unified data platform where all teams can access and manage data seamlessly and combine it with local data is key.

Integration Complexity

Integrating new AI technologies into existing systems is rarely straightforward. Often, it requires reconfiguring infrastructure, leading to delays or compatibility issues.

Strategies

• Phased integration. Implement new AI technologies in phases to ensure smooth integration and minimize disruption to existing systems.
• Invest in adaptable infrastructure. Use cloud-based solutions that offer the flexibility to integrate new tools with minimal friction.
• Partner with integration experts. Bringing in external partners who specialize in system integrations can help navigate these complexities.

Risk: External

The macro environment is also rapidly developing. Best practices are evolving quickly, and new laws are springing up left, right, and center. Portfolio managers would be prudent to keep an eye on these rapid developments. Sources like NIST and industry newsletters will provide great ways to keep on top of the latest developments.

Evolving Standards and Best Practices

AI technologies shift fast, and with that come new standards and best practices. Staying up to date can feel like a moving target, especially when different teams adopt different methods. We discuss this in greater detail in our AI Governance Accelerator .

Strategies

• Centralize AI governance. Establish a dedicated team responsible for keeping up with industry standards and disseminating best practices across departments.
• Adopt flexible frameworks. Use flexible project frameworks like agile, which can adapt as standards evolve without completely disrupting workflows.
• Invest in automation. Automate code quality checks and compliance assessments so new practices are integrated without manual intervention. Also, ensure humans have visibility and transparency with what the automation is doing.

The Evolving Legal Environment

New AI laws, like the EU AI Act, are emerging globally, and staying compliant is becoming a much larger concern. Non-compliance can lead to costly delays, rework, or fines.

Strategies

• Proactive legal reviews. Regular legal audits help keep projects compliant with the latest regulations.
• Stay ahead with ethical AI. Develop internal ethical AI guidelines that go beyond compliance, setting you up for future regulations.
• Incorporate compliance early. Embed legal and compliance considerations from the very start of your AI projects to avoid rework down the road.

Ethics and Compliance

With the introduction of laws like the EU AI Act , organizations need to keep ethical AI use at the forefront of their projects. Staying compliant while navigating ethical concerns is a growing challenge.

Strategies

• Stay ahead of regulations. Partner with legal experts who specialize in AI to proactively monitor changes in regulations and stay compliant.
• Conduct regular ethics reviews. Build bias detection, fairness audits, and transparency checks into the development process to maintain ethical AI.
• Assign a compliance team. Dedicate a team to oversee compliance and ensure projects align with evolving regulatory standards.

Risk: Team

How do you upskill a team in a rapidly changing environment? I strongly recommend learning by doing. Make mistakes and share the lessons learned broadly within the team. Also, take advantage of freely available vendor training as well as use vendors like Marquee Insights to get deep training.

Skill Gaps and Training

As AI technologies evolve, keeping teams up to speed becomes a challenge. Skill gaps can lead to delays or mistakes, especially if teams are using outdated knowledge.

Strategies

• Create a continuous learning culture. Regular training and upskilling programs ensure your team stays on top of AI advancements.
• Leverage AI tools for learning. Use AI-powered learning platforms to deliver personalized training based on each employee’s progress.
• Hire AI specialists. When it’s too much for internal teams to keep up, hiring AI consultants or specialists can fill the gaps temporarily.

Resource Allocation

Allocating the right resources to AI projects can be a juggling act, especially when new technologies demand new skills, tools, and financial investment.

Strategies

• Flexible budgeting. Build flexibility into your budget to account for unforeseen costs, whether that’s new tools or specialist hiring.
• Resource prioritization. Focus on high-impact projects and avoid spreading resources too thin across many initiatives.
• Cross-training teams. Empower teams with cross-functional skills so resources can be allocated more dynamically as project demands change.

Risk: Organizational

There are two primary organizational risks. First, keep a sharp eye on your AI vendors. Consolidation among providers is inevitable given where we are in the cycle. You don’t want merger and acquisition activity to ruin your progress. Also, keep close watch on budgets and costs. While some aspects of AI are getting cheaper, rapidly, the cost of the whole is still high. Combine this with pressure to cut costs in corporations and you could easily wind up in budget trouble.

Vendor Management

The vendor landscape in AI is constantly changing, making it challenging to select long-term partners. Some vendors might seem like the perfect fit today, but new entrants may offer better technology tomorrow.

Strategies

• Strategic vendor partnerships. Form long-term relationships with vendors who align with your AI strategy, rather than simply picking the newest players.
• Diversify your vendor portfolio. Don’t rely on a single vendor for all AI needs. Diversifying your vendor base ensures you’re not locked into one tool.
• Vendor evaluation framework. Develop a formal evaluation process that regularly assesses vendors based on performance, innovation, and support.

Cost Management

While AI usage costs are dropping, the introduction of new, more powerful tools often drives up project costs. Balancing these dynamics can be tricky.

Strategies

• Incremental adoption. Don’t chase the latest AI tech just because it’s new. Start with MVPs and incrementally roll out features to control costs.
• Thorough cost-benefit analysis. Evaluate whether the benefits of the new tech justify the increased cost, ensuring every upgrade makes financial sense.
• Optimize existing tools. Before upgrading to the next big thing, maximize the value of your current AI tools by fully exploring their capabilities.

Thank you for reading! If you like this issue, feel free to share it with colleagues.