Artificial intelligence (AI) has become a major focus for companies across industries who want to leverage these advanced technologies to drive innovation and gain a competitive edge. However, successfully implementing AI is easier said than done. Industry reports indicate that 85% of AI projects fail to deliver on their intended promises.

As more organisations undertake AI projects, there is a growing need to understand the issues that arise during these endeavours and the practices that can help address common challenges. In this blog post, I share insights from an in-depth case study of Consult - a growing North American AI consulting firm founded in 2017 that has successfully delivered custom AI solutions to customers in various industries.

By examining Consult's approach to managing client AI projects, this post provides practical strategies that can help organisations navigate the complexities of AI implementation to achieve better outcomes.

The AI Consulting Landscape

The AI consulting field has expanded rapidly in recent years. AI consulting firms provide advisory services to help organisations:

• Identify AI opportunities that align with strategic goals
• Assess data, infrastructure, and capabilities needed for AI adoption
• Design and execute AI implementations with custom solutions
• Train staff and transition projects into production

Effective AI consultants combine expertise in business strategy, project delivery, and data science. They guide clients through the uncertainties of AI adoption and accelerate realization of its benefits.

Top global consulting firms like Accenture, Deloitte, and McKinsey now have dedicated AI consulting practices. Specialized AI consultancies like DataRobot and Element AI also offer services tailored to AI strategy, implementation, and change management.

With deep experience in AI and analytics, these AI consulting firms and practitioners play a key role in helping organisations navigate AI successfully.

Managing AI Projects at Consult

Founded in 2017, Consult provides AI consulting services focused on helping organisations leverage data and AI to solve business problems. Consult has delivered innovative custom solutions for companies in logistics, supply chain, transportation, and other industries.

The management of AI projects at Consult combines elements of:

• Traditional project management - Used to structure the project into phases for planning and control
• Agile practices - Used to organize work iteratively using cycles of development and feedback
• AI workflow - Used to guide the required data and model development tasks

This integrated approach allows Consult to meet client needs despite the uncertainties of AI. However, it also gives rise to conflicts between the different ways of thinking about and managing the work.

By studying how Consult balances these conflicting approaches, we can derive useful strategies for managing AI projects effectively.

Key Logics in AI Project Management

Our analysis identified three key logics that shape how work is perceived and conducted in AI projects:

1. Traditional Project Management Logic

This well-established logic emphasizes formal planning, controlled execution, and adherence to constraints like budget and schedule. It assumes that following standardized processes will lead to success.

2. Agile Logic

Guiding modern software development, this logic values responding to change, customer collaboration, and frequent delivery of working solutions. It assumes that an adaptive process will succeed despite uncertainty.

3. AI Workflow Logic

Emerging in AI systems development, this logic is guided by an experimental, scientific workflow. It focuses on tasks like data preparation, model building, evaluation, and tuning through testing and analysis.

These logics conflict in their norms, values, roles, and assumptions about the work of AI projects, leading to issues in practice.

Strategies for Managing AI Project Conflicts

The presence of conflicting logics causes difficulties in the coordination, planning, execution, and measurement of progress of AI projects.

By studying Consult, we identified four key strategies they use to balance the logics and successfully deliver projects:

1. Assess Project Viability Continuously

AI workflow uncertainties persist throughout projects, rather than just early phases as traditional project management expects. Consult proactively evaluates project feasibility during each phase to ensure it remains viable or halt projects that are too uncertain.

2. Rethink Definitions of Project Progress

The intermittent, experimental outputs from AI tasks don't always align with agile iteration goals. Consult focuses on scientific learnings rather than only working software to demonstrate progress.

3. Link AI Outputs to Business Value

AI models aim for technical accuracy, while projects target business goals. Consult ensures model outputs are evaluated based on marginal gains in business value versus costs.

4. Foster AI Expert Collaboration

Specialized AI roles make cross-functional collaboration difficult. Consult partners business consultants closely with data scientists to bridge business and technical needs.

These strategies address conflicts between the logics and are key to Consult’s effective management of AI projects for clients.

Implications for Managing AI Projects

The insights from studying Consult highlight important considerations for organisations undertaking AI initiatives:

• AI projects require managers to balance potentially conflicting logics and workstyles that shape AI delivery. Understanding these logics helps identify issues proactively.
• Strategies are needed to continuously evaluate project viability, adapt progress measures, link AI outputs to business value, and foster collaboration between experts.
• AI consultants can play a pivotal role in bridging gaps in business and technical knowledge to guide clients through uncertain AI journeys successfully.

As AI becomes increasingly crucial for competitive success, organisations need to build capabilities to manage AI projects strategically. AI consulting firms like Consult are an invaluable partner in this journey, complementing internal capacities.

By leveraging leading practices from exemplars like Consult, while tailoring execution to their unique contexts, organisations can overcome the pitfalls that plague many AI initiatives and unlock the tremendous potential of AI.

Essentially, effective management of AI projects requires embracing new ways of working that combine proven project management approaches with strategies tailored to the demands of AI development. With the right partners and perspective, companies can navigate uncertainty and complexity to turn AI projects into strategic assets.