Despite years of change management, lessons learned, and proven methodologies, just 15-20% of AI projects that you work on this year will succeed.
Agility is key when working on a topic that’s moving as fast as AI. What’s hot today can easily be old news tomorrow, superseded by faster, cheaper, or higher-performing technology. Yet, despite the rush, it is important to also consider the risks when leading AI projects or programs (a portfolio of projects).
Because, more of than not, is it a few key factors along the way that separate successful from failed AI initiatives. Few of them are really technology-induced. That’s why I created a new course with LinkedIn Learning that helps senior leaders learn about the ins and outs of AI projects.
NEW ONLINE COURSE — Mitigate AI Business Risk
Business leaders are under pressure from their boards and competitors to innovate and boost outcomes using AI. But this can quickly lead to starting AI projects without clearly defined, measurable objectives or exit criteria.
Learn how to implement proven risk mitigation strategies for starting, measuring, and managing AI projects. Along the way, get tips and techniques to optimize resourcing for projects that are more likely to succeed.
There is one stat that has stayed fairly consistent throughout the past 7-8 years of AI hype cycles: according to Gartner, 80-85% of projects fail or don’t deliver the value they’ve originally set out to create.
While it is easy to blame such a high failure rate on a new technology like AI, it is rather the people affected by the change that require the attention. Team members whose way of working changes because of a new AI capability in their workflow are often opposed to this kind of change.
Leaders who have been promised quick and highly successful projects to their superiors are surprised about sudden roadblocks or setbacks in AI projects after status reports have turned from red to orange to green the higher they went in the hierarchy.
Clear success criteria are missing from the get-go. Critical go/ no-go decisions are delayed or not even made, and so, AI projects continue on past the point of feasibility, hoping that the next breakthrough is just around the corner—with a bit more time, a bit more data, a bit more tweaking. But those aren’t the only challenges…
AI Projects are Unlike Any Other Projects You’ve Led
One of the most common challenges in any AI project is the assumption that the project flow is linear. Professionals and leaders are so accustomed to breaking down a goal into work packages, estimating start and finish dates, resources, and budgets, that any other approach seems illogical. (More details on how to overcome the “project mindset” in the AI Leadership Handbook.)
Difference between traditional projects and AI projects
But AI projects are, in fact, iterative. Traditional machine learning projects have been more akin to research projects when models are built from scratch. Generative AI and Agentic AI have lighter requirements on the data science side by using foundation models that someone else has developed.
But ensuring that the AI assistant or agent pulls the correct data, stays within its guardrails, does not hallucinate, etc., still makes the process iterative rather than linear. AI leaders need to manage this expectation with their stakeholders, who may also be under pressure to deliver results from AI.
Your Data is the Foundation for AI—Scary, Right?
When Generative AI was new, it seemed as though this key constraint in machine learning projects had been resolved: data. By crafting a good prompt, users can elicit responses from LLMs that are difficult to distinguish from human-created language. Sadly, these outputs often sound very generic and bland as well.
Difference between data and AI-usable data
The solution to this situation is to add data to the prompt or make it available to the assistant or agent for querying. However, most businesses struggle not with capturing data, but rather with the challenge of making it useful for AI. This includes aspects such as accuracy, completeness, and freshness of data.
Addressing these points as part of a “Phase 0” of an AI project can drive improvement while using the available budget allocated for AI effectively.
Thank you for reading The AI MEMO. Feel free to share it with your network.
These aren’t the only challenges in an AI project—yet, they are all solvable. But addressing them upfront requires awareness, foresight, and credibility.
The single most important aspect is clearly defining the business problem and the project’s success criteria. Without these two, your company will likely treat AI like a hammer (“in search of nails”) when your actual problem requires a different approach (like a screwdriver).
Join my live stream and podcast for leaders and hands-on practitioners. Each episode features a different guest who shares their AI journey and actionable insights. Learn from your peers how you can lead artificial intelligence, agentic AI, generative AI & automation in business with confidence.
July 15 - Jon Reed (Industry Analyst & Co-Founder of diginomica) and I will discuss the state of Enterprise AI and Agents as we head into the summer.
July 29 - Steve Wilson (Project Leader at OWASP Foundation and Chief Product Officer at Exabeam) will share how organizations can secure their AI agent deployments. [More details to follow…]
Watch the latest episodes or listen to the podcast
