Why most AI projects fail — and the playbook for the ones that don’t

Table of Contents

By Christophe Monget, CEO at MoiraCorp. McKinsey and MIT have been telling us why AI projects fail. The findings line up with what we see at MoiraCorp — and most of it is fixable.

I have been reading the recent research from McKinsey and MIT on why a large proportion of corporate AI projects still fail. Both reports converge on the same uncomfortable point — one we have observed across every client engagement at MoiraCorp.

Most firms treat technology-driven transformation as an IT project when it should be a business project. That framing error is the parent of every other problem.

If your AI journey is stalling, it is almost certainly one of these five traps.

The five traps that stall AI transformations

1. Underestimating the front line

Your team is already using GenAI. They are using ChatGPT on personal accounts, Copilot when it appears in their tools, and any number of point solutions that have crept in without procurement noticing. The “readiness gap” that consultants love to diagnose is often an illusion — the readiness is there, just not visible. The gap is between what staff are doing and what leadership thinks they are doing.

2. Steering too slowly

The technology is moving fast. Leadership decision-making, in most regulated firms, is not. By the time the AI committee has met three times and produced a policy, the tools have shipped two new releases and the operational case has changed. The bottleneck is rarely the tech. It is the rhythm at which the firm can decide.

3. The ROI trap

The temptation is to pick small, cool pilots that are easy to demonstrate. The trap is that they prove nothing about the business. A chatbot that summarises meeting notes is a feature. A redesigned suitability workflow is a transformation. Most firms produce a stream of features and call it progress. The systemic, business-wide change is the only thing that actually moves the P&L.

4. Ignoring the ethics

Fairness, privacy, and explainability are not soft topics. They are the conditions under which your staff and your clients will trust the system enough to use it. Skip them at the start and you will discover, somewhere around the scaling phase, that the people you needed to bring with you have quietly decided not to.

5. The skill gap myth

The narrative is that firms cannot scale AI because there is a talent shortage. That is true, but it is not the whole picture. The deeper problem is under-resourcing the talent you already have. Firms hire one AI lead, give them no team, no infrastructure, and no decision rights — and then conclude they need to hire more AI leads. The talent gap is real. The resourcing gap is bigger.

The MoiraCorp Strategic Playbook

The traps describe how firms get stuck. The playbook describes how they get unstuck. Four moves, in order:

Align leadership

Strategy-led roadmaps, not tactic-led experiments. Pick a “Champion” with the seniority and the mandate to bridge the gap between technology, business, and risk. Without that bridge, every decision becomes a negotiation between functions that do not understand each other.

Empower the internal champions

The research is clear: your most enthusiastic AI adopters tend to be your millennial managers, the 35-44 cohort. They have the operational knowledge, the technical literacy, and the appetite. Resource them — give them time, budget, and the authority to redesign workflows. They will move faster than any centrally-driven programme.

Responsible governance, designed federated

Central governance for safety, explainability, and policy. Federated execution at the team level. Teams need autonomy to move; the firm needs visibility to evidence compliance. The architecture that delivers both is federated, not centralised.

Modular scaling

Do not get locked into a single vendor. The model landscape is moving fast enough that the leader today will not be the leader in eighteen months. Build a technology stack that lets you swap underlying LLMs and tools as the market evolves. The firms that commit to one vendor on day one are buying tomorrow’s switching cost.

From stalled to scaled

The honest question for any leadership team right now is this: are you still playing with pilots, or are you reinventing the business? The two activities look similar from a distance. They produce very different results.

The five traps are the default mode. The playbook is the alternative. Neither is complicated. What separates the firms that scale from the firms that stall is the willingness to treat this as a business transformation and resource it accordingly.

Wrestling with a similar regulatory or operational challenge?

We help regulated firms reduce the friction between what compliance requires and what teams actually have to do — through better processes first, AI where it earns its place. A 30-minute Business & Automation Review maps where your time is going and where automation could pay back fastest.

Related posts
Compliance Testing – Fairness Assessment using R
Retrieval Augmented Generation (RAG) augmented by ML can help in Proactive Risk Identification enabling predictive analysis to identify potential issues regarding unbalanced customer selection.
Company default prediction – DLMM internal rating model in R
Most firms are sitting on data that could predict which clients are at risk or which investments are underperforming. Machine learning is the type of artificial intelligence that enables computers to learn from this existing knowledge and data.
Behavioral & decision-making quantification
GenAI can adopt a persona and "make decisions" or "behave" in a way that can be quantified. This technique is used to simulate scenarios, which can then be analyzed quantitatively and used in particular to assess multi-criteria decision alternatives
Prompt for data
Extracting quantitative information using GenAI tools requires to properly structure the prompts used to question them to efficiently use their large language models (LLMs)
Machine learning augmentation: Closing the Data Gap
Machine learning is a type of artificial intelligence that enables computers to learn from existing knowledge and experiment results. These models are traditionally used for prediction and can be augmented by GenAI for training data generation and screening in particular
Retrieval augmented generation (RAG)
Retrieval Augmented Generation (RAG) is a critical technique using proprietary or domain specific documents to augment base LLMs to address specific enterprise or applications needs.