Flash Summary
By some estimates more than 80% of AI projects fail, and each one spends time, compute and budget before it returns anything. Camila Bühler explains why that footprint stays invisible to most organisations, and how the AmPhi Impact Prioritization Framework scores environmental, social and responsible AI impact next to the business case before anything is built. It closes with a Profit × Planet prioritisation matrix you can apply to your own AI use cases.
Artificial intelligence is changing the economics of innovation.
Organizations can now prototype, test, and deploy solutions faster than ever before. Tasks that once required months of development can be explored in weeks. New ideas can be validated quickly, often with relatively limited upfront investment. For innovation teams, this is an extraordinary opportunity.
Yet there is a paradox at the heart of this new reality.
While it has become easier to build, it has not necessarily become easier to create value.
According to research cited by the RAND Corporation, more than 80% of AI projects fail by some estimates, a rate significantly higher than traditional IT projects. More importantly, RAND found that these failures are often not technical in nature. Organizations frequently struggle because they misunderstand the problem they are trying to solve, pursue technology before identifying a meaningful need, or optimize for the wrong objective.
Source: RAND, The Root Causes of Failure for AI Projects (2024)
At the same time, AI adoption continues to accelerate. McKinsey's latest research shows that while organizations are investing heavily in AI and launching increasing numbers of initiatives, many remain stuck in experimentation and pilot phases, with relatively few achieving significant enterprise-wide impact.
This raises an important question.
If so many innovation projects fail to deliver meaningful outcomes, what are the true costs of those failures?
Most organizations can answer that question in financial terms. They know how much was spent on software, infrastructure, consultants, or development resources.
Far fewer can answer it in environmental or social terms.
And yet every innovation project leaves a footprint.
Innovation has a sustainability footprint
When we think about sustainability, we typically focus on products, operations, supply chains, or emissions. Innovation itself is rarely viewed through the same lens.
However, innovation consumes resources long before it creates value.
Every project requires employee time, leadership attention, data processing, computing infrastructure, organizational energy, and financial investment. Even relatively small experiments consume resources that could otherwise have been allocated elsewhere.
- Employee time
- Leadership attention
- Data processing
- Computing infrastructure
- Organizational energy
- Financial investment
The spend is certain from day one. The return is still a bet.
This is not an argument against experimentation. Innovation requires exploration, and not every initiative can or should succeed.
The challenge is that most organizations have no structured way of understanding the broader impact of innovation choices before they commit resources.
As AI lowers the barriers to experimentation, the volume of innovation activity increases. More pilots are launched. More ideas are tested. More solutions are built.
The environmental and social implications of those decisions, however, often remain invisible.
The result is that organizations may become highly effective at building things without fully understanding whether those things are worth building in the first place.
The difference between efficiency and value
Consider one of the most common AI applications today: customer service automation.
The business case is straightforward. AI can reduce operational costs, provide 24-hour support, and handle growing volumes of customer interactions. From a purely financial perspective, the opportunity appears compelling.
Yet the impact extends beyond cost reduction.
- What happens to customer service employees whose roles are significantly altered or eliminated?
- How does automation affect customer trust and satisfaction when complex issues require empathy, judgment, or contextual understanding?
- What governance mechanisms are needed to ensure transparency and accountability?
These questions do not necessarily invalidate the business case. In many situations, AI-powered customer service may be entirely appropriate.
However, they do reveal something important: business value alone does not tell the whole story.
Now compare this with a different application.
A logistics company uses AI-powered route optimization to reduce unnecessary mileage across its delivery network.
The financial benefits are clear. Fuel costs decrease, fleet utilization improves, and operations become more efficient.
At the same time, emissions are reduced, drivers spend less time in traffic, and customers receive more reliable deliveries.
The innovation creates value across multiple dimensions simultaneously.
Both examples involve AI.
Both may generate positive returns.
Yet their overall impact profiles are fundamentally different.
Understanding those differences is becoming increasingly important.
A missing layer in innovation decision-making
Traditionally, innovation opportunities are evaluated through two primary lenses: business value and feasibility.
Can we build it?
Will it create value for the business?
These remain essential questions.
What is often missing is a structured assessment of the broader consequences an innovation may create.
- How will it affect people?Social
- What environmental impact might it have?Environmental
- Does it align with principles of responsible AI?Responsible AI
- How many people will benefit from it?Reach
- How long will those benefits last?Duration
These considerations are increasingly relevant for organizations navigating sustainability commitments, stakeholder expectations, regulatory developments, and growing scrutiny around AI adoption.
Yet they are rarely incorporated into innovation prioritization in a systematic way.
Making the invisible visible
At AmPhi Labs, we have been exploring this challenge through the development of the AmPhi Impact Prioritization Framework (AIPF).
The framework was designed to help organizations evaluate the likely sustainability impact of an innovation opportunity before development begins.
Rather than focusing solely on business outcomes, it assesses three distinct areas: environmental impact, social impact, and responsible AI considerations.
Environmental impact is evaluated across five categories - carbon emissions, water use, air quality, waste and circularity, and biodiversity - ensuring that strengths in one area do not obscure weaknesses in another.
The framework also considers two additional factors: reach and duration. After all, an initiative that benefits a small group for a few months should not necessarily be evaluated in the same way as one that creates meaningful impact for thousands of people over many years.
Each environmental category gets its own score, so a strong carbon result cannot cover a weak water result.
Importantly, the framework does not replace business evaluation.
An innovation with significant environmental benefits still needs a viable business case. Equally, an initiative with a compelling commercial opportunity may warrant further scrutiny if it creates unintended social or environmental consequences.
The objective is not to prioritize sustainability over business, or business over sustainability.
It is to make both visible.
Because better innovation decisions require a more complete understanding of impact.
We love problems
One of the most striking findings from RAND's research is that many AI projects fail because organizations misunderstand the problem they are trying to solve.
This resonates deeply with how we approach innovation at AmPhi Labs.
Our motto is simple: We love problems.
Not because problems are inherently interesting, but because the quality of an innovation is often determined long before a solution is built.
The most important decision in any innovation process is not the choice of technology.
It is the choice of problem.
When organizations start with a clear understanding of the challenge they are trying to address, they dramatically improve their chances of creating meaningful value. For the business. For people. And for the planet.
AI is making it easier than ever to build solutions.
The harder challenge is ensuring those solutions are directed at problems worth solving.
Looking ahead
Innovation will always involve uncertainty. Not every project will succeed, nor should it.
But as AI accelerates the pace of experimentation, organizations have an opportunity to become more intentional about where they invest their attention, resources, and ambition.
The sustainability conversation should not begin once a solution has been deployed.
It should begin much earlier, when decisions are being made about which opportunities deserve investment in the first place.
- Choose the problemAsk it here
- Prioritizeand here
- Build
- DeployUsually asked here
By deployment, the footprint is already spent. At the start, changing course costs a conversation.
Because every innovation leaves a footprint.
The question is whether that footprint creates lasting value - or simply becomes another hidden cost of getting innovation wrong.
- 1Route optimization
- 2Customer service automation, and where the guardrails can move it
Profit × Planet is a multiplication. A zero on either axis gives zero.
Illustrative placement of the two examples in this article.
Start with the problem worth solving
The AI Flash Scan takes 10 minutes and asks about your own operation: where the waste leaks and how close the next compliance deadline is. Free, with no data to prepare. Your first opportunity signals are on screen the moment you finish.
Start the Flash Scan →The Profit × Planet matrix, scored for your own opportunities, is part of the AI Value Deep Dive.