By Louis-David Benyayer and Hao Zhong
Every day, more companies adopt AI, yet profits remain flat. Stuck in the “light bulb phase”, most tweak the old instead of transforming the core. This article shows how to unlock true value.
There is a clear gap between the adoption and excitement about AI and the translation of these investments into corporate profits. Why is this the case, and what can companies do about it?
The Lag in Value Creation
Signs of corporate AI adoption are increasing every day: new software versions are released regularly, start-ups see their list of corporate clients grow bigger, and cloud providers report double-digit growth rates, so much so that the technology can no longer be described as emerging. For example, the 2025 AI Index Report by the Stanford Institute for Human-Centered AI estimates that 78% of organisations now use AI – a significant progress compared to 55% in 2023.
We should expect this rapid adoption to translate into higher corporate profits.
This is true for companies providing the picks and shovels to the AI gold rush. Chip manufacturers and cloud providers are generating substantial profits and seeing their valuations skyrocket.
In contrast, the reality is less rosy for companies using AI.
The same AI Index report states, “Most companies that report financial impacts from using AI within a business function estimate the benefits as being at low levels.”
What’s Behind The Lag?
The impact of technology on productivity has been a stimulating field of study in Economics for decades.
This is especially the case for general-purpose technologies whose impact spreads across all industries, geographies, company sizes and so on.
Some researchers argue that AI has all the attributes of a general-purpose technology.
So, what can we learn from previous examples?
In the case of electricity, research revealed a lag of several decades between the adoption of the technology and its impact on productivity.
This lag represented the time needed for organisations to change their processes to leverage the technology.
When new technology is used to support existing processes, the productivity gain is marginal. But when it’s used for creating totally new ways of doing things, the impact can be radical.
When electricity came to factories in the late 19th Century, it was first used to light bulbs so that workers could work an additional couple of hours a day. However, the major increase in productivity only came when Henry Ford redesigned the manufacturing process with the assembly line.
This example shows that when new technology is used to support existing processes, the productivity gain is marginal. But when it’s used for creating totally new ways of doing things, the impact can be radical.
When it comes to AI, companies are still in the “light bulb phase”: tooling their existing processes with new technology without questioning radically what they offer and how they produce.
Whilst there are signs of greater adoption every day, it hasn’t yet reached the point of changing significant industries or work practices.
What Should Companies Do To Harness AI’s True Potential For Value?
1. Embrace the technology to discover radical new possibilities.
Many companies perceive AI as an addition to what they do already: a way to produce faster, cheaper or more, and to engage better or more often with their clients.
This way of thinking exemplifies the light-bulb approach and will only lead to marginal gains.
AI can open new directions, making previously impossible things now accessible.
Discovering these new possibilities requires a fresh approach. Long-standing companies can find it difficult to engage in such openness. They have established processes, operations, clients, and distributors – each of which serves as a reason against making radical changes to the company’s operations.
By comparison, new entrants don’t have such a legacy and can design new ways to do business, leveraging AI capabilities.
One way for established companies to transform radically is to combine technological expertise with small-scale operations developed from their management structure.
Accessing technological expertise can be done through internal development, partnerships with AI companies or acquisitions.
Developing radically new operations or offers usually involves a different talent pool and is sometimes made easier through partnerships.
2. Make strategic decisions about task allocation.
A major question for companies is where to automate and where to augment with AI.
Automation refers to letting AI handle tasks entirely on its own. Augmentation, by contrast, focuses on enhancing human abilities rather than replacing people.
This opens tricky and nuanced conversations that often depend on the context; for example, what kind of job does AI favour?
A recent research project did a systematic review and meta-analysis of 106 experimental studies.
Each study was required to include an original human-participant experiment that evaluated the performance of humans alone, AI alone and human-AI combinations.
The results partly support augmentation as humans performed on average better with the help of AI than alone. However, they show that augmentation is not a silver bullet as the best AI or the best humans still perform better alone than a combination of AI and humans.
The study revealed an important difference in task type: combining AI and human performance is better for creating content than making decisions. This can be explained in two ways:
- Over-reliance – when people rely too much on AI systems without seeking and processing more information.
- Under-reliance– when people ignore AI’s suggestions because of adverse attitudes towards automation.
Other research has shown a significant shift in job roles due to the surge of Generative AI, with a notable decrease in time spent on initial drafting and an increase in time devoted to editing.
Identifying the right balance between AI and human input becomes crucial for maximising efficiency and effectiveness. It’s paramount to identify the tasks that can be automated, the ones that should blend AI and human contribution, and those that should remain managed by humans.
Deciding on task allocation involves taking the technological developments into account and their cost, but also the strategic positioning of the company. Deciding on what AI performs and what is managed by humans is a strategic choice, not a purely technical one.
Last, the direct environmental footprint of AI is massive and growing. Electricity and water consumption from data centres are now competing with consumers’ domestic energy needs. The choice of task allocation should also be driven by the environmental impact of their chosen AI system.
3. Invest equally in technology and people.
Companies should stop regarding AI as a technology problem and consider the human element.
However, getting clear on the type of human expertise they need versus the tasks they would trust AI to perform is difficult.
Comprehensive training programs can help employees understand the nuances of AI, including its limitations and best practices for its effective use.
Securing data and infrastructure is not enough when managers are not trained on AI systems or how to use the results they produce. Investing in training leads to higher productivity of technological investments. Training employees on AI systems and their limitations mitigates over-reliance and under-reliance, maximising the systems’ impact. Moreover, cultivating in-house talent and processes enhances the uniqueness of internal resources, strengthening competitive advantage.
A recent Financial Times article quoted research from Accenture that even though generative AI is expected to account for 15 percent of technology spending this year, fewer than half of the organisations surveyed had increased training on AI fundamentals or technical skills.
Training is becoming increasingly vital, especially to enhance the skills of lower performers. As AI continues to commoditise certain aspects of knowledge work, it’s not just the low performers who are affected; skilled professionals are also at risk.
Comprehensive training programs can help employees understand the nuances of AI, including its limitations and best practices for its effective use. The synergy between technical and human assets is key to realising the full potential of AI. Success in this domain depends not only on redefining processes but also on fostering an organisational culture that understands and trusts AI.
4. Engage with stakeholders in a conversation about the use and conditions of automation and
Significant efforts should be put into ensuring that users understand the system design.
Transparency around the data used, the models implemented, the results’ limits, and the scope of relevance all foster adoption through higher transparency.
What also supports adoption is sharing a common view about task allocation and developing a common sense of the split between automation and augmentation.
More precisely, these discussions could aim to decide on the details of augmentation, for example, to state the right level of transparency according to the context (e.g., facial recognition for smartphone vs radiology) or the objective (deciding versus performing a task).
These conversations should also involve external stakeholders such as clients and partners. Unions and collectives should also be involved in determining the scope and style of automation and augmentation.
This was demonstrated in September 2023 in the US when an agreement was reached between show writers and studios about the use of AI for writing or editing scripts.
Transformation At Scale: The Next Challenge
Exploring how to leverage AI at companies may lead to two very different types of opportunities: “light bulb” ones and “assembly line” ones.
The challenge, then, is twofold for companies.
First, how do they allocate enough resources to scale the “assembly lines” opportunities?
Second, how do they deal with both types of opportunities? Should the two remain, and if so, what is the right balance? Or should the “light bulb” ones disappear and the company concentrate on the “assembly line” ones – meaning a massive reorganisation?
Deciding how to solve the two challenges will be a critical management decision in the next few years.
Louis-David Benyayer
Hao Zhong
