With the meteoric ascent of AI in the past 5 years, everyone, from CEOs helming corporate giants to team managers striving for efficiency to individuals hungry for more hours in their day, is grappling with a pivotal question: how can AI help me do more with the hours in the day?
Everyone has the same 24 hours in a day. We dedicate 8 of these to work and surrender another 8 to sleep. Slicing our time into thirds leaves precious few hours for everything else in our lives. So it is not surprising that productivity is a big topic on everyone’s minds.
What Does Productivity Actually Mean?
Economists and politicians will focus on national outputs, and the dictionary definition is aligned with this. The rate at which a company or country makes goods. This view can be codified as:
Labour Productivity = Output / Hours Worked
Imagine an office clerk at the HM Passport Office racing against the clock to process a mountain of passport applications, or a Starbucks barista hurriedly crafting caramel macchiato Frappuccino’s by the dozen. This perspective measures productivity in tangible outputs: How much can you achieve within a given timeframe?
But companies cannot limit their view of productivity to output alone. Direct output can be difficult to measure in many cases. So, indirect measures are used:
- Financial metrics like labour or project costs offer a $-sign perspective on productivity: How efficiently is your money translating into results?
- Customer satisfaction and Net Promoter Scores (NPS) reflect productivity through the prism of consumer approval: Are we delivering value that resonates with our customers?
- Operational benchmarks, such as the average time to move from raw materials to shipped products, gauge productivity in the flow of production cycles: How swiftly do we turn concepts into commodities?
- Quality scores focus not just on volume but also on the quality of the service or output being offered.
These alternate vantage points underscore a broader understanding of productivity, particularly where it is difficult to measure the exact desired output and for organisations that understand that output isn’t everything.
Early Indicators Show That AI Is Positively Impacting Productivity
We are seeing numerous early indicators of increased productivity as a result of AI, across a wide range of different professions.
Consultants Are Faster With AI
Early indicators show that consultant productivity can benefit significantly from AI. 758 consultants were divided into three groups: those with no AI access, those with GPT-4 AI access, and those with GPT-4 AI access, with an overview of prompt engineering. Some of the tasks they needed to complete were easily done with AI. On measuring their ability to carry out 18 realistic consulting tasks, consultants were found to be able to complete 12.2% more tasks on average, and they completed their tasks 25.1% more quickly. Additionally, these were found to produce significantly higher-quality results (more than 40% higher quality compared to those without AI). This was also found to benefit consultants across different skill levels. Those below the average performance levels saw performance increases of 43%, and those who were above the average performance level for consultants saw increases of 17% compared to their performance without AI.
Other studies have shown that knowledge freelancers on Upwork are already being impacted by the rollout of AI assistants. Across the board, it was found that freelancers who offer services in occupations most affected by AI experienced reductions in both employment and earnings. The release of ChatGPT led to a 2% drop in the number of jobs available on the platform and a 5.2% drop in monthly earnings. Interestingly, it was found that offering high-quality services did not mitigate the negative effect of AI on freelancers, and in fact, early indications suggest that top employees are disproportionately hurt by AI.
Coders Are Utilising AI Assistants as Part of Their Workflows
Software engineers, data scientists, and coders are already adopting AI into their workflows to improve their own productivity. Products such as ChatGPT and Claude are being used to help developers ideate and solve complex problems, where previously they would have had to spend time carrying out manual research on websites like Stack Overflow. This isn’t just hearsay but is starting to reflect in the data, as highlighted by an analysis carried out by machine learning engineer Ayhan Fuat Celik, who looked at the data shared by Stack Overflow. He visualised this data and found a definite fall in all metrics: page views, visits, questions asked, and votes.
This analysis found that Stack Overflow lost about 50% of its traffic. However, the traffic data turned out to not account for a Google Analytics change. Accounting for this, the drop would be around 35%. Still, the most striking insight from the data is not traffic but the decrease in questions asked and upvotes.
Stack Overflow highlighted that they saw an average traffic decrease of ~15% in April 2023, a month after the release of GPT-4 in March 2023. This is ironic, as many of the AI assistants being used have been trained on data from the likes of Stack Overflow, which has the world’s richest programming Q&A training data. In July 2023, the CEO of Stack Overflow, Prashanth Chandrasekar, announced OverflowAI, a set of AI solutions including the integration of generative AI into their website, the integration of Stack Overflow in Microsoft Teams, and an IDE integration to bring Stack Overflow inside the place where developers get their coding done. Stack Overflow also predicts that this surge in usage of AI tools will simply result in new problems to be solved and new questions to be asked on their platform. They are betting on the quality of content, trust in content, and the power of community and human beings to create and curate content.
GitHub Copilot is an example of a more direct, integrated tool for coders, autosuggesting entire blocks of code and learning through the massive repository of code already on GitHub. This digital coding partner is already shown to significantly reduce the time spent on coding tasks, with developers writing code up to 55% faster than they could without AI.
General Purpose AI Is Estimated to Impact All Roles
A large-scale study by OpenAI found that approximately 15% of all worker tasks in the US could be completed in at least half the time at the same level of quality, provided they have access to a Large Language Model (LLM), and this increased to up to 56% of all tasks when incorporating software built on top of LLMs. 80% of the U.S. workforce could have at least 10% of their work tasks affected by the introduction of GPTs, while around 19% of workers may see at least 50% of their tasks impacted.
These results may be exciting, but they will also likely have substantial economic and societal ramifications. Business leaders will need to carefully examine how to adapt and whether to adopt these technologies, and this presents different choices to be made. AI threatens to indiscriminately disrupt incumbents and erode their competitive advantage, and so having a framework for making these decisions is vital.
Choices to Be Made
If a coder’s productivity surges by 90%, it prompts a strategic crossroads: does this lead to reducing the workforce by a third, or does it pave the way to ramp up digital output by 90%? Yet, the strategic implications extend far beyond this binary choice. The technology does not determine the path taken. Leaders have the power to make different choices on how to adopt this technology, and there are many different approaches they could take:
- Cost Cutting: New operating models, ways of working, working patterns, and the cutting back on legacy functions can support monetary savings, and some organisations will choose to absorb some or all of these savings in their bottom line.
- Cost Reallocation: Many organisations will choose to reallocate resources to growing areas of their business with higher returns, a complex transition that will likely involve reskilling or upskilling initiatives for those staff who are staying, firing staff who are no longer needed, and a whole array of new working practices, operating models, and investments.
- Value Expansion: This approach is about using the technology to do even more without significant reductions in existing activities. This enables diversification of offerings and an opportunity to tailor service to better meet evolving market needs.
For leaders, the integration of AI into their operations is not a simple equation of productivity gains; it’s a menu of strategic choices that balance efficiency, innovation, ethical considerations, and long-term sustainability. Additionally, most leaders will be aware of the hard truth that most technological transformations are much more painful than hoped. It isn’t as simple as rolling out a new technology and waiting for the productivity benefits. This fact will play a significant part in the choices the leader ultimately makes.