Will Generative AI transform our relationship with time and quality: Toward a new definition of productivity?
Intro
Since my last article New ways of working with generative AI: toward an extended or augmented yourself's capability where I shared some personal use cases, the red lines I set as usage guidelines, and remaining questions about #ChatGPT, some new #GenerativeAI tornadoes have already been announced whether they started in big tech shops or in startups.
I'm seeing new use cases almost every day, some of them ridiculous, but others really interesting. Over the last weeks, I have had the opportunity to discuss this with colleagues, partners, and customers and we really feel the excitement about the potential new ways of working, especially directly in some business areas such as Legal and HR who dream of having their own document reference system made up of regulatory and internal/temporary documents.
At the same time, we are at a moment where the economy, and the #tech #economy in particular, is heavily shaken for a lot of different reasons. It is impossible not to read or listen to discussions where the topics of #efficiency, #competitivity, and #productivity are in the spotlight, given the recent layoffs movement that happened in the last weeks.
As we are in front of something that will definitively change our ways of working but also how the information can be accessed and used more widely, I'm questioning if this new Generative AI move may lead to refining the notion of productivity, as few think, in a wrong way from my point of view, that it may replace jobs 1-for-1, drive us to the 10x developer phenomenon or divide by x the time to complete tasks.
I recall that productivity can be defined as the amount of what can be produced per unit of time, per worker, per machine, or per unit of capital invested. Most of the time, the goal of productivity improvement is to produce more output using the same amount of inputs or to use fewer inputs to produce the same level of output.
I'm curious to observe what will be our relationship with #time in such a context: compressed, equal, or extended. And what about the inner #quality of inputs and outputs, and what we will have to set to measure and validate those?
Interviews
I've prepared 3 questions to try to clarify my ideas and see what the results might be in my technical business area:
- Question 1: What kind of impact on productivity can we expect from the last generative AI movement? (i.e. what are the areas where you can foresee productivity impact)
- Question 2: Do you think that our relationship with time will change thanks to those assistants? (i.e. is it necessarily obvious that time will be compressed?)
- Question 3: What could it change in our daily activities? and what are the potential impacts on the quality of what we deliver? (i.e. in the same amount of time, could we expect to be more qualitative on what we will produce)
Then I asked some friends, colleagues, or peers to answer:
- Marie Crappe - Data strategy & implementation expert and consultant
- Didier Girard - co-CEO at SFEIR
- Jason Gulledge - VP of Engineering at BackMarket
- Mick Levy - Business Innovation Director at Business & Decision
- Liam Randall - CEO Cosmonic, Wasm Day Chair, Serial Open Source Entrepreneur, Founder wasmCloud, Investor - Stacklet, Kolide, Bro/Zeek, Horizon3.io
- Alexis Richardson - CEO at Weaveworks
- Sylvain Wallez - Principal Software Engineer at Elastic
Please note that the full answers are at the end of the article, in the appendix, for clarity.
Outcomes
Before I share my personal views on these questions in the Outro section, here is what I gleaned from the answers I received from the people I interviewed, quoting some, on purpose.
Generative AI and Productivity
At a time when our economy is looking for a breath of fresh air in terms of competitiveness, profitability, and performance, I can't believe that no one has imagined that this new and large-scale tidal wave of generative AI could not optimize things in business, potentially improving productivity at different stages of the lifecycle.
On this first question, respondents agreed that ChatGPT and other Generative AI tools have some potential to impact productivity across various businesses and jobs, and clearly in fields where content, data, and software are predominant. However, some contributors argue that caution is needed in terms of expected gains, which may initially be more at the micro level, but at the additional cost of the click workers whose "data worker" work could be even more necessary.
"In the short term, it will be difficult to be more productive than someone who uses an AI as an assistant" - D. Girard
"I think the biggest impact will be on idea generation and scaffolding, e.g. laying out the foundations of a software module" - S. Wallez
It's quite interesting to observe how much Generative AI is associated with ideation, creativity, and scaffolding processes.
"The reduction of cognitive load on engineers is also a significant benefit, allowing them to focus on high-level tasks." - A.Richardson
The idea of an extended or augmented version of yourself through a better together partnership between AI and people is predominant as well.
"Generative AI is poised to be a better together partnership between AI and people" - L. Randall
However, it looks to all important to use Generative AI responsibly and in conjunction with oversight from experts and, thoroughly test and validate before being put into "production", but also in a continuous way if we want to correct the appearance of biases on the duration of use.
Generative AI and relationship with time
On the second question, the contributions suggest that while the use of generative AI could optimize time, it is important to use it with caution, it requires investing time in reviewing and editing the generated content to avoid errors. They warn that the use of generative AI should be tempered with caution, as it may produce low-quality or incorrect content. With generative AI, if you reduce the time to produce by two, but multiply the time to review by two, the total doesn't really change! And even worse with the devil's advocates, who might point out that the time could be more than doubled because we might take longer to convince ourselves of the reality of the facts after having had doubts.
"Just like the internet is full of fake news and low-quality content, generative AIs should be used with caution, thus requiring time for prompting, reviewing, and editing" - M. Crappe
An interesting idea, which was mainly shared, is that the time freed up may also be filled with other activities that are not done enough or less, and it may take some time to learn how to use the technology effectively.
"In the end, this freed up time to do other things, other jobs were born." - D. Girard
For some, the use of Generative AI will lead to a shift in time allocation, with more time spent reviewing and analyzing AI-generated material and less time spent brainstorming. Put another way, it's more about cost-saving than time-saving.
"If you're a curious person and you invest your time and mental energy into it to try to learn from it, you'll get quite a lot out of it. If you just want to casually consume it, you might not learn much, but it will serve your needs." - J. Gulledge
In addition, Generative AI has the potential to dramatically increase the accessibility of content creation, similar to what Google has done for access to information. Remember when some of us were skeptical about the results of that query, which returned answers faster than queries from their own databases?
"With generative AI, once the art of prompting is mastered, content production is within everyone's reach." - M. Levy
The emergence of Generative AI is part of a broader trend in human history toward increased productivity and time efficiency. They argue that while there may be some adjustment required, the overall effect will be positive and lead to new opportunities and areas of growth. Some of them highlighted that there is a risk of over-reliance on AI-generated content, which can potentially lead to a lack of creative variation or content standardization.
"As our assistants emerge we run the risk of overwhelming ourselves with the output of generative models, of relying upon common and undifferentiated models, and for early adopters the challenge of balancing the partnerships between generative AI and creative professionals correctly." - L. Randall
Finally, some put the focus on the fact that Generative AI could lead to increased knowledge acquisition for curious individuals, but could make lazy learners even less interested in learning; but hey, we are talking about humans after all!
Generative AI and relationship with quality
Regarding the third question, the contributions suggest that if generative AI has the potential to increase productivity, it's more challenging in terms of quality.
Even if generative AI has the potential to automate repetitive tasks and increase the quantity of what we produce, it also has downsides and can produce low-quality content.
"I don't think there's a way to predict the quality of the product coming out of Generative AI." - J. Gulledge
This is where we can really feel how much the improvement in quality is linked to the level of skills, even if its apparent ease of use makes it feel accessible to all, and the skills it requires to imagine a serious boost. So the impact of generative AI on quality depends on how well the user masters the tool, and whether they use the time saved to deliver more or to improve quality.
"The question is: will you use the time saved to deliver more, or to deliver the same amount of work but with higher quality standards?" - M. Crappe
The use of generative AI raises questions about the future of work, and more precisely the way we could work and the related impacts.
"Generative AI does not come without downsides - oftentimes it is the very struggle of the creative process where the most fundamental insights are delivered" - L. Randall
"But, the effectiveness of generative AI in improving the quality of what we deliver also depends on how well the engineer understands the technology problems and effectively interacts with the AI model." - A. Richardson
Recommended by LinkedIn
"Developing analysis skills requires some level of seniority. How will people be able to reach that level if the junior-level tasks are mostly automated?" - S. Wallez
Outro
So, like everyone else who's contributed their views, I'm really convinced that we're on the verge of a major evolution in the way we work, one that will see us doing things differently, and where Amara's Law could certainly apply!
"We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run" (from Roy Amara, past President of The Institue for the Future).
And because the devil is in the detail, I expect that the productivity lever will be hidden behind the word "differently". Even if it's not obvious to quantify the potential level of productivity gain, we can imagine a different linear distribution of tasks and activities in this new human-AI tandem. Generative AI technology has the potential to increase productivity for those who master it, but it can also widen the productivity gap with those who don't. There are also some warnings about the risks of over-reliance on AI-generated content and the potential for reducing creative variation. As time is a key parameter in the productivity equation, it will be interesting to observe the extent to which there should be a rebalancing, as AI-generated content may take time for humans to analyze, understand and use efficiently, with the expertise of practitioners certainly increasing to drive the process. I don't think the improvement in quality can be taken for granted. I see it more as a consequence than a precondition, and there are several parameters at stake.
In conclusion, I have the feeling that we're in the midst of a global live experiment in which we're all guinea pigs!
As everyone becomes more 'hands-on', we are seeing a new equation being established with productivity, time, and quality as the key parameters for success in moving from a technology that makes a buzz to a technology that makes businesses win. If time is not saved, the potential will not be realized. If production quality is not maintained at a minimum level, it will grind to a halt.
As we better understand the transformative potential of this technology, we also understand the implications for skills, the risk of uniformity, and our relationship with creativity.
Here, we talked about generative AI that generates access to open-world knowledge and reformats this knowledge in response to questions. To work at scale and in a rather exhaustive way, it is based on a huge volume of data and that is why it impresses everyone. But the daily productivity of a company depends relatively on the global transverse knowledge base: a company has a lot, and especially, needs to manage the documentary corpus that belongs to it and that is, a priori, private. What will be the capacity of these generative AI to efficiently process this data in relative volumes? And above all, what about the secure cross-referencing of open data with this corporate data?
Let's let some time pass, but here is already an idea for a future article! In the meantime, all these are the key elements that I will use as probes in my daily watch to analyze and understand all this.
Appendix
I'm sharing here the accurate and comprehensive answers from all the contributors, in surname order:
- Marie Crappe - Data strategy & implementation expert and consultant
Q1: "I won't give a full list of impacted job areas, as many articles already cover that, and will instead focus on the areas I know best. First, I can foresee productivity impacts on data exploration and basic analysis. So many tools on the market today are trying to bridge the gap between business users and the big amounts of data they're sitting on. After testing the new capabilities of Python code and SQL query generation, this looks quite promising. More generally, I strongly believe in the code generation potential, especially for bootstrapping projects. The other thing that comes to my mind is a huge productivity boost in administrative documentation generation. E.g. as a University teacher, I love spending hours designing activities and preparing anecdotes, but compiling documents like the syllabus or the abstract always feels like a waste of time. Those assistants do that super well."
Q2: "The way I see it, we will just expect some things to be achieved faster than before, but I'm pretty sure the freed-up time will get filled by other activities straight away! Moreover, you'll still need to invest a significant amount of time reviewing the generated content, especially as it can hold big mistakes! I had chatGPT tell me, assertively, that Clermont-Ferrand was connected to Paris with a 2.5h high-speed train. It also gave me "real-world examples" of data governance failures that were completely fake. Just like the internet is full of fake news and low-quality content, generative AIs should be used with caution, thus requiring time for prompting, reviewing, and editing."
Q3: "Generative AIs are another type of technology that can benefit those who will master it, widening the productivity gap with those who won't. I see the AI assistants a bit like I see the IDEs for software development. If you have a nice IDE and a fairly good mastery of its features and shortcuts, you will definitely be more productive than someone who doesn't. The question is: will you use the time saved to deliver more, or to deliver the same amount of work but with higher quality standards? I like to think it will be a bit of both!"
- Didier Girard - co-CEO at SFEIR
Q1: "The impact on the productivity of certain tasks will be significant. The scope remains to be defined, but we can already cite content writing (meeting minutes, content creation, recruitment announcements, test writing, production of GUI skeletons, production of boiler code, etc.), in the field of graphic arts (production of visuals for presentations, production of characters for video games, production of 3D animation), in the field of ideation and brainstorming. The field of application is very vast, not a day goes by without a new application of these tools appearing. In the short term, it will be difficult to be more productive than someone who uses an AI as an assistant."
Q2: "We have always tried to improve productivity. In his childhood, it took my grandfather's family several months to make hay. A few decades later, everything was done in a few days. In the end, this freed up time to do other things, other jobs were born. The arrival of these new tools is in the same vein. It is both an upheaval in our daily lives and at the same time a process that has been going on for centuries."
Q3: "I don't see any direct impact on quality. As always it is the individuals who are the masters of the tools, as each new tool comes along, it will be possible to do wonderful things as well as sloppy things. For example, sanding machines allow us to produce nice surfaces for furniture very quickly, but sanding machines do not make us carpenters. The same will be true for AI assistants."
- Jason Gulledge - VP Engineering at BackMarket
Q1: "I don't think we know yet how large the impact could be. Between ChatGPT and Constitutional AI (anthropic), There's a lot. Some areas I think we'll see the impact on are: 1. Advertising / Marketing / GTM strategies outlining frameworks for generating ABM campaigns. You can feed chatGPT a LOT of data (from your CRM, hubspot, etc) and have it identify LTV activities to pinpoint accounts that should be marketed to. It can identify buying cycles, and can tell you the best buying windows for accounts. Generating predictive algorithms from being fed historical data (and it will already do this *as well as* generating the python/R code you need). helping you in other areas related to ad spending. You can give it paid ad spend data, competitor data, competitor reviews, related online articles, and other social data. 2. Content Generation - I think this is what chatGPT does well but it's honestly just the tip of the iceberg of the power it will bring in terms of productivity. 3. Learning. It's one step closer to "plugging your brain in" like it was done in the Matrix. You can quickly learn with generative AI. You ask it questions, it distills a TON of information and will answer your questions. Also like stackoverflow but not just for engineering topics. ANY topic. 4. Customer Support - you'll be able to instantly train generative AI on your entire knowledge base to help answer customers' questions. 5. Anomaly detection / Data Quality issue detection - Numbers appear in data in a specific ratio. I think generative AI will be able to automate data quality issues, and be able to identify anomalies, and perhaps equal to the level of quality you get in anomaly detection you get with topic modeling. 6. Investing. Generative AI attempts to predict human behavior. This will very likely be leveraged to make investing decisions. and possibly in an automated way. 7. ??????? - I think there may be an unending list of areas where major productivity improvements will be made."
Q2: "For me, not in the short term, but perhaps in the long term. ChatGPT has really made me curious, and I find myself spending more time working on it just to learn what it can do. But I think for mankind, it definitely will have a huge impact in terms of how quickly someone can level up. On the flip side though, for people who are uninitiated, chatGPT will make the lazy even less interested in learning. They'll lean too heavily on chatGPT's ability to generate content, or "give them an answer", and they may just use the answer without considering whatever topic they're looking at any deeper. I suppose, like with any tool, you get out of it what you put in. If you're a curious person and you invest your time and mental energy into it to try to learn from it, you'll get quite a lot out of it. If you just want to casually consume it, you might not learn much, but it will serve your needs."
Q3: "1. The price of "BS" will go to $0. People will use generative AI to create content, and some of it will not be good. ChatGPT, for instance, will use fake scholarly citations when giving an answer. That's not good. 2. People who don't know excel can just paste data into chatGPT to generate the formulas they want. 3. Marketing made easy. 4. Lots of business operations/decisions can be made easy by feeding chatGPT with data. 5. Prompt hacks - People will come up with prompts to make chatGPT super useful for their needs. I don't think there's a way to predict the quality of the product coming out of generative AI. I think the "GIGO" rule applies here. We'll have to trust humans to use it responsibly, which won't happen universally."
- Mick Levy - Business Innovation Director at Business & Decision
Q1: "I'm going to go against the grain here, but I think we need to be cautious about the impact on overall business productivity. Several studies show that the promise of productivity gains from technology is rarely fulfilled... And I'm not sure that ChatGPT is an exception. What will change is at the micro level, for the individuals themselves. With ChatGPT, the speed and modality of access to information, as well as the availability of powerful tools to a wide population, could change the way a number of jobs are done."
Q2: "Most certainly. At the end of the 90s, Google made it possible to divide the time it takes to access information by 1000. Before Google, you had to go to a library and do a painstaking search, accompanied by a librarian. After Google, everything is just a click away. The same could happen for the production of any content with ChatGPT, and more globally with generative AI. Before generative AI, the production time can be long, and, above all, the skills are not accessible to everyone. With generative AI, once the art of prompting is mastered, content production is within everyone's reach."
Q3: "Our business will change because we have a new tool, that's all! This tool is powerful and opens up new possibilities for everyone. The global quality of production could increase if users make the effort to review, complete, and humanise the content they produce. Without this, the overall quality of human production will only decline. As with the arrival of any tool, some knowledge and skills are required to use it correctly... even if its apparent ease of use makes it accessible to all."
- Liam Randall - CEO Cosmonic, Wasm Day Chair, Serial Open Source Entrepreneur, Founder wasmCloud, Investor - Stacklet, Kolide, Bro/Zeek, Horizon3.io
Q1: "Generative AI is poised to accelerate and increase the velocity for those creative workers that invest the time to partner with their software solutions. Highly creative tasks such as writing, architecture, or developing software are all very nuanced and situational in how a valid solution to a problem may be more or less appropriate. Generative AI is poised to be a better together partnership between AI and people - generative AI quickly roughs out a set of possible solutions that are then further shaped and refined to meet the requirements of the specific implementation."
Q2: "In the near future running your business without generative AI will sound as crazy as running your business without electricity, computers, or the internet. The early adopters of these technologies will absolutely experience productivity gains and have competitive advantages, however, these advances will not come without a price. As our assistants emerge we run the risk of overwhelming ourselves with the output of generative models, of relying upon common and undifferentiated models, and for early adopters the challenge of balancing the partnerships between generative AI and creative professionals correctly."
Q3: "Generative AI will certainly increase the quality of what we produce across a huge range of creative fields. It is poised to make certain common repetitive tasks, such as writing documentation, obsolete and, when used judiciously, is already today accelerating common workflows, such as the writing of these comments. Generative AI does not come without downsides - oftentimes it is the very struggle of the creative process where the most fundamental insights are delivered."
- Alexis Richardson - CEO at Weaveworks
Q1: "The last generative AI movement has the potential to result in significant productivity gains in the field of software engineering. By leveraging AI to automate many manual and repetitive tasks, engineers can focus on more critical and complex tasks, leading to more efficient and effective software development. Our preliminary observations suggest that the productivity gain from using AI in software development is at least three times higher than traditional software development methods. The reduction of cognitive load on engineers is also a significant benefit, allowing them to focus on high-level tasks. Generative AI also has the potential to play a significant role in operational automation by generating new scripts, workflows, and processes that automate tasks. However, it must be used responsibly and in conjunction with engineers' oversight, as the outputs generated by generative AI systems must be thoroughly tested and validated before being put into production."
Q2: "The use of assistants powered by generative AI has the potential to significantly reduce the time required for certain tasks in software development. However, it is not necessarily obvious that time will be compressed for all tasks. While certain tasks may be completed more quickly, others may require more time and resources to train the AI or review its output."
Q3: "The use of generative AI in software development can significantly change how engineers work, impacting daily activities and potentially leading to higher-quality products. As previously mentioned, AI-powered assistants can offload repetitive and time-consuming tasks, freeing up more time for creative problem-solving and high-level tasks. But, the effectiveness of generative AI in improving the quality of what we deliver also depends on how well the engineer understands the technology problems and effectively interacts with the AI model."
- Sylvain Wallez - Principal Software Engineer at Elastic
Q1: "From my usage of it so far, I think the biggest impact will be on idea generation and scaffolding, e.g. laying out the foundations of a software module. Some businesses and jobs will be impacted, including those that don't require expert skills. Some examples may be 99designs (low-cost logo design services), and the job of a lot of lesser-skilled developers coming out of boot camps. Idea generation will change from brainstorming and sketch boarding to choosing among a large number of AI-generated proposals. We will see "prompt experts" who will master the craft of talking to AI to refine its production in a controlled manner. Scaffolding in the area of software development touches almost everything: application code, test generation, and even documentation. I have first-hand experience with GitHub's Copilot generating entire paragraphs of documentation that were surprisingly good."
Q2: "The distribution of time will change. We will spend more time reviewing and analyzing the AI-generated material, and less time on brainstorming and scaffolding. Despite this shift in activities, the net balance will certainly be positive. And if it's not, people will stop using AI generation. It's interesting BTW to see how brainstorming and scaffolding are on the extreme ends of the creativity spectrum. This comes with a danger though, as it can potentially lead to a lack of creative variation if idea generation relies too much on AI. This may also be exacerbated in a mid-term future when AI engines will be trained with hybrid content that was bootstrapped by an AI and refined by humans."
Q3: "The use of AI may backfire in areas where precision is required, e.g. software development or legal documents. A tool like chatGPT can produce content that is very wrong but with a style that is very convincing. The critical analysis of this content is hard as it requires looking past the assertive tone of AI-generated content, and strong expertise to find mistakes or blatantly invented facts. In other words, an AI has no malice but can easily fool you. In software, a recent study has shown that AI can produce code with more security issues than a human. So, as stated previously, this will require fewer less-skilled people and more people who can perform this critical analysis. This however raises a question: developing analysis skills requires some level of seniority. How will people be able to reach that level if the junior-level tasks are mostly automated?"
Also published on Medium
Principal Software Engineer at Elastic
1yThanks Philippe for inviting me in this panel! It's interesting to see people with quite different backgrounds ending up having relatively similar views on this topic!
Head of Strategic Planning at Orange Business
1yNice piece of generative human brain! 💡
VP jobline ITN&Data AI @Orange | Expertise in Telecoms |MBA | Change management | Mentor | Start-up Entrepreneur
1yProud to be part of the reviewers, thanks Philippe Ensarguet for adding me to the group! This disruptive topic might change our lives in the coming months/years. It appears precious to exchange, and gather opinions, and visions from each other! it's worth the 20min read!
🔥 On parle Data & iA ? 🎤 Conférencier | 📚 Auteur | 💡 Directeur Stratégie & Innovation chez Orange Business
1yGlad to have been part of this collaborative article! Thanks Philippe for the initiative, the result is rich of interesting thoughts and information 💡