The Rise of the Prompt Engineer
Prompt engineering is a job – ?and a skill
What does it mean to be a prompt engineer?
Frankly, if you've ever typed a message into a conversation with ChatGPT (or Bard, Claude, HuggingChat) and sent it, you're a prompt engineer. Kinda.
Summarizing the responsibilities companies look for, we conclude that a prompt engineer will:
As for the requirements, they're quite technical:
Some of them require knowledge of specific tools, which are quite technical: PromptLayer, Honeyhive, Vellum, Promptitude. One may conclude that prompt engineers are a breed of software developers, data scientists, business analysts, - and won't be wrong. Those are the jobs that language models, esp. with code interpreter capabilities, affect the most. The change is rapid and unstoppable, and if you're 16 and dream of becoming a software engineer, you have to understand that your job won't be much about writing code.
However, those are tech jobs. What about other jobs, like an accountant, procurement manager, customer service agent? Are those going to transform into prompt engineering?
These 3 jobs in particular are at extremely high risk of full automation. The reasons are simple: they are fairly standardized, they work with intangibles (information captured in documents),... For most knowledge workers, prompt engineering won't become their job, but one of the skills they need to learn during major changes in the nature of work.
Let's have a look.?
Specialist to generalist shift
When you have a question you don't know the answer to, who you gonna call?
Ghostbus... I mean, some kind of subject matter expert.
How to do my taxes, how to design a logo, how to write a business plan - you know who to go to get help.
What about the questions you have in your daily job?
How to upload an invoice, where to change master data, and who is responsible for X??
Many of those questions require a certain level of expertise to be answered. And the fact of the matter is, that GPT-4 is close to the human level of expertise. A reasonable question to ask oneself is, is my expertise deep enough, unique and irreplaceable? The reality for most people is that they have a job because there is a human with a certain level of knowledge and skill needed to complete the tasks. But what happens when AI can do those just as good, 100 times faster, and 1000 times cheaper?
It won't be sufficient to simply do one job that consists of several tasks in a sequence. Humans will have to take on tasks that AI is not capable of doing. Currently, and for quite a while will stay so, it is difficult to imagine an AI running a whole project. As in, let's set a clear purpose and common goals for the project, understand what people from the organization would fit into the roles needed for this project, and organized them in a way that fits with the interests of other departments. It is also difficult to integrate across business functions, such as transitioning from sales into order management into delivery into customer service. And those transitions are difficult for AI because they require human-to-human interaction instead of simply creating documents. This gives human workers an advantage while freeing them from the mundane work of rewriting texts that have been written hundreds of times before.
On the flip side, how the hell is one person supposed to fulfill their functional role and be a project manager at the same time? How they are supposed to communicate with a bunch of people, manage orders and requests, and work through customer tickets?
Clearly, this will necessitate special skills and tools. Prompt engineering means being able to make AI do things you want it to. Human has to provide a clear definition of the process to be executed, knowledge base / background information, requirements, decision criteria, and the like, and an integration into multiple systems in file formats.
Prompt engineering means being able to make AI do things you want it to
Knowledge jobs that currently deal creation and processing of documents will expand their borders, requiring more and more responsibilities from the workers supported by additional knowledge and tools. Thus the age of the specialists will be largely over, and general skills, strategic thinking, organizational ability will become even more important. With this, we can transition into the next major change in the nature of work.
领英推荐
Employees as managers
As described above, with AI rising to the level of human expertise for many jobs in the knowledge economy, humans will need to broaden their knowledge and apply organizational skills to stay in the workforce.
A helpful approach for process automation with generative AI is to look for ways to automate small individual tasks, rather than complete jobs or end-to-end processes. AI is simply not good enough yet to succeed in such a feat. Anybody who worked with ChatGPT or other similar tools knows that they can do anything you asked them to, but only on a very superficial level. They tend to run into loops and fail to deliver an output that is beyond a simple regurgitation of other articles.
For example, it can easily assess if your invoice has all the necessary information if you give it a structured template to compare with. To generate the template, you can use company internal guidelines with a mix of legal requirements. This would be a task. However, if you simply ask AI to run your whole accountable process, it will inevitably fail.
Therefore, automating an end-to-end process involves orchestrating the work of dozens and dozens of agents. The human role would be the one of oversight, authorization, and resolution of complex cases:
Sounds familiar? This is exactly what managers do. In our case, the employee becomes similar to a team lead, guiding their agents towards a common goal. The employee would check the work, correct the prompts, teach the system on the complexities, and resolve some interpersonal issues. And while some parts of communication can be automated too, any customer or colleague will prefer an honest e-mail over an AI-generated response - so that stays with the human as well.
Prompt engineering in organizations
So some of you might be wondering, how's that one going to work in companies? Is everybody going to be replaced by someone else who knows prompt engineering?
Scenario 1
While everybody needs to learn how to talk to AI, it's going to be the most critical for engineering professionals such as software developers and data scientists. For these, the nature of work has already changed dramatically. Young people entering universities now, will not know how enterprise software development used to work, because, by the time they finish, the current system will be gone in all companies but laggards that didn't go out of business who knows why. The famed shortage of software engineers will be a thing of the past. And here I'm torn about the board may view of the future. Could be that the automation of software development will allow us to build 10 times more software products but it's not clear if there is demand for it. If not software engineers will be the first job that automated itself out of existence. NB: not completely but rather similar to the next case. ?
Scenario 2
Functional roles, such as the aforementioned accountants and customer support agents, will undergo a massive shift into covering end-to-end processes instead of their narrow areas and become managers of AI agents to complete most of their tasks automatically at scale. In these areas, massive job cuts are to be expected, as one person will be able to do the work of 5 or 10.
Scenario 3
People roles, such as sales, teaching, management aren't likely to go away anytime soon, because while AI can create project plans and assign tasks, it cannot effectively solve organizational issues, interpersonal conflicts, alignment of incentives, and long-term strategic planning. In the case of teaching, AI can create personalized learning plans and check homework, but it cannot replace the guidance a teacher can give in terms of developing a work ethic, team communication, grit, and resilience needed for professional success.
In short, the changes within the next five to seven years will be quite diverse.
After that, prompt engineering will probably fade away and become irrelevant again, as the AI’s capability to understand us will surpass our talent for miscommunication. Plus, tooling for prompt engineering will mature and mostly happen on the back end, without the average user ever editing the prompts.
So when I think about prompt engineering as a skill and a job, I think of it as applicable only for a relatively short transition period. But this transition will define decades of career success for a lot of people, as those that master prompt engineering now will have compounding value effects over the next years by becoming more productive today and staying more productive for the years to come.
Prompt engineering will probably fade away as the AI’s capability to understand us will surpass our talent for miscommunication.
Prompt engineering education
The impact of AI on the job market it's difficult to predict. Goldman Sachs estimates 300 mln job losses. We all know that those predictions aren't worth much, but it is clear that an unprecedented number of jobs will be cut due to automation, and those people will need to work elsewhere and will need to acquire a modern skill set to be attractive employees. There are some articles and courses online, incl. some from Google and Microsoft, but none of them seem very applied or useful on the job.
For me and my team at OSNOVA, this realization became the reason why we expanded our workshop “ChatGPT for Business” into ACADEMY. In particular, “Prompt Engineering for Process Automation” is a workshop aimed at operational employees. Our thought here is simple: today, the employer would invest in educating the people and will make them productive - which is an employer's interest! While the employees gain the skills that give them long-term job security.
If this resonates with you, get in touch. We're looking for companies to run workshops for their teams and partners to help us get the message out.
Entrepreneur | Consultant | Product Manager | Tech addicted early adopter | Long life learner | Edutainment and LXD Curioser | AI follower
1 年Thank you for the article! I found Denis's idea about moving from very specific jobs to more general ones really interesting. We used to ask experts for advice, but now, with things like GPT-4, machines can almost give advice like people. In the tourism industry, where many of my colleagues operate, there's a significant presence of agency firms offering tailored, niche approaches to client needs. Some of their expertise could potentially be replicated by AI, like matching travel services to a client’s preferences, previous purchases, and budget. However, an AI-based transaction loses the personal touch, an external reference component of the purchase process. The travel agent adds a sense of comfort to clients through customer-centric communication. The consultation process is often a much-anticipated and highly valued aspect by clients when planning travel. Plus, when you buy from a person, you feel safer knowing there's someone to call if something goes wrong. Travel agents need to start using AI and the big question is how exactly. Companies need to help their workers get ready for these changes. It'd be great to talk about how AI can be used in the travel business and hear what people think.