Generative AI’s use has exploded across industries, helping people to write, code, brainstorm and more. While the interface couldn’t be simpler — just type some text in the box — mastery of it involves continued use and constant iteration.
GenAI is considered a game-changer, which is why enterprises want to scale it. While users have various resources available, like OpenAI and Gemini, proprietary LLMs and GenAI embedded in applications, companies want to ensure that employees are not compromising sensitive data.
GenAI’s unprecedented rate of adoption has inspired many individuals to seek training on their own, often online at sites such as Coursera, EdX, and Udemy, but employers shouldn’t depend on that. Given the strategic nature of the technology, companies should invest in training for their employees.
A Fast Track To Improving Prompt Engineering Efficacy
Andreas Welsch, founder and chief AI strategist at boutique AI strategy consultancy Intelligence Briefing, advocates starting with a “Community of Multipliers” — early tech adopters who are eager to learn about the latest technology and how to make it useful. These multipliers can teach others in their departments, helping leadership scale the training. Next, he suggests piloting training formats in one business area, gathering feedback and iterating on the concept and delivery. Then, roll it out to the entire organization to maximize utility and impact.
“Despite ChatGPT being available for two years, Generative AI tools are still a new type of application for most business users,” says Welsch. “Prompt engineering training should inspire learners to think and dream big.”
He also believes different kinds of learning environments benefit different types of users. For example, cohort-based online sessions have proven successful for introductory levels of AI literacy while executive training expands the scope from basic prompting to GenAI products.
Advanced training is best conducted in a workshop because the content requires more context and interaction, and the value comes from networking with others and having access to an expert trainer. Advanced training goes deeper into the fundamentals including LLMs, retrieval-augmented generation, vector databases and security risks, for example.
Andreas Welsch, Intelligence Briefing
“Function-specific, tailored workshops and trainings can provide additional level of relevance to learners when the content and examples are put into the audience’s context, for example, using GenAI in marketing,” says Welsch. “Prompting is an important skill to learn at this early stage of GenAI maturity.”
Digital agency Create & Grow, initiated its prompt engineering training with a focus on the basics of generative AI and its applications. Recognizing the diverse skill levels within its team, the company implemented stratified training sessions, beginning with foundational concepts for novices and advancing to complex techniques for experienced members.
“This approach ensures that each team member receives the appropriate level of training, maximizing learning efficiency and application effectiveness,” says Georgi Todorov, founder and CEO of Create & Grow, in an email interview. “Our AI specialists, in collaboration with the HR department, lead the training initiatives. This dual leadership ensures that the technical depth of AI is well-integrated with our overarching employee training programs, aligning with broader company goals and individual development plans.”
The company’s training covers:
-
The basics of AI and language models
-
Principles of prompt design and response analysis
-
Use cases specific to its industry and client requirements
-
Ethical considerations and best practices in AI usage
-
Educational resources including online courses, in-person workshops, and peer-led sessions, and use of resources from leading AI platforms and collaborations with AI experts that keeps training up-to-date and relevant
To gauge individuals’ level of prompt engineering mastery, Create & Grow conducts regular assessments and chooses practical projects that reflect real-world scenarios. These assessments help the company tailor ongoing training and provide targeted support where needed.
“It’s crucial to foster a culture of continuous learning and curiosity. Encouraging team members to experiment with AI tools and share their findings helps demystify the technology and integrate it more deeply into everyday workflows,” says Todorov. “Our commitment to developing prompt engineering expertise is not just about staying competitive; it’s about harnessing the full potential of AI to innovate and improve our client offerings.”
A Different Take
Kelwin Fernandes, cofounder and CEO at AI strategy consulting firm NILG.AI says good prompts are not ambiguous.
“A quick way to improve prompts is to ask the AI model if there’s any ambiguity in the prompt. Then, adjust it accordingly,” says Fernandes in an email interview.
His company defined a basic six-part template for efficient prompting that covers:
-
The role the AI should play (e.g., summarizing, drafting, etc.)
-
The human role or position the AI should imitate
-
A description of the task, being specific and removing any ambiguity
-
A negative prompt stating what the AI cannot do. (E.g., don’t answer if you’re unsure)
-
Any context you have that the AI doesn’t know (E.g., information about the company)
-
The specific task details the AI should solve at this time.
“[W]e do sharing sessions and role plays where team members bring their prompts, with examples that worked and examples that didn’t and we brainstorm how to improve them,” says Fernandes.
At video production company Bonfire Labs, prompt training includes a communal think tank on Google Chat, making knowledge accessible to all. The company also holds staff meetings in which different departments learn foundational skills, such as prompt structure or tool identification.
“This ensures we are constantly cross-skilling and upskilling our people to stay ahead of the game. Our head of emerging technologies also plays an integral role in training and any creative process that requires AI, further improving our best practices,” says Jim Bartel, partner, managing director at Bonfire Labs in an email interview. “We have found that the best people to spearhead prompt training are those who are already masters at what they do, such as our designers and VFX artists. Their expertise in refinement and attention to detail is perfect for prompting.”
Why Developers May Have an Edge
Edward Tian, CEO at GPTZero believes prompt engineering begins with gaining an understanding of the various language models, including ChatGPT, GPT-2, GPT-3, GPT-4, and LLaMA.
“It’s also important to have a background in coding and an understanding of NLP, but people often have minimal knowledge about the different language models,” says Tian. “Understanding how their learning concepts work and how they are structured can help significantly with prompt engineering. Working with pre-trained models can also help prompt engineers really hone their skills and [gain] a further understanding of how it all works.”
Chris Beavis, partner and AI Specialist at design-led consultancy The Frameworks suggests using the OpenAI development portal versus ChatGPT or Gemini, for example.
“It offers a greater level of control and access to different models. The temperature of a model is particularly important, allowing you to flex the randomness [and] creativity of answers over determinism [or] repeatability,” says Beavis in an email interview.
Chris Beavis, The Frameworks
“[The user] should start by identifying an idea or a challenge they are facing to see what impact AI can have. Try out different approaches, remember to give specific instructions, provide examples, and be clear about the format of the result you are expecting. Some other tips include breaking problems down into steps, including relevant data sets for context and prompting the AI to ask you questions about your request if it’s not clear.”
Most employees are experimenting with AI at The Frameworks in different ways, from image generation and summarization to more advanced techniques like augmented information retrieval and model training.
“I certainly think there is an initial barrier to overcome [when] familiarizing yourself with how to prompt, which may suggest the need for a beginner level of training. Beyond that, I think it’s a learning journey that will depend on your area of interest. A developer may want to explore how to connect AI prompting to data sets via APIs, copywriters may want to use it for brainstorming or drafting and strategists may want to use it to interrogate complex data sets. It’s a digital literacy question.”
His company is finding the most useful applications are where they use code to combine prompts with data sets, like mail merging. That way, AI can be treated as a step in a repeatable problem-solving process.
“As with most companies, we started by simply seeing what the technology could do,” says Beavis. “As we become more familiar with the capabilities, we are finding interesting uses within client projects and our own internal processes.”
Intelligence Briefing’s Welsch says for software developers, mastery is a cost function such as getting the optimal output with the shortest possible prompt (to consume the least amount of tokens). For business users, he says proficiency could be measured by awareness of common prompting techniques and frameworks.
“Prompting is often portrayed as a glorified science. While teaching techniques is a good start for laying a foundation, Generative AI requires users to think differently and use software differently,” says Welsch. “[Trainees] can learn about examples of what these tools can be used for, but it is experimenting and iterating over an open-ended conversation that they should take away from it.”
Engage Specialized Trainers
Brendan Gutierrez McDonnell, a partner at K&L Gates in the law firm’s AI solutions group, says his company uses a multifaced approach to prompt engineering training.
“We have relied on experiential training provider AltaClaro’s prompt engineering course as an introduction for our lawyers and allied professionals to the world of prompt engineering. We have supplemented that foundational training with prompt engineering courses tailored to the GenAI [and other] AI solutions that our firm has licensed,” says McDonnell in an email interview. “These more tailored programs have been conducted in tandem by the vendor providing the solution and by our internal community of power users familiar with the specific solution.”
At present, the firm is building its own internal database of prompt engineering questions that work well with the various GenAI solutions. Over time, he expects the solutions themselves will recommend the best prompt engineering guidance to solve a particular problem.
“The best way to develop a degree of mastery is through education from outside educational vendors like AltaClaro, solution vendors like Thomson Reuters, and by learning from your colleagues,” says McDonnell. “Prompt engineering is best approached as a team sport. Most importantly, you must dive in and use the program. Be creative and push your own limits and the program’s limits.”
Brendan Gutierrez McDonnell, K&L Gates
K&L Gates has training programs for beginners that cover the basics and nuanced programs for advanced users, but before jumping into prompt engineering, he believes the user should have a fundamental understanding of how a GenAI solution works and whether the information input into the program will remain confidential or not.
“The user [should] understand that the output needs to be verified as large language models can make mistakes. Finally, the user needs to know how to vet the output. Once the user has these basics in order, she or he can start to learn how to prompt,” says McDonnell. “The user should be given problems to solve so that the user can put his or her prompting to the test and then review the results with peers. Having a training partner like AltaClaro can make sure that the training experience is effective, as they are experts in building programs tailored to the way lawyers learn best.”
Bottom Line
Organizations are approaching GenAI training differently, but they tend to agree it’s necessary to jumpstart better prompting.
Where to get that training varies, and the sources are not mutually exclusive. One can hire expert help on-site, create their own programs and invest in GenAI online courses depending on the level of existing knowledge and the need to provide training that advances GenAI proficiency at varying levels of mastery.