Since the public launch of ChatGPT in November last year, the generative AI-powered chatbot become the online service that has reached fast adoption worldwide. To put into perspective, Netflix, launched in 1999, took 3.5 years to reach one million users. Dropbox, launched in 2008, took 7 months. ChatGPT took 5 days in 2022.[1] No wonder generative AI and related issues were discussed in every session I attended at Collision Conference last month in Toronto, Canada. Technology and business leaders, regulators, and other opinion leaders shared equal amount of enthusiasm and concerns.
One undeniable fact with generative AI is that the technology will continue to develop quickly and will change the way we work and think. As such, I want to better understand how this emerging technology is affecting my profession and what my teams need to stay competitive. Since ChatGPT is often billed as a co-pilot to its users, I want to share with you a few things I have learned to work with this “smart hire.” What do I need to know as a manager to get the best of out of this new team member? By the way, this framing of ChatGPT’s role in a workplace is suggested by Professor Karim Lakhani at Harvard Business School, in the first episode of a special four-part series on generative AI (episode 909) on HBR Ideacast.
First, we should understand the fundamentals and limitations of ChatGPT as a tool:
- ChatGPT is built from a large language model. It is very good at digesting a large volume of text quickly that only a machine can do. The responses it generates are based on existing text that it has digested. ChatGPT will readily acknowledge that its “knowledge cut-off in September 2021,” meaning, all references used are on or prior to that time.
- To circumvent that limitation, you can use the version integrated with Microsoft’s Bing. Or there are now new plug-ins that can be added to other browsers, such as Chrome to support the search function. These features enable ChatGPT to search the web and to provide some attribution to information sources. Many users have experienced ChatGPT’s hallucination, meaning the machine making up responses. As such, human fact-check is still paramount.
- Given ChatGPT is built on a large language model, it is good at predicting the next word, phrase, and paragraph based on existing content in the public domain and additional learning from users’ feedback. That’s why it can produce reasonable template-like documents for many disciplines. That does not mean you can expect original ideas from the machine instantly. And again, you should still check the work.
That third point often sparks some of the most contentious debates. Complex and important topics such as copyright, transparency of how the content is generated, or potential impact on the cost of content creation in the future, warrant deeper and more extensive discussions on another occasion.
These bigger concerns are being addressed by our colleagues at Omnicom. In June, we announced that Omnicom would be one of the first companies to get direct access to the Open AI algorithms through Microsoft cognitive services. This partnership has allowed us to set up sandbox environments to enable our colleagues and clients to use powerful Generative AI tools that avoid the legal pitfalls that come with the publicly available services. We have set up a multidisciplinary group with representation from agencies, IT, security, training and development, and legal to make this a reality.
Meanwhile, in the spirit of learning and curiosity, I want to experiment with the technology to know what kind of efficiencies and scale it can offer and if it aids or stifles creativity. I used the free version of ChatGPT 3.5. Some of these examples may seem rudimentary to avid users of ChatGPT. But for those new to the adoption, I hope these discoveries may be useful.
Light-lift content creator
As PR professionals, how often do you and your team create short-form content for social platforms such as Twitter or LinkedIn? A Twitter post on average has a shelf life of 20 seconds unless something goes viral. For this kind of content, ChatGPT can be a great co-pilot to drive efficiencies and scale.
I chose a recently published contributed article by my colleague Sandy Skees. The title is “The Super Bowl Ads Aimed for Funny but Few Made it Purposeful.” I then wrote a simple prompt:
Act like a social media specialist. Create 5 LinkedIn feeds to promote an article. I will provide the article text. Key words should include “Porter Novelli” “Purpose Marketing” “Sandy Skees” “Sustainable brands” “Advertising”
I fed the article text to ChatGPT. The feeds were too long. So I asked ChatGPT to shorten them to 25 words or under and it came back with the following:
- Discover why Super Bowl ads missed purposeful themes. Sandy Skees, EVP at Porter Novelli, discusses the impact. #PurposeMarketing #SuperBowlAds #SustainableBrands
- Super Bowl LVII: Humor vs. Purpose. Sandy Skees from Porter Novelli explores meaningful messages in ads. #PurposefulAds #SuperBowlLVII #PorterNovelli
- Driving Purpose: Electric Vehicle ads shine. GM and Jeep lead with meaningful themes. #ElectricVehicles #PurposefulAds #SuperBowlLVII #PorterNovelli
- Reduce food waste: Purposeful Super Bowl moments. Sandy Skees on impactful ads. #FoodWaste #PurposefulAds #SustainableBrands #PorterNovelli
- Beyond the Game: The Role of Purpose in Super Bowl Ads. Sandy Skees reflects on advertising’s social message. #PurposeMarketing #SuperBowlAds #SustainableBrands
One lesson I learned here is telling ChatGPT to think like a social media content generator to give the instruction context. Then the prompt should be as prescriptive as required to get the results you want.
This exercise leads me to think about the skills that PR practitioners will need to potentially incorporate generative AI into their workflow and what kind of work the technology should handle. As demonstrated above, the feeds may not be the strongest yet but the potential is there.
A draft starter companion
I considered many directions when preparing for this blog and got lost a bit in my own research. Since my goal is to explore how ChatGPT can improve efficiencies and foster creativity, I put the machine to test and ask it to suggest an outline. I wrote the following prompt:
I am a marketer. I need your help to write a 500-word blog on generative AI and productivity. The blog should be helpful to marketers who are new to generative AI. The goal is to dispel fears and help readers focus on what they can do with generative AI. Do not start writing yet. Do you understand?
I learned this grounding technique to tell ChatGPT to think in context. Then I wrote another prompt:
Write the blog based on the bullets below:
- Why marketers should focus on increasing productivity and unleashing creativity with generative AI.
- What marketers should be concerned about using generative AI.
As expected, ChatGPT generated a 488-word blog. For the first question, it suggested three areas: efficiency and time savings; enhanced personalization and customer engagements; unleashing creative potential.
The content was generic but it gave me some suggestions on areas of focus, which led me to how I structure this piece.
While I had no intention of asking ChatGPT to write this article for me, the content it generated unblocked my thinking. Call it the writer’s block cure if you will.
A divergent thinking facilitator
Many of us have experienced brainstorming sessions that take hours but end up in group think or the “is that it” so called big idea. ChatGPT can help facilitate divergent thinking. You can use it as an icebreaker or a speed brainstorming session. Here’s an example.
A typical brainstorming session for a naming exercise takes 60-90 minutes, requiring several participants. The traditional approach would require a facilitator to select words from the brand architecture and then ask the participants to do an association of words exercise. Then everyone in the room will select the words that appeal to them. And the facilitator will start looking for patterns and connections after the session.
I used ChatGPT to conduct a simulation of the above exercise. For the purpose of this article, my imaginary assignment was to name a new water beverage. Here’s my prompt:
Act like an ideator. You will randomly generate 50 adjectives associated with water. You will then randomly select 5 of the adjectives. You will ask me for one more adjective. You will then suggest names for a beverage based on the selection.
ChatGPT generated 50 adjectives and gave me the following:
Now, I will randomly select 5 of these adjectives: 11. Cooling, 15. Icy, 26. Luminous, 33. Fresh, and 41. Misty. Please provide one more adjective of your choice to complete the selection.
I added the adjective Vital. ChatGPT generated these names:
- Vital Splash
- Icy Vitality
- Luminous Quencher
- Fresh Vitality
- Misty Revive
Was I happy with the names? No. However, the above process took less than 5 minutes. I could afford to iterate until I could find something useful for the next stage. I could also look for divergent connections such as luminous (light) and quench.
My lesson here is that I can take advantage of ChatGPT’s strengths at randomly pairing existing content to uncover “happy accidents” to spark surprising ideas.
Conclusion
This exercise has enabled me to better understand the current potential applications of the technology in a professional setting. We are only experiencing the infancy of the technological development but it is accelerating quickly. And there’s a lot to learn and to think about regarding copyright, data privacy, attribution, and other imminent issues. We have the responsibility of fostering a culture of curiosity and learning to encourage our teams to explore this new frontier of working. A wise colleague said to me, “Don’t be afraid if generative AI will replace your job. Be afraid if you are not skilled at using the new technology to do your job.”
One last important reminder, before you use ChatGPT, or other generative AI chatbots, for work projects, check your company’s policy. Corporations are concerned about proprietary information being shared on an open-source tool and rightly so. Even if you use the incognito mode on ChatGPT, it may not satisfy your company’s requirements.
ChatGPT offers immense potential as a valuable tool for enhancing efficiency and fostering creativity. By understanding its strengths, limitations, and integrating it thoughtfully into our workflows, we can harness the power of generative AI to achieve remarkable outcomes in our profession.
[1] Artificial Intelligence in Plain English, May 1, 2023: Chat GPT: Achieving 100 Million Users in Just 2 Month — A Deep Analysis