Does ChatGPT dream of digital marketing plans?
Discover how insurance agencies can leverage generative AI to better engage & educate policyholders.
This was not written by AI.
As the public plays with ChatGPT and new developments in GPT-4 and competitive platforms continue, it seems that we are peering into the future dreamt by Asimov, Bradbury, Dick and other writers of the sci-fi ilk.
While artificial intelligence and machine learning have long pervaded the public’s imagination — from the humanity-crushing Skynet of the “Terminator” franchise to AI/human love story that is Spike Jonze’s “Her” — we have reached a point where science fiction is becoming one with reality. For better, for worse, until death do us part.
Chris Cline, executive director of the Big-I’s Agents Council for Technology, points out that with 50 sets of regulations and a range of varying policies, the U.S. property & casualty industry should proceed into generative AI with caution.
“Trusting ChatGPT to go A to Z with accuracy, compliance and E&O risks mitigated would be a mistake, but it could certainly get you started,” Cline says. “It can expedite content creation for blogs, or even vlogs if they want to do videos and need help with scripting.”
He adds that a short time spent constructing a decent prompt can result in several paragraphs of content that is “reasonably well constructed and articulate,” and can serve as a great jumping-off point for agencies looking to up their content game.
‘An AI is trustworthy, loyal, helpful, friendly . . .’
While ChatGPT and its promises for the future are filling headlines, it is also causing some consternation among leading engineers, academics, researchers, industrialists and business executives.
In an open letter, which has garnered more than 50,000 signatures, people from across industries have called for a six-month pause on work at all AI labs with systems that are more powerful than GPT-4, the newest version of OpenAI’s language model systems.
“I’m a huge follower of Elon (Musk), and here is a dude that wants to colonize Mars and connect us all with Neuralink, and he is telling us we need to stop,” Cline says, continuing: “If one of the most forward-looking humans alive tells us that this is scary, we should pay attention.”
The heart of the letter puts forth questions such as: “Should we let machines flood our information channels with propaganda and untruth?” and “Should we automate away all the jobs, including the fulfilling ones?”
However, the letter isn’t calling for a complete halt to all AI development, but rather a “stepping back from the dangerous race to ever-larger unpredictable black-box models with emergent capabilities.”
The letter goes on to suggest that current AI researchers take up an ethos akin to the Scout Law, and focus on making systems that are trustworthy, loyal, accurate, safe, interpretable, transparent, robust and aligned.
The generative AI-enabled future is producing seemingly equal amounts of trepidation and jubilation, yet the age-old question — the real nut of the matter — persists: “What have you done for me lately?”
‘Write me a LinkedIn post’
Although ChatGPT and similar AI technologies are still developing, and their full potential has yet to be unlocked, there are practical applications that insurance agents can deploy today.
For example, Agency Revolution recently launched a mobile app that merges ChatGPT with the Vestorly content platform to help agents find content that matches their professional interests as well as personal interests. From there the generative AI creates a social media post to accompany the content.
“Social is big from a selling perspective but the usability and adoption across all of these platforms are low. The reason why is because we are asking agents to do another job in the plethora of jobs they are already asked to do on a daily basis,” says Jason Walker, president of Agency Revolution. “We are trying to take that onus off of them while still giving them the benefit of social selling.”
He says that agents should view ChatGPT as a way to create a first draft, which can be further tweaked to better align with the agency’s branding or the agent’s personal voice.
Once the post has been reviewed, and checked for accuracy, agencies can sync the app up with LinkedIn and Facebook and push posts out to the public.
“As part of the algorithms native to LinkedIn and Facebook, consumer behaviors and interactions with those posts dictate their relevance and prioritization in the social media platform’s feed,” Walker says. “Agencies are able to use the data provided by the social media platforms and configure their preferences in our app’s curation tool to hone future posts/content.”
This was written (sorta, kinda) by AI
What follows is a Q&A with Agency Revolution’s Jason Walker. The questions were written by the public version of ChatGPT. Follow-up questions were generated (in real time) by the program based on Walker’s responses during a phone conversation with PC360.
It was edited (by a human) for length and clarity.
ChatGPT: How do you plan to train the AI model to generate content that is relevant and engaging for your target audience?
Walker: Just like we do with our website platform, we go through a process of asking the agent about them and what makes them unique. We then capture that information and feed it into the sites that ultimately launch so they don’t read the same for every agency. The same process is conducted for this product as well. It asks them about specific topics that they are interested in and specific words or colloquialisms that they may use.
It is all about making it as personalized as possible so it feels very custom to each of the users and they have something to say that is relevant and different from the next agency or carrier, whoever the end user is.
ChatGPT: Can you provide an example of a challenge you faced while using generative AI for social media content creation?
Walker: As we’ve seen when going through the development phase and doing demos and it was a little bit more generic, the length could come out a little too long or too short. There might be too many emojis or not enough. There might be no hashtag or closing question.
We know that if an agent is talking to a consumer base and just says something like, “Here is the article and this is what it says,” they are telling consumers to click on the article. At that point, the agency is less likely to interact with that consumer. Whereas if you add a question to the end of the caption, you can keep that consumer locked in with you.
If you are doing the open-ended version and just allowing it to create content on your behalf, there is a lot of generic there. It is not trained to do anything but create a summary. It isn’t following the best practices of social, it is just creating a summary and the consumer will click on the article. We want the agent to keep the consumer captured within that experience and not lose them to a third-party site.
ChatGPT: What are some of the best practices for creating engaging and interactive social media content, and how do you plan to incorporate these practices into your content strategy, while still leveraging the efficiency gains of generative AI?
Walker: When you are thinking about content that is interactive, you have to have a little bit of narrative with a beginning, middle and end. In our case, we are explaining the content of a particular post, we are then giving the agent a question to interact with the consumer.
Another piece we know is important in the social space: Hashtagging is still around, and we want to make sure we have relevant hashtagging around industry-specific terms that are in line with the article being pushed out.
We also want to make sure to bring social back to social media. We understand there is a balance between the content you are putting out there and your personality. Using emojis, little images or different types of emoticons within a post is germane to social media. Adding specific content; emojis; a beginning, middle and end; hashtagging — those are all best practices that should be trained into a system. Once that takes place, it should set up a good conversation for the business user and their audience.
The content being generated is a great first draft. At any point, our users can go in and edit the content to ensure it fits their voice and tone.
ChatGPT: In your opinion, what are some of the potential ethical implications of using generative AI for social media content creation, and how do you plan to address them in your strategy moving forward?
Walker: It’s having a tool with no parameters or guidelines exposed to an audience, and that is exactly where we fit in all of this.
The tool (ChatGPT) is creating based on a bunch of public content and uses natural language processing to put something together that is coherent. There is the opportunity for something to bleed in that a business owner would not want to put in front of their audience. A solution with no parameters around it presents the opportunity to expose content that is not favorable.
Having the ability to edit posts before they go out also puts in additional guidelines and rules. If you just have a solution that has ChatGPT baked in and you just press go with a set-it-and-forget-it mindset, you are opening yourself up for content that you won’t see as favorable. And neither will your audience.
Related: