I got on the AI user train pretty early.
We were synthesizing nearly 4 months of work on a large project and had to write nearly 11 long reports that brought together insights about our client organization, its stakeholders and general trends and macro category data.
If you have written a long form report from scratch, you would understand the amount of work it involves. First is making sense of all the unstructured and structured data to arrive at key insights for the report. Then the focus required to write it in a cohesive and readable manner. Then there is the absolute numbness of the mind that can follow countless corrections in punctuation, capitalisations, spacing.
The news of ChatGPT arrived as a ray piercing through this fog of report writing. Ever excited by the power of technology in enhancing our work, I thought I would let ChatGPT fetch me some data from reports available in the public domain to enrich a report. I gave specific prompts – on the data source, the variables to look for etc. I was ecstatic with the result – I got all the data needed for a few charts in a fraction of the time it would have taken me otherwise.
We created a beautiful looking report with our insights and the new data charts. A day before we were to share the reports with the client, my colleague messaged me, “Could you tell me where you got the data for the charts? I can find part of the data but the other part is not published anywhere else.” It is then that I realized, ChatGPT had provided made up data for part of my request!
We were saved, just in time, by human intervention! However, if we had not realized this error, we would have continued to rely on ChatGPT to provide us data without a doubt. I wonder at the quantum of error prone reports that have been published with the help of generative AI since then. AI definitely needs a fact check! It is useful, without a doubt. However, it needs to be used with a consciousness that is absolutely human.
Let AI assist, not direct
After the reality check very early on, I have continued my cautious exploitation of AI tools. Since then I have used them to help me start on an article, generate ideas of taglines, do quick research for general information and check for grammatical correctness of pieces that I have written. I have also dabbled in them to generate images to illustrate a point I want conveyed.
Here is what I have learnt based on a year of dabbling with generative AI.
Manual content output > Bard content output > ChatGPT content output
When compared with output from ChatGPT, Bard seems to generate content that is of better quality, seemingly less artificial sounding and with better adherence to prompts. It is not surprising, given the amount of data Google has.
But nothing beats thoughtful content created by humans! It is not only a personal opinion. Content created by experienced content creators have shown better results across studies. They garner much higher Impressions, Clicks, as well as Position in search results. They also get much better conversions.
Use AI to embellish content, but with caution
I have found AI useful in getting me missing pieces of data or information to support a line thought in human-created content. It saves an incredible amount of time for such quick searches. I am still wary of relying on it completely, especially for quantitative data. Since it gets its information by scraping the content on the internet, accuracy is not guaranteed. The Internet is the place where misinformation plays out well, after all. So the prompts need to specify which data sources it should search to get the information needed.
Tagline brainstorm, starters and options
There is very interesting and productive back and forth one can do with Generative AI in creating campaigns, especially short form posts and taglines. It can generate a number of options quite quickly that provide a good stimulus for brainstorming. We have used bits and pieces of many options created by AI chat to arrive at campaign taglines that have worked quite well.
Decision making on speed
There are many tools for marketing data analytics to aid quick decision making. It is now inbuilt in many of the tools we use for digital ad management, for instance. This is where AI is going to transform marketing actions. It increases productivity and speed manifold with its lightning data analysis capabilities. So, us humans can no longer keep playing with data! We have to shift attention and skills to understanding business implications of data and making sound decisions.
A concluding thought. Most challenges in life and business cannot be solved by knowledge alone. Intelligence based on past knowledge has its limitations. We live and work in a web of human relationships and have to learn about them all the time. New thinking, openness and fresh approaches is what will take us ahead, and that needs us to be very human!
Exciting times ahead, for sure!
