𝗔𝗜 𝗠𝗮𝗱𝗲 𝗘𝗭: 𝘍𝘪𝘨𝘩𝘵𝘪𝘯𝘨 𝘐𝘯𝘴𝘶𝘳𝘢𝘯𝘤𝘦 𝘋𝘦𝘯𝘪𝘢𝘭𝘴 𝘸𝘪𝘵𝘩 𝘈𝘐

(3 Min Read)

👂 𝗛𝗮𝘃𝗲 𝘆𝗼𝘂 𝗵𝗲𝗮𝗿𝗱 𝗼𝗳 “𝗗𝗲𝗹𝗮𝘆, 𝗗𝗲𝗻𝘆, 𝗗𝗲𝗳𝗲𝗻𝗱?”
I experienced a health insurance claim denial and am $1,600 out-of-pocket. Over six months, I spoke to six reps by phone. Each rep listened to what I presented and agreed that the insurer was responsible. They resubmit the claim and say the Claims Dept. will take 14–21 days to review it. Of course, there’s no phone # or email for the Claims Dept. I login at 21 days: 𝗗𝗘𝗡𝗜𝗘𝗗! And the process starts again.

𝘌𝘹𝘩𝘢𝘶𝘴𝘵𝘪𝘯𝘨.

🆚 𝗗𝗲𝘁𝗲𝗿𝗺𝗶𝗻𝗶𝘀𝘁𝗶𝗰 𝘃𝘀 𝗣𝗿𝗼𝗯𝗮𝗯𝗶𝗹𝗶𝘀𝘁𝗶𝗰
I recently finished the book 𝘈𝘨𝘦𝘯𝘵𝘪𝘤 𝘈𝘳𝘵𝘪𝘧𝘪𝘤𝘪𝘢𝘭 𝘐𝘯𝘵𝘦𝘭𝘭𝘪𝘨𝘦𝘯𝘤𝘦 by Pascal BORNET. He discusses these concepts:

• 𝗗𝗲𝘁𝗲𝗿𝗺𝗶𝗻𝗶𝘀𝘁𝗶𝗰 𝗽𝗿𝗼𝗰𝗲𝘀𝘀𝗲𝘀 follow defined rules and return the same result each time for a given set of inputs. Consider an accounting process for a SOX-compliant public company. Accuracy is key.
• 𝗣𝗿𝗼𝗯𝗮𝗯𝗶𝗹𝗶𝘀𝘁𝗶𝗰 𝗽𝗿𝗼𝗰𝗲𝘀𝘀𝗲𝘀 incorporate randomness and produce different results with the same input. Think of a marketing team needing a creative list of blog post ideas.

That’s a helpful way to think about AI solutions

📝𝗨𝘀𝗶𝗻𝗴 𝗖𝗵𝗮𝘁𝗚𝗣𝗧 𝘁𝗼 𝗛𝗲𝗹𝗽 𝗜𝗺𝗽𝗿𝗼𝘃𝗲 𝗠𝘆 𝗔𝗽𝗽𝗲𝗮𝗹 𝗟𝗲𝘁𝘁𝗲𝗿
So, I drafted my appeal letter and imported it into ChatGPT. If you’ve used an LLM to help revise writing, you know the process can involve a few “spin cycles”—revise, review, repeat. Many times, I request comments from GPT but rewrite myself. This time, I received a handful of comments. Exhausted, I said: “Go ahead and just rewrite the letter.”

✅𝗔 𝗖𝗮𝘀𝗲 𝗪𝗵𝗲𝗿𝗲 𝗮 𝗣𝗿𝗼𝗯𝗮𝗯𝗶𝗹𝗶𝘀𝘁𝗶𝗰 𝗥𝗲𝘄𝗿𝗶𝘁𝗲 𝗪𝗮𝘀 𝗨𝘀𝗲𝗳𝘂𝗹
We’ve all had experiences where an LLM rewrites and the result was completely off. In this case, the randomness factor  pushed the tone of frustration that I felt. Here are three of the juiciest parts:

• 𝘈𝘵 𝘴𝘰𝘮𝘦 𝘱𝘰𝘪𝘯𝘵, 𝘵𝘩𝘦 𝘳𝘦𝘢𝘴𝘰𝘯 𝘧𝘰𝘳 𝘺𝘰𝘶𝘳 𝘥𝘦𝘯𝘪𝘢𝘭 𝘤𝘩𝘢𝘯𝘨𝘦𝘥! 𝘛𝘩𝘪𝘴 𝘸𝘢𝘴 𝘯𝘰𝘯𝘴𝘦𝘯𝘴𝘦!
• 𝘛𝘩𝘪𝘴 𝘦𝘹𝘱𝘦𝘳𝘪𝘦𝘯𝘤𝘦 𝘩𝘢𝘴 𝘧𝘦𝘭𝘵 𝘭𝘪𝘬𝘦 𝘢 𝘵𝘦𝘹𝘵𝘣𝘰𝘰𝘬 𝘦𝘹𝘢𝘮𝘱𝘭𝘦 𝘰𝘧 "𝘋𝘦𝘯𝘺, 𝘋𝘦𝘭𝘢𝘺, 𝘋𝘦𝘧𝘦𝘯𝘥"—𝘢 𝘵𝘢𝘤𝘵𝘪𝘤 𝘸𝘩𝘦𝘳𝘦 𝘪𝘯𝘴𝘶𝘳𝘦𝘳𝘴 𝘴𝘵𝘢𝘭𝘭 𝘤𝘭𝘢𝘪𝘮𝘴, 𝘦𝘹𝘩𝘢𝘶𝘴𝘵 𝘱𝘰𝘭𝘪𝘤𝘺𝘩𝘰𝘭𝘥𝘦𝘳𝘴, 𝘢𝘯𝘥 𝘵𝘩𝘦𝘯 𝘥𝘦𝘯𝘺 𝘱𝘢𝘺𝘮𝘦𝘯𝘵 𝘰𝘯 𝘢 𝘵𝘦𝘤𝘩𝘯𝘪𝘤𝘢𝘭𝘪𝘵𝘺. 𝘐𝘴 𝘵𝘩𝘪𝘴 𝘳𝘦𝘢𝘭𝘭𝘺 𝘩𝘰𝘸 𝘺𝘰𝘶 𝘰𝘱𝘦𝘳𝘢𝘵𝘦?
• 𝘐 𝘦𝘹𝘱𝘦𝘤𝘵 𝘢𝘯 𝘢𝘤𝘵𝘶𝘢𝘭 𝘳𝘦𝘷𝘪𝘦𝘸 𝘰𝘧 𝘵𝘩𝘪𝘴 𝘴𝘪𝘵𝘶𝘢𝘵𝘪𝘰𝘯—𝘯𝘰𝘵 𝘢𝘯𝘰𝘵𝘩𝘦𝘳 𝘦𝘮𝘱𝘵𝘺 𝘳𝘦𝘫𝘦𝘤𝘵𝘪𝘰𝘯.

🟢 𝗜𝗡 𝗖𝗟𝗢𝗦𝗜𝗡𝗚 🟢
This appeal letter is a great example of when AI’s probabilistic nature can deliver the punch a situation demands.

𝗗𝗲𝘁𝗲𝗿𝗺𝗶𝗻𝗶𝘀𝘁𝗶𝗰 𝘃𝘀 𝗣𝗿𝗼𝗯𝗮𝗯𝗶𝗹𝗶𝘀𝘁𝗶𝗰: Know when each applies in AI solutions. The randomness of a probabilistic output may be best!