GenAI Summit Is Coming Soon

Providing financial services that educate the customer is a desired outcome of all banks, but when does that education slide into information overload? How does a financial advisor strike the delicate balance between exuding customer-centered transparency while still maintaining legally required client confidentiality? When does the unyielding due diligence required of financial analysis depart from reasonable flexibility? These questions have been challenging the industry since First Bank opened in 1791 during the George Washington Administration. 

Multiply these concerns by infinity because of Generative Artificial Intelligence (GenAI)! Sure, you can bet your assets that GenAI will be always more efficient, usually more comprehensive, and occasionally more rigorous than the best human researcher. But the problems GenAI poses, from downright refusing information requests to generating hallucinations, or seemingly credible lies, lays a crucible on the industry unlike any other.

The response to these conundrums is not easy, but you've got to start somewhere. A smart launching point would be the all-day AI in Banking Summit, presented by OnCourseLearning, on Wednesday, June 3, from 9:00 a.m. to 4:00 p.m. During that time, speakers (including me) will deliver six presentations ranging from 30 to 45 minutes over five hours on a wide range of topics merging banking concerns with the powerful, transformative force of GenAI. Do not miss it. You can register here.

Using AI at Each Step of the Writing Process, Part 7: Concerns

The last six posts covered general and specific concerns about depending on AI throughout the writing process. Do I find AI useful to plan, draft, revise, edit, and proofread? Yes. Do I trust AI completely? No. The previous posts explain why. 

In closing this series, I share insights based on AI research in business, law, science, psychology, and medicine to summarize the state of AI in the work world.

Benefits

Undoubtedly, AI provides at least four immediate benefits to users:

• Expanded content. AI possesses an unmatched capacity to capture information on any imaginable topic from a vast range of global databases. A researcher can then ask for AI to organize this data into manageable chunks if it already has not.
• Increased speed. AI retrieves data at a remarkably fast pace that a human being cannot. What takes the savviest researcher hours, days, or weeks to collect takes AI seconds.
• Enhanced quality. AI writes well enough in virtually every orthographic language. This quality check closes the articulation gap between nonnative writers of a language and their more fluent counterparts.
• Broadened scrutiny. AI efficiently uncovers plagiarism. Such a feature is invaluable for teachers, editors, and proprietary businesspeople in assigning original work.

Concerns

Every research article I have read about AI has concluded that the writer must beware of many issues, five of which I mention here:

• Credibility. We may get from AI contrived, imprecise, or unreliable information. Therefore, we must verify content we get from multiple, reliable sources.
• Originality. Nothing from AI is original, and though it provides sources from where it gathered material, it might not be citing the primary source. This task has always been the researcher's job and continues to be. 
• Transparency. Despite AI programmers' best efforts, AI at times does not provide sources, challenging writers to investigate the source material for themselves.
• Accessibility. AI may provide information that is nearly impossible or impractical to trace. This dilemma forces writers to practice the old adage from journalism: When in doubt, leave it out.
• Compliance. While AI has been around for a few years, it is still a new technology that has left a lot of organizations deciding on how to best use it. In some places, using it equates to an unethical breach of company policy, so writers should know the organizational policy. 

As I wrote in "Using AI to Improve Writing Creativity, Productivity, and Quality," writers must remember that they are the boss and AI is just their assistant.

Using AI at Each Step of the Writing Process, Part 6: Proofreading

As of now, AI is better at proofreading than it is at editing, with a caveat. In distinguishing between editing and proofreading, you can use as an example having written in first draft:

Evlyn impacted me.

When editing, you focus on command of language. In this case, you decide to change the verb impacted:

Evlyn affected me.

When proofreading, you search for overlooked errors, such as spacing, font, spelling, and number inconsistencies. Here you pick up the spelling error Evlyn:

Evelyn affected me.

AI ensures you spell, capitalize, and punctuate properly. You can also count on AI to make every sentence comply with standard grammar rules. Your typos now are less tolerable since you have AI to run a quick check for you before pressing send. 

But beware! AI loses its effectiveness for various reasons, some programmatic and some bandwidth. I recently prompted it to "Proofread only for overlooked errors without changing the content" of a third draft that AI and I collaborated on. It responded that the draft was error-free yet reverted to an earlier draft. Clearly, AI and I have a different definition of proofread. Also, it might have been suffering from information overload because of the multiple drafts we covered. In addition, here are four proofreading points it did not detect:

• Repetitive phrasing. One of my sentences read Effective immediately, a one-year lease a $90 per square foot is available immediately. AI did not report the repetitive opening and closing phrases.
• Spelling inconsistencies. I mistakenly included alternate spellings of smartboard and smart board in the same document, which AI did not call out.
• Number-letter spacing. I inserted a phone number without a space on either side of two words, at732-718-3361to, which AI ignored.
• Spacing inconsistencies. I drop only one space after a period, but I inadvertently double spaced after one sentence in a document of about 20 sentences. While I do well at revising and editing, proofreading is my biggest weakness in the quality control phase of the writing process. Luckily, I saw this error, which AI overlooked.

Once again, the theme of this series has been not to get lazy. The final look of your draft is on you, not AI.