A friend recently posted a short opinion on Facebook:

“ONLY THREE KINDS OF PROFESSIONAL WRITERS would lean on AI to puke out their prose: the lazy, the cynical and the hopelessly overwhelmed. Tell me I’m wrong.”

This post was, in fact, a promo for a longer blog post. His blog, by the way, is excellent. My friend is an excellent writer. He wrote an excellent book, about which I interviewed him on my podcast. He’s a smart guy.

But about this, he’s wrong. I said so in a comment on his Facebook post. He asked me to elaborate. This post is that elaboration.

What AI Writing Tools Are—And Aren’t—Good For

This whole topic arose when another friend posted something about AI writing tools. I responded with a hastily produced example using the Paragraph Generator template on Jasper. On reflection, it was not the best choice to demonstrate what an AI writing tool can do, which is not to say that it sucked ass. My friend copied that example into his post along with my belief that a good writer would edit this output rather than simply publish it, and then he invoked Joan Didion on the creative process.

A college professor introduced me to her through her essay collection, “Slouching Toward Bethlehem,” leading her to become one of my favorite writers. (Number one on my list, in case you’re wondering, is Gabriel Garcia Marquez.) I have read every book and essay that Didion wrote, some more than once. I do not recall seeing a single corporate earnings release among her considerable catalog of work, nor a single press release announcing the promotion of an employee to an executive role. I have yet to see an AI writing tool with a fiction component or an essay module. Novelists and essayists are clearly not the targets for these tools.

They can, of course, be a real boon to people who are not writers but who need to produce clear, understandable written content. Any communicator who has read a report from someone in a non-communication job knows just how much work their writing needs to make it presentable. But my friend’s post focused on professional writers, so let’s stick with that theme.

Unlike my friend (and Joan Didion, may she rest in peace), some of us need to produce large volumes of content every day. In his post, my friend raked me over the coals for my suggestion that a busy communicator might use a tool like this to produce a paragraph about, for instance, an employee who earned a certification.

I shall digress because my friend suggested we internal communicators should not be demeaning ourselves with this lowly type of content in the first place. But employees like reading about their colleagues’ accomplishments; intranet analytics reveal this type of content is popular. Beyond popularity, it serves business purposes. Once reading one of these recognition pieces, employees learn whom to turn to when they need the expertise embodied in the certification they earned. People like congratulating work friends and teammates on hard-earned achievements. And being recognized is a significant intrinsic reward, motivating employees far more than extrinsic rewards. Watch Daniel Pink’s TED Talk if you don’t believe me. In any case, “recognition stuff” is not, contrary to my friend’s view, “dumb old things”.

Back to the matter at hand. My friend dismissed the prospect of using an AI writing tool to generate drafts of this kind of content—it was over this idea that he rubbed my face in Joan Didion’s approach to the creative process—or for producing potential outlines or idea starters.

A Place for AI Writing Tools for Professional Writers

Despite my friend’s objections, it is not hard to justify professional writers (and others) using these tools. It is crystal clear that my friend penned his screed without researching AI writing tools or taking one out for a spin. That’s rather like writing a review of a movie you haven’t seen.

Another digression. You may, in your mind, hear me say “my friend” with a sarcastic sneer in my tone of voice. That would be a mistake. We have enjoyed long evenings drinking and talking. He invited me to speak at the annual conference of a marvelous organization he built and runs. When my wife and I visited Chicago recently, we had dinner with him and his wife on our wedding anniversary. I like him a lot. We have been through these disagreements before, almost always over the viability of technology for communicators. These squabbles don’t affect our friendship.)

Before I address the value of AI writing tools, consider IBM’s view that the AI label is unfortunate. It should not be artificial intelligence. Augmented intelligence is closer to the mark. AI does not exercise judgment; it does not know how. (Nor can it convey empathy, creativity, or emotions. Yet.) But it effectively augments the skills and talent you bring to the table.

AI’s power is in its predictive capabilities. Based on the massive amounts of data it has consumed, it can take your input and return something it believes you’re looking for.

A new AI graphics tool called DALL-E 2 has just entered beta. (There’s a waiting list to use it. I’m on it.) You provide some input, like this:

In response, it creates several images it thinks you’re looking for, including (based on the inputs in the image above) this one:

My friend might ask, “What self-respecting graphic designer would ever use such a tool?” But look at that image. This was not cobbled together from graphics stored in a database. This was invented from whole cloth, completely bespoke, based on what the AI knows about graphics, graphic styles, and the intent behind the words input by the user.

Another AI-based image-producing tool creates photo-realistic images of people who do not exist. Each time you refresh the screen, you get another image of a completely made-up person. Try it, then try to convince me it’s clear hackery that nobody will accept as anything but a subhuman attempt to mimic what a skilled professional special effects wizard can do.

On a sillier level, an AI image tool is available that turns any image into a “Stranger Things” poster—like this one (the AI comes into play in creating the title based on what it sees in the image):

The image of a robot typing in the banner for this post was also produced by yet another AI graphics program responding to my request for an image of, well, a robot typing.

AI writing tools are still in their early days. They do not—not the ones I have tried, at least—simply take your input and regurgitate some computer-generated words. The paragraphs Jasper wrote for me were based solely on the answer to the question, “What is your paragraph about?” (Why construction companies should care about the metaverse) and keywords to include (metaverse, construction, building, general contractors). The three options Jasper returned to me contained a considerable amount of accurate information that was not inherent in those inputs.

A Writer’s Assistant

The software gathers data from millions of online resources in mere seconds based on the user’s inputs. That research could take the most professional of business writers (or their research assistants) hours, even days. Because these programs already grasp the rules of grammar, punctuation, sentence structure, and the like, they can assemble the information they have uncovered into a coherent narrative that matches your parameters (e.g., word count and tone of voice). It is not, unsurprisingly, the most creative content you will ever see. On the other hand, it is unbiased because it has no opinions. (Remember, AI cannot make judgments.) AI can even detect whether words you have written have already been written by someone else, protecting you from plagiarism allegations.

While I have not yet used Jasper for an actual work-related product—just testing and playing so far—I do use another AI tool regularly. Grammarly checks my work as I’m writing. I am not ashamed to admit that it often suggests better sentence structure and catches me in subject-verb disagreements. I may well have found and fixed these errors on my own during an edit, but getting it right as I’m writing saves me time. As one member of a two-person team doing the work of a team that should be larger, time is precious. I am not embarrassed that an AI tool helps me out. I am, on the contrary, grateful for the tech that makes it possible.

Mainstream Journalism Adoption

My friend calls any professional writer who would turn to an AI writing tool a “hack.” That must include Reuters and the Associated Press, both of which use AI writing tools. AP announced its use of an AI tool in 2019 to produce data-driven text previews of all Major League Soccer games. AP’s decision was not based on laziness or cynicism. It was based on the simple fact that AP can not employ enough writers to crank out a preview for every MLS game. Before AI, they simply did not even try. Today, AP also uses AI writing tools to report on high school games—again, because there are not enough journalists at AP to cover every high school game.

Reuters and AP also use AI writing tools to produce corporate earnings reports, which has freed up 20 percent of its reporters’ time that can be allocated to the in-depth analysis only professional writers can do. Sometimes, that in-depth analysis results from insights the AI discovered in the corporation’s numbers.

A Reuter survey that included 200 publishing executives found that 91 percent cited productivity efficiency as one of their top priorities. If there are better solutions than AI writing tools, I’m sure they would be happy to learn about them. Absent alternatives, Reuters has been using AI to produce corporate earnings reports since 2018, helping its reporters discover insights from corporate data that would have taken much longer to surface without it, all while improving accuracy.

The Washington Post’s proprietary robot reporter published some 850 articles way back in 2018. The software, called Heliograf, created 300 short reports and alerts from the Rio Olympics. Having proven its capabilities, it was used later to cover congressional and gubernatorial races on election day, along with D.C.-area high school football games. How serviceable are these sports stories? Judge for yourself:

The Yorktown Patriots triumphed over the visiting Wilson Tigers in a close game on Thursday, 20-14.

The game began with a scoreless first quarter.

In the second quarter, The Patriots’ Paul Dalzell was the first to put points on the board with a two-yard touchdown reception off a pass from quarterback William Porter.

Wilson was behind Yorktown 7-0 heading into the second half. Wilson’s Anton Reed tied the score with a two-yard touchdown run. The Patriots took the lead from Wilson with a two-yard touchdown run by Tanner Wall. The Patriots scored again on Adam Luncher’s 29-yard field goal.

Yorktown maintained their lead going into the fourth quarter, 17-7. The Patriots extended their lead over the Tigers on Luncher’s 27-yard field goal. Wilson cut into the Patriots’ lead with a three-yard touchdown run by Amir Gerald. The game ended with Yorktown defeating Wilson, 20-14.

Yorktown’s top passer was Wilson, who completed 6 of 10 passes for 91 yards and one touchdown. Yorktown’s top rusher was Wall, who had seven carries for 57 yards and one touchdown. The Patriots’ top receiver was Wall, who had four catches for 54 yards.

Yorktown will play Wakefield High School (1-0, 0-0) on Sept. 8. Wilson will play McKinley Technical High School (0-1, 0-0) on Sept. 8.

Clearly, this is not the work of Frank Deford, Jim Murray, or Tim Withers, but then again, no sports editor would ask an A-list sports writer to crank out copy like this. The choice was an AI writing solution or no report at all. Fans of the Yorktown Patriots are doubtless happy to have this report of their team’s victory over no report at all. (Tigers fans may feel differently.)

Last year, AP announced a two-year partnership with the Knight Foundation to help local newsrooms employ AI tools, which can help these cash-strapped, skeleton-crewed bulwarks of democracy continue to crank out simple reports, extracting information from data, so their few reporters can focus their time on more in-depth, important stories.

Business Buys In

Salesforce now owns Narrative Science, one of the pioneers in AI-generated writing, integrating it into Tableau, the business intelligence software it also owns. Already acclaimed for its ability to generate charts, graphs, and other images from company databases, Tableau can now auto-generate text analysis of that same data. This move by Salesforce follows Microsoft’s entry into the space with Smart Narrative, a free tool for Microsoft Power BI, allowing the business analytics software to auto-generate a text explanation of a chart or a graph with a single click. In the image below, the text on the right was generated automatically based on the charts and graphs Power BI produced from data.

Business communicators and journalists who have used AI writing tools cite a variety of uses, even if—unlike corporate earnings reports and high school sports coverage—they would not copy and paste an AI output into a website, an intranet, or a newsletter. They can help generate ideas, structure your thoughts, brainstorm ideas, break through your writer’s block, assemble an outline, and simplify complex narratives (Jasper offers one template called “Explain It to A Child.”)

Improving a Process

I knew I wanted to be a writer when I was 8 years old. Sixty years later, I have no plans for AI to craft my articles for me. On the other hand, I am delighted to have AI help me in ways that cannot be quickly or easily replicated with analog resources. User interface expert Jared Spool once told me technology can do only three things: solve a problem, improve a process, or let you do something that wasn’t possible before. I am just fine employing technology to help improve my writing process.

In fact, I have no doubt my friend uses technology to aid him in his writing process. I would bet real money he has searched Google, for example, and has probably perused synonyms using an online thesaurus. Still, he is obstinately opposed to the idea that an AI can search for him and assemble its findings into a cogent, coherent paragraph. And while I would not cut and paste that paragraph into my copy, I am more than happy to let it save me hours of searching and note-taking. My friend probably has no problem with a human assistant hired to do exactly the same thing: conduct research and deliver a report a writer can incorporate into the article they’re crafting. That assistant and a writing AI tool augment the writer’s work in pretty much the same way. One is just cheaper and faster than the other.

My friend may not like it, but AI will play an increasingly big part in the kind of writing business communicators and journalism outlets produce. It will continue to improve. That doesn’t fill me with scorn. It does not make me feel like a hack. It excites me.

Note: Some of the AI writing tools available right now include Wordsmith, AI Writer, QuillBot, Writesonic, and Article Forge. They all offer free trials. Unlike my friend, you should actually try them out to see how they can help you get your job done. You may be pleasantly surprised even if you’re an unrepentant purist, like my friend.