Skip to content

Businesses are not rushing to adopt AI

Here's an AI chart you probably haven't seen:

After rising through 2019, AI adoption by businesses has been dead flat for the past five years. What's more, the vast bulk of this is old-school "AI" that doesn't deserve the name, not generative AI of the kind that's been the focus of all our attention for the past couple of years. Despite its almost miraculous abilities, we're still struggling to find compelling real-world uses for it.

22 thoughts on “Businesses are not rushing to adopt AI

  1. Brett

    That lack of reliability in answers is a real problem, although I think it's gotten better. It limits the business cases to stuff where you have a relatively easy time doing error corrections or the level of errors is acceptable.

    And of course, one wonders what "adoption" means in this context. I used Google Gemini to write some foreign language customer service email templates - does that count as "adoption" even if I almost never use it otherwise, and only used the free version?

  2. bbleh

    Really not sure about the numbers, because they depend so much on the respondents' understanding of what "AI" means, and the term has become so wildly hyped, and therefore debased, that the range of such perceptions likely is VERY wide.

    But assuming a business owner/manager has a reasonably good understanding of what actual AI applications are and can do -- that is, they're not sucked in by the hype, or going with the flow because they think their superiors are sucked in by the hype -- what can AI actually DO for the "median" (ok, 60th-percentile) business?

    I can see a couple plausible things for certain broad categories of businesses -- automating routine clerical functions, maybe increasing marketing reach -- but there would be real costs and a potentially costly period of "burn-in" and adjustment. I'd still want humans involved in anything customer-facing or with legal implications, and for many businesses I can't see the actual benefits amounting to much, even if they panned out (which they might not). I'm sure the salesmen can hypothesize many more, but I'd still be careful about how much is hype and how much would be a real effect on the bottom line.

    So I'll sign on to the conclusion of the post, and also color me skeptical that the actual numbers are that high.

    1. Pittsburgh Mike

      I think you've already seen some of the issues with using error prone AI for routine stuff -- the airline whose chatbot told someone they could make a change against the actual policy. The airline was sued and lost when they tried to disown the chatbot's statements.

      This stuff is fun to play with, but isn't good enough yet for even routine purposes, because it is unreliable.

  3. somebody123

    My husband is a director at a Fortune 500 that rhymes with “faster bard.” They’ve been pitched a ton of LLM products, but none with a working demo that makes fewer mistakes than a person. A lot of AI companies can’t provide a demo at all- they want you to sign a contract for them to create a custom product, and no company in the finance sector is going to take a risk like that.

    That said, they do use Microsoft products, so they have Copilot, and my husband uses it to do first drafts of the sort of documents you have to write but no one will ever read. He just used it to write everyone’s employee evaluations. He had to skim the documents to make sure it hadn’t hallucinated, but it was a huge time savings since they’re trash anyway. So I guess that counts as adoption?

    LLMs are turning out to be like the blockchain- intellectually interesting but with very limited practical uses.

    1. Art Eclectic

      That's the same kind of stuff I use it for, but saving time is a valid use case. Getting jump on the outline for a 20 slide presentation matters and gets you to the stage of deciding whether slide 7 really is necessary or there's a cleaner way to get your point across on slide 12.

      I find it to be like having both an intern and editor rolled into one.

    2. Pittsburgh Mike

      Bingo. I wouldn't trust it for anything important, but it generally produces an adequate summary of the notes from a meeting that didn't look particularly worth attending.

      Some relatively simple things have worked for years now: reading hand written addresses, reading checks, along with understanding spoken voice statements on phones and the like.

      There are some machine learning applications with much narrower domains that look promising -- the oft-written about radiologist replacements or skin cancer screening apps.

      But there's a decent chance that the generative AI stuff isn't going to get much better than we see today, and it will have to get a lot more reliable before you'll be able to trust it to answer questions from customers definitively.

  4. Scott_F

    I imagine a lot of the "adoption" is using AI to summaries documents before diving into the details or to spiff up you emails.

  5. Scott_F

    I was going to find this chart and share it on another cite when I came across the 2024 version which shows an increase from 55% in 2023 to 72% in 2024. This blows Kevin's post out of the water. Remind me why I come back here...

  6. golack

    Phone apps love chat bot help.
    A summary/paraphrase of a recent encounter.

    Me: How do I report a problem with the app?
    Response: You know, you're in the app right now.

    In case you're wondering, the "feedback" button did not respond anymore, so I tried the help feature.

  7. Justin

    I don't use social media like facebook and instragram, so I find this hilarious. Before too much longer, all these bots will be talking to each other and drive the political and cultural conversation as if we humans didn't even exist! Hilarious.

    https://nymag.com/intelligencer/article/meta-wants-more-ai-bots-on-facebook-and-instagram.html

    "The Silicon Valley group is rolling out a range of AI products, including one that helps users create AI characters on Instagram and Facebook, as it battles with rival tech groups to attract and retain a younger audience. “We expect these AIs to actually, over time, exist on our platforms, kind of in the same way that accounts do,” said Connor Hayes, vice-president of product for generative AI at Meta. “They’ll have bios and profile pictures and be able to generate and share content powered by AI on the platform . . . that’s where we see all of this going,” he added."

    Entertainment!

    My employer is probably using AI for something, but the only thing I've seen personally are people using copilot or teams to send email summaries of meetings which no one reads. I think maybe there's some app coming out to do expense reports soon. Meh.

    1. Chondrite23

      I imagine students using AI to write papers that are graded by AI bots. Or HR AI bots are reading resumes written by other AI bots.

      1. Justin

        No doubt. Happy New Year to you and all. I'm sure all the bots will have fun. I'm reminded of that scene in the Poseidon Adventure movie where everyone is celebrating and singing as the wave hits the ship and flips it over. Seems like we're about to get some sort of disaster! Good luck.

  8. OldFlyer

    Any chance they could use AI to understand me when I scream "LIVE REP!!" to the customer service phone Bot.

    Or will they just program him to say "I'm sorry Dave, I'm afraid I can't do that"

    1. iamr4man

      Funny, just read this:
      “In dealing with such robotic logjams, there's usually a magic word or phrase you can say that will get them to kick the call over to an actual human being. I tried "agent" and "human being" and "representative" and "pharmacist" and everything else I could think of and finally, "she" said she'd connect me with someone. I think but am not certain the magic words were "live agent." I've jotted that down in the Notes section of my phone listing for this company. But wouldn't it be nice if they told us what those magic words were up front? Or if they all worked on the same magic words…like "Open Sesame" or "Swordfish" or "Human being?"”
      https://www.newsfromme.com/2024/12/30/magic-words/

      1. SwamiRedux

        Now there's a use case for AI: use Open AI to talk to the company's customer service to find out the magic words. That way you don't have to go through the list.

        You keep notes in the phone listing for companies?

  9. different_name

    Despite its almost miraculous abilities, we're still struggling to find compelling real-world uses for it.

    I disagree. The most consequential uses for LLMs have already been found:

    - centralizing information control
    - cratering the cost of generating propaganda and poisoning public knowledge

    Whizzy miracle cures and bizarre-looking jet engines might be neat, but as folks use to learn those two capabilities, we will see a very different world. (Well, some of us. I rather expect more rather than less conflict over the next little while, too.)

    Poisoning the internet and putting more low-skilled folks out of work are just bonuses, I guess.

  10. SwamiRedux

    Here's a prompt I'd like to give a generative AI:
    Create a realistic avatar for me, train it on my past behavior in Zoom meetings, then attend future meeting for me. After each meeting prepare a 1-minute summary for me, and tell me if it's worth my time to read the summary.

  11. Pittsburgh Mike

    I've read descriptions from Cory Doctorow or Ed Zitron that describes the uses as writing pointless memos to pass up the management chain to someone who uses AI to summarize the memo.

Comments are closed.