Stories
Slash Boxes
Comments

SoylentNews is people

SoylentNews is powered by your submissions, so send in your scoop. Only 10 submissions in the queue.
posted by mrcoolbp on Thursday September 24, @12:18PM   Printer-friendly
from the sauce-for-the-goose dept.

There's no reason the 'AI revolution' shouldn't also impact the C-suite, right ?

A group of developers who were let go due to AI efficiency savings have struck back at the executives who are often shielded from AI-induced layoffs by creating OpenExecutive, an AI tool designed to augment and replace executive decision-making, TechRadar reports.

"Work happens around the clock. Leadership doesn't scale with it," a demo video says. Using the kind of language that executives have grown comfortable throwing around about AI replacing workers over the past years, it suggests the bottleneck to productivity and enhanced profitability at companies isn't a lack of worker enhancements; it's that leadership can't keep up.

So why not replace them with AI ?

OpenExecutive presents as a singular AI for interactive purposes. However, in making its C-Suite-grade decisions, it leans on a virtual board of eight AI agents designed to emulate experts within an organizational structure. The singular voice is backed by an emulated CSO, CFO, HR officer, COO, marketing chief, a product officer, and a board comms director. Each of them can lean on the main executive AI to guide its actions and suggestions.

They can all draw from company documents and data, remember their decisions, and trace back reasoning for that if you want to follow up on why they decided to do certain things.

[Source]: PCMag


Original Submission

Related Stories

BMW Just Announced It’s Replacing Several Management Roles With AI 20 comments

BMW announced Wednesday it planned to slash its management structure by 20% while expanding the use of artificial intelligence across its operations:

The luxury carmaker said it will reduce the number of divisions and associated management roles by 20% by mid-2027, with comparable reductions at lower organizational levels. BMW said AI will play a central role in making the company more efficient, speeding up decision-making and automating routine work across development, manufacturing, purchasing, sales and other operations.

The restructuring comes as BMW faces weak European demand, growing competition from Chinese automakers and U.S. tariffs, while its shares have fallen more than 30% over the past year, according to Reuters. The automaker is also seeking to restore its automotive operating margin after it fell to roughly 2.3% in its latest results.

"Consistent use of agentic AI applications across all areas of the company will be a game-changer for more agile and efficient development, leaner structures and faster decision-making," BMW finance chief Walter Mertl said.

[...] The company already uses AI throughout its factories for digital twins, quality inspections and autonomous transportation systems and plans to have digital AI agents perform increasingly challenging tasks autonomously.

Related:


Original Submission

This discussion was created by mrcoolbp (68) for logged-in users only, but now has been archived. No new comments can be posted.
Display Options Threshold/Breakthrough Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
(1)
  • (Score: 5, Insightful) by esperto123 on Thursday September 24, @01:01PM (4 children)

    by esperto123 (4303) on Thursday September 24, @01:01PM (#1454772)

    The easiest jobs to replace with AI, after the junior code monkeys, is the C-suit, AI can work 24/7, is not inaccessible while traveling, less prone to taking decisions to benefit itself, would have more impartial decisions, can have its objectives easily adjusted by the board, easy to blame if something goes wrong, would probably hallucinate less and will not try to fuck the secretary!

    • (Score: 4, Informative) by VLM on Thursday September 24, @01:15PM

      by VLM (445) on Thursday September 24, @01:15PM (#1454776)

      I've delt with a company experimenting with this and the biggest problem is GIGO, if your business plan is not up to date or explicit instructions are missing, the LLM will suggest pretty stupid stuff. Humans would be too embarrassed to ship BS but its kind of a LLM's job to try its best.

      You'll get lots of suggestions to work extra hard on cancelled projects, nobody has been working with that client (that closed, ceased operations) lately, etc.

      I remember a funny one where it complained its an emergency that no one at that employer is processing accounts rx (from a former client) LOL.

      Another funny one was some suggestion for a technical improvement, that was a good suggestion, it just wasn't documented online and had been implemented years ago (I think it was MFA... that client had it, it just wasn't explicitly in their business plan and to do lists, it was just 'how business was done' so the LLM suggested in the current year it might be a good idea to pilot program MFA lol.)

    • (Score: 5, Insightful) by Anonymous Coward on Thursday September 24, @09:44PM (2 children)

      by Anonymous Coward on Thursday September 24, @09:44PM (#1454810)

      I get it’s fun to joke that C suites can be easily replaced by artificial intelligence, I’ve been wheeling that out too, on occasion.

      But I don’t actually want it. Putting AI agents in a place of power over people is absolutely the wrong thing to do in my opinion.

      • (Score: 2, Informative) by aafcac on Friday September 25, @06:51AM (1 child)

        by aafcac (17646) on Friday September 25, @06:51AM (#1454823)

        Over people? We're talking about replacing C suite executives, not people.

        • (Score: 2, Touché) by Anonymous Coward on Friday September 25, @11:27PM

          by Anonymous Coward on Friday September 25, @11:27PM (#1454892)

          I mean the AI agents being in charge of real people.

  • (Score: 2, Interesting) by VLM on Thursday September 24, @01:07PM

    by VLM (445) on Thursday September 24, @01:07PM (#1454773)

    You can do this with ClickUp, assuming your company/client uses clickup, and you'll already have your full to do list and business plan and records of historical workflow for the agents. I've seen this with my own eyes and it works pretty well.

    Got some interesting ideas from this.

    Its certainly hyper verbose when set up this way. Just paging thru to get some ideas is difficult.

    One thing thats EXTREMELY unrealistic is I've worked at megacorporations that stack rank management so their main enemy/competitor is other departments not competing "in the market" so decisions made by AI will be very unrealistic compared to IRL. The AI will maximize total company provide, minimize internal friction between departments, its all very unrealistic, so companies with that "reward system" can't use AI assisted decision making. "I don't care if it'll make the company $10M, I won't make my $100K bonus so we're going to sabotage the other department procedurally" etc. So its kind of 100% AI or 0% AI.

    I have used LLMs for planning. "Turn this yearly project plan into a monthly plan" then turn around and "turn this monthly plan into a weekly plan for this week" and that works VERY well.

  • (Score: 4, Funny) by chucky on Thursday September 24, @01:11PM

    by chucky (3309) on Thursday September 24, @01:11PM (#1454775)

    Where will they get the physical part? So far I didn’t read about any humanoid robots which could play golf.

  • (Score: 4, Funny) by DadaDoofy on Thursday September 24, @01:24PM

    by DadaDoofy (23827) on Thursday September 24, @01:24PM (#1454777)

    I, for one, welcome our new c-suite overlords.

  • (Score: 2) by Username on Thursday September 24, @01:48PM (1 child)

    by Username (4557) on Thursday September 24, @01:48PM (#1454780)

    Now all the C$$ people are going to have the AI run the company while they do nothing but spend money on hookers and blow entertaining investors. It's not like they will get fired. It's a club, a social cast, a group. You can either join them or work under them.

    • (Score: 4, Funny) by krishnoid on Thursday September 24, @05:20PM

      by krishnoid (1156) on Thursday September 24, @05:20PM (#1454796)

      while they do nothing but spend money on hookers and blow entertaining investors

      But can an AI properly identify the part of speech of "blow", and its subject if it's a verb?

  • (Score: 5, Insightful) by Thexalon on Thursday September 24, @02:19PM

    by Thexalon (636) on Thursday September 24, @02:19PM (#1454784)

    There are 3 modes the ExecuBrain needs to have:
    1. Subordinate Mode: Tells you to work harder, no matter what the situation. When picking out raises and promotions, simply randomly selects from eligible people.
    2. Investor Mode: Always says that we're on a path to massive future growth, no matter what the situation.
    3. Sales Mode: Promises the client that of course their company's product or service can give them the moon, sun, and stars, and whatever else they want. The legal department will ensure that there's no consequences for this.

    I'm quite sure an LLM can do this.

    --
    "Think of how stupid the average person is. Then realize half of 'em are stupider than that." - George Carlin
  • (Score: 5, Insightful) by VLM on Thursday September 24, @03:18PM

    by VLM (445) on Thursday September 24, @03:18PM (#1454789)

    I looked it over while having breakfast tea, not super in detail not overly vague.

    1) There's a huge architectural mistake right at the beginning "a single coherent executive persona backed by 8 specialist agents" My client uses their agents as employees, if you assign a "Marketing Manager" or a "Operations Director" to a task you'll get wildly different answers if in context you ask it what the impact of the task will be. Or scheduling issues resulting from pushing this up or down a week, etc. They're missing out on a huge opportunity, its all just bits and credits run each in parallel. There's less of a SPOF and you can safely experiment with each separately. This also helps with access control. If you let anyone tell any AI agent what to do, they'll ... do it and you probably don't want that. Access control is a pretty big deal if you use this for real. Its fine if the CEO has an open door policy but you can't let project manager abc order the ceo that all of the abc projects are top priority and F projects for xyz, because project manager xyz will freak out.

    2) Can connect to MCP so it can do "anything" but someone has so write/alter the backend which looks painful. Its easier to implement if you start with a manual-ish workflow automation system that has AI bolted onto it as a feature, than to have AI system with workflow automation added to it later on IMHO.

    3) There's some substantial flowchart level mistakes. Note that company documents (business plan, current project planning system etc) is mixed in at the end after the conclusions were made. Which unfortunately matches failing legacy corporate flowcharts. A better way, since AI is pretty good with large input contexts, is to mix the "company documents" in at THE BEGINNING not the end of the flowchart. Or in some cases of approximation run a loop like three times optimizing the answer including passes thru other agents, which my client does.

    4) I get the feels that this is a second system. First isn't nearly enough, Second goes wildly overboard, third will be just right, although it doesn't exist yet. Not necessarily a problem if you know going in, but just saying, you're buying into what feels like someone elses second system so its going to have the usual outcome of a second system. Oh well.

    5) Not sure why they're using ChromaDB if you just want to store MD files full of stuff, stick in in a MCP equipped wiki/knowledge base (notion?) or stick it in a MCP front ended file server. Its more open for debugging and backing up that way than stuck in ChromaDB. (throws hands up in the air) "Whatever". Client does a lot of debugging their agents using plain text in a knowledge base of docs.

    6) The onboarding process is hilarious. One of the agents my client has is a virtual "competitor analyst" who does the grunt work of researching competitors. The idea of using humans for grunt work like that to set up automation is kind of funny they got the cart before the horse. A "lot", if not all, of the onboarding process should be done by the AI agents themselves. They need an agent that whines about data not being included "report the most impactful missing data to be added next". If the agents don't have access to your business plan, thats a BIG problem day to day, not a mere hassle to type in by hand during onboarding. They got this ALL wrong.

    7) The client IRL doesn't allow writes to anything without explicit approval to as a top level unalterable rule. They probably need something like this and I missed it if it exists. Its probably a very good idea that the agents can't modify stuff on their own.

    8) Its missing day two, day after deployment. I assure you the second day after a new agent is added is total chaos. Things settle down fast. This is another point in favor of #1 above roll out one agent doing one role at a time.

    9) I found the runtime for the evals, where'd they hide the actual evaluations?

    10) Ha ha added functionality for tailored personalities. That had to get banned at $client because people tried to make a joke out of implementing their favorite Shark Tank tv personalities and useless stuff like that. Sure financial reporting doesn't work, but the new addition was one of the agents gained a tagline along the lines of "and because of those reasons, I'm out" and something about Mr Wonderful and similar stupid stuff.

    11) They seem to have built it for Q+A whereas my client seems to have built their agent family / posse / gang / mafia around document processing, mostly batch process. The operations manager agent will look at new feature type tasks in the project tracker and then look at existing schedules across the entire company and express an opinion that may or may not be followed about schedule adjustments to other things if the due date is already set or it'll set a due date for this based on capacity and priority (I didn't code it, but I see it in use; it does alright...) Whereas these guys seem to be implementing the "Shark Tank" Q+A section of the show.

    12) They need time triggering. My client's process automation system does this for free; these guys need something to kick off scheduled "meetings" etc.

    I don't know if it made my breakfast tea taste any better or worse, but clicking around the github repo was at least entertaining this morning. Yeah I'm wide awake now LOL. Has anyone else seriously looked at this? I know a certain "Mr Merchant" here who uses a lot of AI for enduser programming tasks, so surely he has an interesting opinion on this very different, but AI related topic.

  • (Score: 3, Funny) by cmdrklarg on Thursday September 24, @09:58PM

    by cmdrklarg (5048) Subscriber Badge on Thursday September 24, @09:58PM (#1454811)

    You don't need any fancy AI to do that. All you need is a Magic 8-Ball.

    --
    The world is full of kings and queens who blind your eyes and steal your dreams.
  • (Score: 2) by jb on Friday September 25, @06:42AM (1 child)

    by jb (338) on Friday September 25, @06:42AM (#1454822)

    If executives are replaced with AIs, then who will pay all the bribes?

    • (Score: 3, Informative) by Dr Spin on Friday September 25, @09:12AM

      by Dr Spin (5239) on Friday September 25, @09:12AM (#1454832)

      The usual suspects.

      --
      Warning: Opening your mouth may invalidate your brain!
(1)