I’m used to seeing articles about AI being used for either highly scientific uses or for generating semi-entertaining nonsense. For a personal business involving managing appointments, documenting meetings, tracking payments etc, can AI help with any of that? Other things include undertaking CPD training, occasional advertising as well as maintaining a website from time-to-time.
The people I know who don’t think AI has any use for them belong in this category and work in the area of mental health, yoga teaching / training, nursing and massage therapy.
Honestly, I don’t think it can in any meaningful way.
LLM are regurgitation machines prone to hallucination. Other tools exist for business management which are more suited for a sole trader.
Thank you. That doesn’t surprise me, but with all the hype I wondered if I was missing something.
AI right now is in that phase where we find out how to improve it and where to use it. It will be useful somewhere in the next 5 years, but not now. And I don’t even know if it will ever be useful in finance. We’ll see. Now is not the time.
The only reason to use it rn is if you’re a scientist or an enthusiast. You are neither I’d assume.
My company uses an LLM to summarize a report. Basically give it all the data and tell it what normal looks like so it can write stuff like “26,000 users this month, a minor decrease from last month. This is not an issue as it is within normal bounds.”
I don’t see the point, but management felt we had to put AI in SOMETHING so other companies don’t have a marketing advantage.
We’ve been working on anomaly detection for a while with Machine Learning, which is really more impressive but much harder.
You can use LLMs to, well, do what they’re designed to do - generate text. Need to write a marketing text? Summersie a meeting or make a summery more readable? Rewrite an “about” page to incorporate something new? Just be sure to read through the generated text and make sure it’s correct.
The only way I’ve seen it used is to write slop articles that are highly SEOed to drive traffic to a website.
You can use speech to text to transcribe meetings then have an LLM generate summaries. You can make an automated tagging system with the transcription or summaries as well.
I wouldn’t have it manage appointments, just in case of errors. I don’t know for cpd, not really sure what that entails. For advertising, image generation can be a real boon but it still requires someone with some advertising and photo editing knowledge to get good results.
It really depends on the specific task, it can fail monumentally for certain things but can be a real time and cost saver for others. It’s important to review what it does as well and test it thoroughly.
Likely depends on your definition of AI. Most of the replies seem to be about LLMs… But there are many other possible uses of machine learning for business… Forecasting trends in sales, customer churn prediction etc. Entirely depends on your business.
A very good point. Thank you.
I’m super excited to see machine vision for inventory management personally. A tons of trades is just finding the right connectors, plates, wire for a job.
Software can. Using intelligent, well structured tools can make a huge difference in the ability to keep organized.
But AI is mostly bullshit.
As of right now, there’s no AI product that actually do any of those things you need without a human in the middle to curate and correct the output. So you have to pay for AI and also a person to wrangle the AI.
For example, the company I work for has the whole office 365 package. It includes an AI enhanced automatic audio transcript. It is supposedly capable of transcription of Teams meetings and to generate a summary and minutes of the meeting. Except it consistently insists the meeting was held in German, despite the fact we only speak Spanish or English during meetings, we are not in Germany, nor we are anywhere close to Europe. It also can’t cope with low quality microphones, slow internet lag and sound cutoffs throw it out of whack and it requires all the meeting to be recorded and stored at MS servers (not our in-house exchange server) which means that it is useless for some meetings were confidential information is discussed and recording is disallowed.
It would take one of us a couple of hours to correct the transcript to make the summary function work and then make the minute and summary actually be in a useful format for us. Or, that same person could participate in the meeting taking notes, then use half an hour to write the minutes directly.
Your best bet to do those things you need is to find and pay for a good personal assistant. Cut the middleware.
Generating documents full of corporate jargon when the substance isn’t as important as the presentation.
LLM are used particularly to process big amounts of text. I remember my first encounter with it in 2009, somebody giving a talk about observing topics on Twitter, e.g. to track the source of fake news or figure out why some particular topic became viral.
You might already be using it regularly with a translation tool. Yesterday I just saw a foss app called receipt-wrangler, which uses LLM to parse shopping receipts, because a simple scan and ocr would still leave you with a highly unstructured heap of text, which is hard to parse into anything useful.
Stay as far away from it as possible. Let your competitors waste their time/money/effort buying into scams. When the bubble pops, you won’t have lost your shirt the way those of your compeditors who did fall for it did.
Promise you won’t use it.
Black hat and Defcon just ended and I’ll share my impression from LLM related talks given there. Microsoft VPs charged additional money to CISOs attending the summit talking about how AI will disrupt and be the future and blah blah magical thinking.
Meanwhile Microsoft engineers and others said things like “this is logarithmic regression for people who are bad at math, and is best for cases where 75% accuracy is good enough. Try to break use cases into as many steps as possible and keep the LLM away from any automation that could have any consequences. These systems have no separation between the control plane and user input, which is re-exposing us to problems that were solved 15 years ago.”
I think there are some neat possibilities that are lost in marketing hype as venture capitalist anger grows that they might have been scammed by yet another hammer in search of nails.
As others have noted, LLMs are prone to error and need review. They can be very helpful in taking large amounts of information and making it more concise.
To your question about how AI can generally be of use, AI is good at taking large amounts of information we humans normally would have issue consuming and reacting to and giving us reactions based on the algorithm we gave it related to the info. I’ll give you an example from my life that may make sense to you as a fellow small business owner.
I used to work for a large company that had a travel agent. I used to travel every week and was told to use the agency to book my flights. The corp has specific requirements about 2nd best cost and non stop vs 1 stop. I could end up with a crappy seat and extra flight stops and had no control officially. Back then I would use Expedia to stitch together what flights I wanted and then casually provide constraints to the agency so that I would get exactly the carrier, non stop flights and decent seats… But it required time to figure these things out. I was the AI taking all of the info thay mattered to me and crafted my result based on what I wanted. I would game the agency/corp’s constraints to get the one flight I wanted.
AI could be useful for managing your schedule (personal and work) to book sessions for clients. One could argue that how hard is it to just let clients book into your schedule, but consider Motion app. Motion can watch how you book clients, when you take lunch, when you normally start or end days, and help move appointments around or create availability on the fly. You could tell an AI app like motion that you need prep and clean up time, you take longer lunches two days a week, etc…
As with all AI you can do what it does. What’s interesting is AI can learn your processes and start to anticipate like a good Admin executive… But it takes training and maintenance. That’s the expensive part. Maybe time, maybe money, maybe both.
To make a long post longer, AI could help with marketing. It’s like the old school mail inserts, but with extra smarts. I’m sure there are AI marketing tools that take certain data and constraints and then shoot out various emails, calls, texts and posts to the demos you need at a frequency and amplitude that the tool thinks matches your business model. I’m speculating at this point tho.
Wrong question. “I have a solution (‘AI’), what’s the problem it should solve?” This is the path towards micromanaging stuff that’s not core to the enterprise.
Instead, try to identify specific problems in the specific context, or factors that are most relevant for success. Then see what the solution could be. That solution might be “AI”, or a bunch of sticky notes, or whatever else.
Other than that: Wherever you use a new tech like ‘AI’, also consider the risks. For example, do you really want to outsource part of your customer relations to an unpredictable thing that sends them the implicit message that you don’t care to directly communicate with them? Etc.
For your use case I would use ai to:
Summarize meetings creating built points.
Analyze client list habits and “predict” what services potential future clients will need.
Help with designing the website maybe incorporating new tools or features.
Help draft advertisements or maybe write a script for a potential television or radio commercial.
Use it to brainstorm thoughts and ideas to take the business to the next level.
Thanks. It has been helpful in drafting an advert - it had to be modified heavily but it did save time.
For summarising notes, there are privacy and IP concerns which supplying those as input sometimes.
