AI-Based Business

[Dannydee]

I have 4 years of marketing experience and attempted online business several times. Those business attempts did not do great but the last one technically gave me credibility to transition into the remote marketing career I'm in now.

Anyway, I have had my eyes on AI for a while and now I want to attempt another online business utilising the power of it. Initially, I turned away from the idea of building chatbots, but I think I have a viable idea for one now that I believe doesn't exist in the market in which I intend to enter.

It's probably not wildly different than anything you see on most websites that you hop on to, but it would be quite unique to the world I'm talking about here. I think it's worth exploring.

Has anyone tried anything with AI in this way? Or used AI to help start an online business? What tools should I use or could I use for this?

I'd appreciate any advice on this. I will of course answer more questions as I know I've been rather simplistic in what I've put here.

 

[fisicx]

There will be those who think it’s a great idea and those who think the opposite.

I’ve not yet seen a successful application of AI to anything webby. Chatbots in particular have failed in almost all cases when AI has been used.

 

[WaveJumper]

Have to say IA chatbots drive me mad, had an issue using my server software at the weekend and thank goodness I actually spoke to a human. Docent really answer your question but as mentioned above there are those who swear by them and those who don't, what I am sure about is they not going away anytime soon so please make yours is user friendly

 

[AndyReco31]

I've seen plenty of chatbots that train the AI on company documentation, is that what you are thinking?

 

[gpietersz]

Its impossible to say anything useful without knowing why your idea is different from all the other "AI" chatbots out there.

I have done "AI" several times in my life, starting with a holiday job in the late 1980s, and most recently just a few years ago. Nothing I have seriously worked with is what is currently being called "AI"! They might be called GOFAI though!

 

[Dannydee]

I’ve not yet seen a successful application of AI to anything webby. Chatbots in particular have failed in almost all cases when AI has been used.

Have to say IA chatbots drive me mad, had an issue using my server software at the weekend and thank goodness I actually spoke to a human. Docent really answer your question but as mentioned above there are those who swear by them and those who don't, what I am sure about is they not going away anytime soon so please make yours is user friendly

I use Hubspot in my main job and recently used the AI chatbot, which was really useful. So they aren't all bad. For sure, they are still in relative infancy but they must be getting better.

... but yes, I will do my best to make the UX of mine nice and pleasant

 

[Dannydee]

I've seen plenty of chatbots that train the AI on company documentation, is that what you are thinking?

I guess conceptually we could say it's along those lines. This case would be more about industry and company information/data. I'm a beginner with this, and learning, so it's probably wayyy more complex than I understand at this point.

 

[Dannydee]

Its impossible to say anything useful without knowing why your idea is different from all the other "AI" chatbots out there.

I have done "AI" several times in my life, starting with a holiday job in the late 1980s, and most recently just a few years ago. Nothing I have seriously worked with is what is currently being called "AI"! They might be called GOFAI though!

I expect you will understand and appreciate my reluctance to give more away than I should since I believe the angle I'm coming at with this might be relatively new to the space I'm entering.

As I said, it's probably not a ground-breaking new development. The technology is the same, but I believe the direction I want to come at this has not been done yet. So I want to take it before someone else does.

That's what I think, anyway. I need to get stuck in further and see what I come up with.

 

[fisicx]

so it's probably wayyy more complex than I understand at this point.

It is. And needs a lot of resources to implement. Which means a lot of money.

 

[Dannydee]

Which means a lot of money.

... that's not fun to hear

I'll just have to see how far I get on a shoestring and see what happens/go from there

What should one do in such a circumstance where there isn't a whole lot of capital to work with currently? Forget it and do something where the financial barrier to entry is lower?

 

[Paul Kelly ICHYB]

Here I am, in Tenerife, having a fantastic time with the family. In 5 days, I have given business pointers to 3 people (one might become a client/mentee!) and told one person to keep their day job.

With several failed projects behind you, I would say analyse why these failed. Sadly, it may be a simple as self employment/starting a business is not for you.

Yes, start with a business plan. Expect the best but plan for the worst

 

[ecommerce84]

What should one do in such a circumstance where there isn't a whole lot of capital to work with currently? Forget it and do something where the financial barrier to entry is lower?

Can you raise capital? Bank loan, friends and family, 2nd job?

Without sufficient capital most businesses are destined to fail from day 1, not least because of marketing costs - getting a product in front of the people that are going to buy it can be expensive and you may not even break even for years.

IMO lack of capital is the main reason that many small businesses fail - sometimes because they have a great product and nothing to market it with and other times because you open that shop you’ve always dreamed of and don’t have enough to cover your living expenses for the first 2-3 years so you end up closing it down and getting a job so that you don’t end up losing your home as well.

 

[Baines Watson]

Before you even begin to explore 'how' to solve the problem you are trying to solve with AI, make sure your answer to the 'why' question is solid, if it isn't then there is no problem to be solved and therefore no viable business plan to explore further.

 

[Kerwin]

How good are you with maths and programming? Without those skills you won't really be able to achieve much. Bear in mind that most people working on AI have at least a BSc or a masters or even a PhD.

 

[Mark T Jones]

I expect you will understand and appreciate my reluctance to give more away than I should since I believe the angle I'm coming at with this might be relatively new to the space I'm entering.

As I said, it's probably not a ground-breaking new development. The technology is the same, but I believe the direction I want to come at this has not been done yet. So I want to take it before someone else does.

That's what I think, anyway. I need to get stuck in further and see what I come up with.

I think there is only one question that I - a potential client - would need to know the answer to:

Why should I care that your not groundbreaking development is a different angle?

 

[gpietersz]

I expect you will understand and appreciate my reluctance to give more away than I should since I believe the angle I'm coming at with this might be relatively new to the space I'm entering.

Not entirely. Its not usually ideas that make a business successful, its execution and fair bit of luck. I have done a lot of consultancy and implementation work over the years, for many different businesses, and I have come across only one idea that I thought worth keeping confidential (it was original, there was no competing existing product, and it was relatively easy to implement).

The problem is that without more detail no one can give you any useful advice. Its impossible to say whether "AI" (by which I assume you mean an LLM) is an appropriate way of doing what you want to do, nor how much work it will take. A lot of people are just writing thin wrappers around big providers services (like ChatGPT), which is cheap, but has no barriers to entry so if it works expect competition or potential customers just implementing it for themselves.

As I said earlier, it is execution that really matters, and you do not seem to know enough about the technology to be able to be able to hire the right people. I think you need to find someone you can trust, is competent in the field, has a wide enough view to tell you whether LLM is the best way to do it (rather than older natural language processing or whatever), and can oversea the implementation. Given people with that set of skills are in very high demand right now, its going to be hard to get one interested, and given there are a lot of snakeoil salesmen selling AI at the moment its going to be hard to idenfity someone competent too. Its hard enough to get those right in tech fields that are well established. With "AI" its going to be impossible.

 

[Nick@Daydot]

I think there's a good chance that whatever your idea is that someone is already working on it. If seen a lot of chat about chatbots and, given how crap many of them are, AI if done well could be an obvious way of doing them better.

The normal way that a successful business would be set up would be to have a business plan and some sort of demo, enough to get an investor. You'd need a generic demo that's good enough to persuade a potential real client to let you train the AI on their content and then try it out. You'd need success metrics and a load of user testing. As well as the technical resources to build and run it.

Or, you could cobble something together and get very lucky with a launch client who funds the development if it really is something special. That doesn't happen much and is a good way to lose money if it doesn't work out.

 

[AndyReco31]

I guess conceptually we could say it's along those lines. This case would be more about industry and company information/data. I'm a beginner with this, and learning, so it's probably wayyy more complex than I understand at this point.

Check this guy out @pbteja1998 on X he has built an AI chatbot called SiteGPT, might give you some ideas.

 

[fisicx]

Check this guy out @pbteja1998 on X he has built an AI chatbot called SiteGPT, might give you some ideas.

Just had a play and it's dreadful. In some way's it's worse that a normal chatbot. Certainly not worth paying $39/month.

 

[gpietersz]

Just had a play and it's dreadful. In some way's it's worse that a normal chatbot. Certainly not worth paying $39/month.

I did too. I asked it to compare itself to NLP based chatbots and every answer was prefaced with "often" and "may not offer"

It also refuses to answer any generic questions about chatbots.

 

[fisicx]

Every response suggested I visit the website I was currently viewing. And I agree, some of the answers were very vague or evasive.

 

[JEREMY HAWKE]

I am dumping Vonage due to their use of chat box only
Imagine if it was fully AI based nobody would deal with them

 

[pavp]

Here's some really practical things you can try:

1. Validate the business idea further: speak to customers, describe what your product will do.
   1. Mockup what a workflow would like like eg "problem" -> "solution" workflow steps, drawn with basic boxes in Miro
   2. What's the excitement in the meeting? would they 100% pay for it if you had it tomorrow? Would they pay today (prepayment) for you to create it? (thats what a 10/10 looks like here)
   3. Understand how they are solving this problem today? Google sheets / people / stitching 4 things together - that's good it indicates pain. But eg: there's one solution which handles it all for them, that's not so good (at least for this person and customer profile)
2. Start light: setup a chatbot on a no-code builder like relevance.ai or botsonic. When you get to a strong pull on 1. tweak the workflow
3. Start niche: you said that your solution is novel for your market - that's perfect. It should make 1. (validation) easier: you have a small, defined target group of people AND less competitive to build

There are awesome implementations of AI chatbots for business usecases: if anyone's requested support from Intercom, they've likely interacted with it. And for marketing software like Jasper and copy.ai are great.

We are extensively using AI to help small business owners create websites and start marketing their products - and a lot comes down to who are you targeting, how and with what features.

Good luck!!

 

[fisicx]

There are awesome implementations of AI chatbots for business usecases:

Can you give some examples of these?

I tried the intercom chat and it was awful.

 

[pavp]

can't off the top of my head, but remember 2 customer support chatbots and they were using intercom and my question was answered by a LLM/AI reply from information in their Helpdesk Article / Blog pages - this is a key usecase for AI chatbots as relieves pressure on customer support people for queries which the user could have answered and for me. Great experience without too much searching!

 

[gpietersz]

can't off the top of my head, but remember 2 customer support chatbots and they were using intercom and my question was answered by a LLM/AI reply from information in their Helpdesk Article / Blog pages - this is a key usecase for AI chatbots as relieves pressure on customer support people for queries which the user could have answered and for me. Great experience without too much searching!

I find the opposite. It is very frustrating to find a chatbot repeats information that I have already read because its on their website, but I cannot get through to a human without wasting that time with the chatbot first. The DVLA are a great example of this.

Its even worse than listening to messages telling me I can do things on someone's website or app when the reason i am phoning them is because the website or app told me I needed to phone.