Thoughts on Fin Pricing?

I was extremely dissapointed in the pricing model. Our support team is offshore and, at $12 per hour, is comparable to Fin. I was thinking the pricing would be less than 10 cents per resolution.
I’m going through the trial and have other issues. Posted that here (better support for providing answers when your company has multiple products).

Thanks @Chris C11 - I share your disappointment. For what it’s worth, our expectation was that pricing would be in the $0.10 - $0.20 range as well.

@Jacob Cox appreciate the follow up on the pricing. I want to clarify that the charge per resolution includes both those confirmed by the customers with a “thumbs up” as well as no response. Based on the evidence I have seen in my account 100% of the “no responses” in the trial period could be considered as not resolving the customer issue.

So for this to be truly priced at cost I can only assume OpenAI is charging Intercom $1.90 per resolution as well. That seems extremely high and I would recommend Intercom negotiate that price.

I absolutely think that the price of 1,90$ per resolution is completely out of the market. We could hire directly a human agent to solve issues manually at that price. I think that at this price, Fin will exclusively be used by extremely small and well budgeted companies.

Hey @Eli & @User511 (And any other people reading this message!)

Mark from Intercom’s product team here 👋

Its very early in this wave of AI technology. As @Jacob Cox mentioned, Fin is currently priced at cost but we absolutely expect costs to come down in the not too distant future. We are determined to make great AI powered customer support accessible to all companies. Price shouldn’t be a prohibitive factor in most cases.

Would you be open to getting on a call to help me understand your POV better? My email is mark.nealon@intercom.io if you’re open to chatting. I can work around times that work for you 🙂

@Jacob Cox 

You mentioned earlier

Our Custom Answers feature is a great tool to use to help costs scale for your repetitive or more advanced questions as it's priced per seat versus being priced per resolution. You can set your Bot up so it will answer these repetitive questions before the AI Answers Bot kicks in, reducing the volume of questions that are going through that “per resolution” path.

How do you set custom answers up so they answer before AI? Does it default to custom answers already and only switches to AI answers if there isn’t a custom one set?

Hey @Cary Melly 👋🏼 Shauna here from Support! Firstly, apologies for the long wait - your question seems to have slipped through.

Currently it’s not possible to put a cap on the cost of Fin. What I will do is flag this with our Team 🚀  Appreciate you sharing this with us and for providing a little bit of context too! 

Shauna | Customer Support Specialist at Intercom

Can I plug my own openAI API key into Fin? 

Your pricing of $0.99 is too pricy for what's barely even a 500 token response.

For example, I can have my chatgpt 4.0 API give a 2,500 character response and that would only cost me $0.06 for a custom trained model.

Your current pricing of $0.99 per resolution would equate to roughly 41,250 characters, or roughly ~6,875 words (according to OpenAI’s Tokenizer). We can all agree, very rarely will any "resolution" take ~6,875 words.

With that in mind, Fins pricing model is definitly not "At Cost" as you claim here -  https://community.intercom.com/fin-tips-and-tricks-connect-with-the-intercom-team-54/thoughts-on-fin-pricing-4479

Until we can get the resolution cost to less than 0.25$ we are more tempted to plug our own OpenAI API key in and pay as we go.

Hi David,

Thanks for your comments.

No, we don’t support this, and have no plans to.

It would be quite complex, and difficult to support on an ongoing basis.

There’s a lot going on under the hood in Fin to deliver a high-quality experience.

We’ve got to embed the support content, retrieve it, manage many parts of the conversation not just the answer, try and run at professional standards of latency etc.

There’s really a lot more going on in Fin than just a single 500 token response.

As just some examples - sometimes it takes multiple responses to resolve a users query, and we’ve to pay for each of them; or sometimes Fin gives the best answer that’s in the help center, but it doesn’t fully resolve the users issue, and so we don’t bill; sometimes Fin does a great job but the user has a second query they need to talk to human support for, and that isn’t counted as a Resolution, etc.

These and other factors also vary across our customers!

So, we’ve got to take all of those into account in our modelling. I think if those are taken into account, with the factors I mentioned above, its not as expensive as it might seem from a more simple calculation.

In addition to trying to deliver great value, I also feel that it compares very favourably in terms of the overall cost:quality ratio to most alternatives, including human support; obviously there might be unusual cases though, and while we believe its a very fairly priced product, while we don’t provide a ‘bring your own API key’, there’s our API and app store for customers who want to do something different.