Why Traditional SaaS Pricing Fails in the Era of Agentic AI (Ashok Subramanian)

Full transcript

Welcome back to the podcast guys. Today we are joined by Ash from Carotex. Ash, welcome to the podcast. >> Glad to be here, Zach. >> Let's start simple. Tell us a little bit about yourself. >> I am a software entrepreneur. I started my career as a software engineer after my education in India. I worked in multiple countries like India, Baharin, Canada, and the US. I work for public companies like AT&T Labs, what used to be called Bellabs, Oracle, Aspent, and Do Systems, which is a French company. They have a US headquarters as well here, as well as many private companies. I was an entrepreneur and had a successful exit in my first startup that I started in Sunnyville, California with a couple of other co-founders. Since then I've been with some public companies and since pretty much 2018 I have been associated with kerotex the chairman of this company was my yet co-founder in sunnyale California and managed to keep in touch and we discuss pretty much every other day what we can do agentic or any CRMotex is not a single company it's a group of companies they do in manufacturing and assembly very big in e-commerce call printer, customer care and professional services and software development value generation predominantly has become marketing term and many many things I hear about in the world SAS and then they complain oh SAS is dead SAS doesn't work no you have to look into how you shape the software >> I read your post and you say that AAS isn't a new religion It completes SAS.

Can you explain the core idea and what problem are you actually trying to fix? >> Sure. It's not a brand new category as you mentioned right. It's a SAS when it is forced to evolve its monetization. Take AWS which is agentic AI as a service. How do you think about monetizing your product? Let's say somebody built a SAS sales automation, marketing automation, call center automation, customer car automation and what they built is a product. If they have 30 users, they pay for 30 users. So seatbased license is predominantly the only manner they talk about. When they add agentic, agentic is not human. Human beings are not logging in. Agentic AI is automation work. bot software bot operating at your customer premises agent if it is a company how could you deliver agent as a service customers think oh this is the magic no it's not magic how do they pay for it if you run it for 1 hour you pay money they say go fly kites you know I cannot give my data to you into your SAS so you pay me $5,000 for this month what did you do for $5,000 you Understand >> what you are trying to say is that the seat pricing punishes the vendors for making customers more efficient. >> I don't mean to use punishes.

It confuses the customer. They don't know what they are paying for. And suddenly your very intelligent person always think it's a magic. I don't believe in magic and I'm not paying for magic. You got to be transparent about what you are going to be charging for. The only weapon they had is per user charge. Okay. 10 users. Okay. Pay uh they say 10 user license, 20 user license, 100 user license is very easily believable. If you say agent is magic, no. What agent? I'm not giving you data. Number one. What is magic? They don't want to pay for magic. Number two. Then are they buying a product? No, because agent it can go from sales to support to marketing to customer care. It can go across product. That's a workflow, right? It's a workflow stream goes from product to product to product. You cannot charge for one product either. You understand? It's not per product base. It's not per user base. Then what is it based?

You cannot cheat the customer. You have to convince the customer that when we send you a invoice, we send you with authenticity, the usage record. How many times from lead to generating a paying customer, how many times our software helped you to automate going from this end to that end? We helped you 3,000 times in the month of January. So 3,000 times $20, $50,000, something like that, right? Wow. 3,000 times you transformed a lead coming to your website into a paying customer. That's the workflow usage record. On January 6th, 10 customers on January 7th, 25 customers. Your invoice is going to have more workflows named workflow strains to prove why are you charging them that much you are putting on the invoice. If you show them 60,000 oh we really cannot afford 60,000 can you work on a discount then it is between it is between you the software vendor and the customer to negotiate a term if you want to be tough pay now and then we will take away then then it becomes a negotiation but when AI comes into play there are two things customers are not going to give you the data you have to take the logic to the customer >> you are saying that you shouldn't price for work executed.

>> You should price for work executed only. Many customers still pay for users. They dance around. They already have users. They added agent. They didn't think about agent in the beginning. It's an add-on. When they add it on, customer get confused. Let's say take a customer called ABC. Example, hypothetical customer. ABC is a software vendor. CRM software they developed they have uh about 100 users from a sample customer called XYZ let's call it XYZ or XY is right is a customer they got 100 paying users per month and then ABC automated lot of automation agentic AI which can help customer to transform from lead generation ation into paying customers. Lot of things like this. When they automated the XYZ customer for the second month, they are not using 100 users seats users anymore. They were using only 60 of them. What happened? They just reduced 40 users. Usage reduced because the 40% is filled in by agentic AI. The only blind mechanism the ABC company does is increase the price on a per seat. That's a blind right now customers say don't increase. If you increase I will go away.

I will churn out because there are other alternatives they have. So how do they price it? How do they package it? 40% of the users don't have to log in because it is done by the bot. read that you don't sell agents, you sell named workflow streams. What do you think makes a workflow stream actually sellable and adoptable in this day of age? >> By saying named workflow stream, you can convince the customer, hey, this workflow stream will transform them from lead to paying customer. That's what this workflow stream number one does it to you. When that stream executes for every execution pay $20, you are not saying, "Oh, gimmick. Our sales product is going to generate value, is going to generate your leads into paying customers very quickly." Yeah, I heard about before. This is nonsense, right? >> It's like buzzwords, right? >> Yeah. Buzzwords, right? Buzzwords and marketing words. They have been overused and customers are tired of hearing from it either. But this agentic AI is lot has got lot more promise and it can try various avenues. If one approach doesn't work, it can try the second one.

They're all validated leads though. They're not cold calling. The email is valid. The phone number is valid of the lead. Customers have a reason to believe. You will know over a period of time how many seats you have to purchase. So gradually you get to increase the number of agentic AI services you have and customers will start paying for workflow streams that go across products. So product managers don't have to be insecured and quit the job or they are laid off. You understand? Product manager don't think oh sales is my product, marketing is my product. They don't collaborate right now because they hold on to their domain. They are your colleague. Now what this opens up? Agentic AI is not one product. It can cut across many products. What is cutting across many products? Able to define and design a workflow stream that cuts across many products. What it means Zach is very simple.

Every line of code my organization develops, I can very creatively come up with workflow streams to touch every single line of code. And I can go completely agentic then I can say agentic AI as a service. They don't have to dance around. Agentic AI is nothing new. It complete SAS. SAS is software as a service. It all depends on how crafty you are, how creative you are to package. I'll tell you one quote. Jim Box Day, CEO of Netscape from the Silicon Valley, a great entrepreneur. He said one particular phrase that remember I'm paraphrasing it now in software you make money either by bundling the software or unbundling it's all about your creativity how we package it how you unpackage it that's why product management >> exactly product management has to play that creative role >> who pays when the agent misses when it hallucinates for example or crashes or there is some kind of bug. >> What should you do? >> Exactly. That means we have to be diligently transparent as ABC software company. We have to eat the cost. When you do that, customer will trust you.

Sorry, I made a mistake. You got to say, let's say there's a soccer match going on. Somebody missed a penalty kick. The person missed a penalty kick cannot give excuse because hundreds of cameras looking at them. I have to scratch my face. That's why I missed the goal. No, you cannot do that. You got to be transparent. When the agent is running, customers are not looking. So, there are no cameras watching you. You got to audit your own work and be diligently transparent. If I tell 10 times I messed up, uh next month that they will not come to me, they go to somebody else. Don't lie. Tell the truth and be diligently transparent. So, give the usage record. Here we missed because it could be a network latency issue and you have to provide the evidence the payment did not go through because of network latency and this is the evidence that's why on that time our software was not able to complete the payment transaction I'm just creating a scenario right a payment did not go through we apologize it is unfair for you to ask the customer to pay for it that's why the software got to eat the cost if it is a network latency something that you cannot control.

You have to put it in the contract which is trustable. So these things are critical because it gives a credibility. Yet when the agent misses you got to pay up currently you know what they do they do this or they they believe in God they go and pray and agent can hallucinate too. Who's paying for it? You got to take it. So by taking agentic AI into your business you are taking the responsibility when it does some mistakes you need to pay for it. If you do that it brings credibility. I cannot be sitting idle and twiddling my thumb. I have to do whatever I have to do to make my platform my offering better and better and better. But bottom line is construct workflow creatively. Assign a dollar value or euro value on how you are going to monetize it. Be diligently transparent with the customer and tell them I'll be sending you an invoice. This many executions will cost this much. Give them examples. You're not selling we generate value. No marketing buzzwords anymore. You are very clearly articulating leads to paying customer. I got you 10,000 in the month of January 2026.

Pay $60,000 $100,000. Some SAS companies if I help you to generate 10,000 customers for every customer give me 10% of whatever you are taking from all those loosey goosey business school bling bling is not going to work. are the same. Every single idea you come up with, you articulate it so that customers will get the best out of it or you spend more and more money automating it and no customers yet and you are expecting them to pay per seat and good luck. >> Yeah. Yeah. completely understand your perspective like transparency is key in anything not just business >> but so many companies have so many good ideas they everybody has a chance to win just be creative and come up with a workable monetizable workflow stream come up with how you want the customers pay for it you want to be able to make money out of every single single line of code you write it is across the board workflow. >> Okay, Ash, thanks for coming on. I will add links in the description so people can check you out.

Thanks everyone for watching and we will see you in the next one. >> Thanks folks. Appreciate it. Bye now.