Q&A: Agentic pricing - Startups weigh in on the top question on everyone’s mind

As enterprises look to agentic AI to augment their operations, the topic of how to pay for and price agentic capabilities has become a top area of focus for everyone involved, from startups to incumbents. For decades, enterprise software pricing was designed around human-led execution: per-seat licenses, human productivity assumptions and predictable labor-based cost structures. In that dynamic, software supported the workforce. Now, the idea of agentic “labor” is upending longstanding conversations around how to price these technology offerings and all parties are adjusting.

Organizations across industries are asking foundational questions: What is the unit of value when the executor is digital? How do companies measure agentic vs human outputs? And how do startups charge for systems whose value comes not only from execution, but also from their ability to accumulate context, learn from historical behavior and continuously improve? To help answer these questions, Citi Ventures spoke with Avinash Misra, CEO and CoFounder of Skan AI, and Nikola Mrkšić, CEO and Co-Founder of PolyAI, two leaders building the foundations of enterprise agentic execution. Both Skan AI and PolyAI are Citi Ventures portfolio companies.

Citi Ventures: Why is pricing for agentic systems so challenging compared to previous generations of software?

Misra: There is a whole shift that is happening. Over the years, we have developed pricing mechanisms in a theater of execution that is essentially human-led, and humans were augmented by software, so you had per-seat pricing. And now, suddenly, the software itself becomes the execution, so therefore you are forced to reckon with many forms of pricing.

The challenge in all these mechanisms comes down to who understands the outcomes that are being produced and who understands the process by which those outcomes are being produced. Because I think that’s where the levers of pricing lie.

The depth of understanding of work and the outcomes produced with all the nuances, exceptions and permutations allow you to price properly. And I think that this whole bottleneck is the continuum of what is work, how is it organized/orchestrated, and how it is delivered and measured.

Mrkšić: A lot of the talk in the industry, especially by some of our competitors, is about how to price, and they act as if it’s simple. The truth is, it’s anything but.

Startups have to go and build a financial model that is really from the customer’s perspective, and people don’t always want to share all that data with you. You need to have very compelling modelling of their cost basis to know how to price. And people will tell you stories about how it’s super simple.

People have been sold a lot of software in the past, and it takes time before internal teams agree and say, ‘Hey, this is working!’.

Citi Ventures: What considerations are customers weighing when pricing agentic AI?

Misra: Any pricing model requires a deep understanding of the work being done and the role of people that work. Even on the SaaS side, where human labor is involved. Companies know the most about their people and their costs. So when we are talking about agentic pricing, it helps for them to bring to these insights to the table, since the role of people is changing.

This is a fundamental unit: from human execution to bot execution. Humans are now part of a larger conversation that includes deploying digital labor. So I now have customers asking me: “‘What are you charging me for? Are you going to charge me for learning and context? Or are you charging me for execution? Or are charging me for both?’”

Mrkšić: Every company looks at ROI calculations differently. The main thing we look for when discussing pricing is metrics: do companies know how to measure what their people are doing versus what software or, now, agents, will be doing?

If they have this, then pricing can be collaborative because it’s about information sharing and coming up with the best solution. It helps to have credible partners. It helps to get the CFO’s office involved. The good news is, a lot of people want to go agentic so they are willing to think through conversations as they happen.

Citi Ventures: What kinds of ROI are companies seeing today?

Misra:We are seeing two kinds of ROI emerge. There is “hard” ROI, which is about the people on a team. And there is a secondary ROI that we have started to notice across many customers, which relates to the quality of work.

The agents we use are not only able to understand and perform work like the existing workforce, but they are able to do it quicker and with complete understanding of all the rules and compliance issues.

From the get go, many of our customers see that value even in the testing phase, because there it becomes clear how congruent your workforce was to the rules and regulations and the business processes as opposed to a model which is now rock solid and has a full picture of all the work. The model is interrogating each case historically and saying my outcome is different from what this human did for five reasons. And these five reasons are disproportionately in favor of the model because it is applying the rules more properly.

This is quality ROI.

Mrkšić: We work with customers in a number of different verticals. They are looking at ROI from three perspectives. First is around cutting costs. Then there is increasing profits. The final one is where they know agentic is the right way to go but they don’t know yet know how best to determine ROI. It’s hard to price for that. What I would say, is that a good conversation about agentic pricing is rooted in the people who have full ownership of all the data streams in a company. They are best placed to determine ROI, since they know the value of their data and how it can best be used to determine ROI and outcomes. If you don’t have that central knowledge or person, that’s when it gets really difficult.

Citi Ventures: What is your long-term view on agentic pricing and what factors will affect the overall dynamic?

Misra: Everyone is trying to think two steps ahead. We have customers who are explicitly asking us to bring their entire agentic execution into local models on local hardware. No pricing discussion is complete without the notion of control and independence and choice. People don’t want to be beholden to anyone.

At the back of that calculus is the build up toward ‘How will costs be controlled?’ and ‘How will they be paid for?’

Mrkšić: There are a lot of people talking about how SaaS is dead, but it’s more nuanced than that. For example, I can tell you that we have a very stable customer base in hospitality, which is used to paying for us in the same way they pay for software, which is per site, per month. That unit of hotels is not going away, or that unit of restaurant.

What’s our job? To build the technology that allows companies to have leverage over their business and fix their problems disproportionately well with the help of technology. That’s the role of technology. At the end of the day, it helps to have competition to be priced in a way that makes our businesses attractive, otherwise we’ll all just be owned by hyperscalers.