Monetization for AI and Agentic Business Models:

5 Takeaways from Salesforce, Norwest, and RightRev

AI and agentic business models are here, and the rapid pace of innovation is astounding. 

It wasn’t so long ago that transformer technologies came on the scene in 2017. Then it was the launch of ChatGPT at the end of 2022. Businesses started thinking about how to build large language models (LLMs) into their own products, so 2023 and 2024 saw a proliferation of co-pilots and chatbots. 

Fast forward to 2025, and the evolution of generative AI and agentic systems is moving faster than ever. Gartner predicts that by 2028, 33% of enterprise software applications will include agentic AI. 

But because these AI and agentic AI tools increasingly make use of usage-based pricing (UBP) or outcome-based frameworks, they require teams to monitor high volumes of usage data – creating major headaches for the back office. 

To better understand how back office teams can navigate these complexities, we recently sat down with three experts in the space: Jagan Reddy , CEO of RightRev ; Mike Aaron , Sr. Director of Revenue Cloud Solutions at Salesforce; and Scott Beechuk , Investor and Partner at Norwest .

Tl;dr: AI and agentic products necessitate some level of consumption pricing, but that drastically complicates revenue recognition and billing. Not to worry – this conversation digs into the key aspects you’ll need to think through. Read on for the top takeaways from our webinar:

1) Monetizing AI services calls for elements of usage-based pricing.

The traditional subscription is easy and familiar in SaaS. But as Mike notes, the value in AI comes from the interaction, aka the number of times you call it. That means the monetization strategy needs to account for consumption.

Why is this so important with AI and agentic business models? Because when customers upgrade beyond a common “freemium” offering and into the paid version, traditional pricing may not be practical for a high-COGS AI product.

Many companies automatically go with a subscription model because of their familiarity with traditional SaaS. But Jagan says those businesses will start realizing that they're going to lose money on a subscription model. Usage can vary – widely – and a fixed rate subscription won’t necessarily cover the cost of high volume users.

As an example, look at OpenAI founder Sam Altman’s recent announcement that ChatGPT is losing money on their $200/month Pro subscription – because they didn’t expect people to use it as much as they are. 

But while all three panelists agree that AI offerings require some level of consumption pricing to protect from overuse, these pricing models bring their own unique challenges.

2) Revenue recognition gets even more complex with consumption pricing.

Especially from a revenue recognition perspective, Jagan says that every company’s accounting model is either super simple or super complex – there’s no in-between.

And that’s certainly the case with consumption pricing models. Here are just a few challenges mentioned in our conversation:

Credits, rollovers, and prepayments

While a pay-as-you-go model can be simpler (you collect usage data, bill at month-end, and recognize the revenue) more complexity is introduced when the customer prepays for services and draws down from those amounts. 

But what if the customer has unused credits? How do those roll over from the old contract to the new contract? Not to mention all the companies offering free credits in some cases – how should free versus paid credits draw down, and how will you recognize that revenue?

ASC606 

Something engraved in many of our minds is that billing is revenue. But as Jagan points out, the introduction of ASC606 put billing and revenue recognition each on their own path. While the set of accounting standards developed by the U.S. Financial Accounting Standards Board (FASB) aims to standardize how companies recognize revenue, it also completely detached billing from revenue recognition.

Contract modifications

On a consumption business model, Jagan says that in his experience, contract modification can completely collapse your accounting when not done properly. For established companies selling subscriptions who launch new AI products, this can be a huge challenge for the customer to understand:

• How the contract changes

• How unused portions get carried to the new contract

• How to assign SSP for unused credit and new credits

• How to reallocate the revenue

Jagan believes that traditional contract modification simply won’t work for usage-based products – something that many companies unfortunately aren’t thinking about. But as an example of what to look for, Jagan says that the way Salesforce Revenue Cloud enables contract modifications is a “dream-come-true model.”

Why? Because the connectivity between original contract and new activity is crucial – and it's all tightly aligned in Salesforce Revenue Cloud. All of the relevant line items under a contract, both new and old, stay grouped. That way, accounting treatments are easier to reallocate across the changes, prospectively and retrospectively.

“When contract adjustments necessitate backing out original bookings, altering revenue, or canceling billings, the aggregation and reconciliation burden lies with the back office team. The dynamic nature of contracts makes data continuity nearly impossible, increasing the risk of revenue leakage or inaccurate revenue recognition.”

Ebook: Elevating Quote to Revenue Operations with Salesforce and RightRev

Manual revenue recognition 

Trying to manage all of this in a spreadsheet is going to be nearly impossible for accountants, from tracking usage data to applying different discounts to tracking terms for revenue recognition. Even building something in-house is a nightmare to develop and maintain.

This is why RightRev has embedded all of this in our purpose-built engine, so you can handle all the complexities of consumption business models. 

Mike notes how important it is to have strong links between your quoting, billing, and revenue recognition system. He says they’ve seen customers really struggle in this area because of how complicated those handoffs can be.

Salesforce and RightRev joined forces so you can run your entire Quote-to-Revenue process on the Salesforce platform, offering complete visibility into bookings, billings, and revenue. Learn more here. →

3) One of the most important tasks for consumption-based AI pricing will be revenue forecasting.

For businesses moving into a UBP model, a key task is not only to look at what revenue you recognize, but your revenue forecast. 

While this is fairly straightforward to do in a subscription model (even with adding the necessary variance), usage-based forecasting is more complex because consumption is never a straight line. There can be seasonal variance and any number of other impacting factors, so predicting a customer’s usage one, three, or twelve months out is very difficult.

There’s no easy answer to this problem, but one thing is for certain: Usage data quality matters. Jagan says you’ll need to have the right processes in place to capture that data, classify it, attach it to the contract, and rate it.

He also notes how the evolution that’s happening with AI models will help with these revenue forecasting challenges. Now, we can ingest these high volumes of historical data and have AI start analyzing patterns – seasonal and otherwise. We can then use that to start forecasting what usage might look like in the future.

4) Investors are evaluating usage-based AI companies on ROI more than anything else.

Annual Recurring Revenue (ARR) has held the title of North Star metric in SaaS for the last 10–15 years. But now, investors have to identify the best ways to assess companies with usage or outcome-based models.

According to Scott, the investment world is still working through how companies should be evaluated when revenue can be so variable. He points out that it might take a few years before we start to understand how seasonal patterns impact consumption for those usage-based businesses.

This all comes back to the point of forecasting capabilities, which Scott notes are improving every day. Companies are experimenting with a blend of subscription and consumption – often referred to as “hybrid pricing” – as well as credit models where companies can prepay for a certain number of credits for a period. He says some buyers really appreciate these models, especially CFOs who love more predictability around what they’ll be billed for the year.

From an investment standpoint, Scott says evaluation – especially for AI companies – is coming down to ROI more than anything. We’ve all seen stories in the news about AI-native B2B companies and whether they’re getting renewals and generating real ROI for customers.

For him, case studies are one of the most important things. He notes that RightRev and Salesforce are good examples of showcasing case studies on their AI products’ success in the market. Not all startups are able to illustrate that success, he says, so that’s something they’re looking out for. 

5) Trust and security will be key to choosing AI vendors. 

It’s not just investors that should be looking at a vendor’s proof points. In the new world of generative AI, trust becomes one of the most obvious topics to prioritize. As Scott points out, we’re handing over a lot of data to AI models that is both 1) proprietary and 2) sensitive. So the question becomes: How do we know we can trust these models and the vendors who leverage them? 

It comes down to the vendors and their experience – most importantly, he says, how well they understand their domain.

Scott suggests going back to first principles:

• Does the vendor understand how the data flows?

• Does the vendor understand how valuable that data is and how it must be stored and how it must be treated?

• Do they understand the regulations?

• Do they understand the nuances of compliance?

This isn’t the first time the economy has worked through these questions of security and trust. Jagan notes that at the dawn of the cloud, it was a similar story of resistance among finance teams. 

The resistance to AI is tenfold, especially in finance. But the best vendors will have founding teams with deep experience in their domain and the proper guardrails to ensure the highest level of data privacy and security.

For more insights on AI monetization and how it's transforming the back office, plus practical advice on applying these ideas to your business, check out the full webinar with Jagan, Mike, and Scott.