AI is transforming the front office—now it’s time for the back office to catch up.
In this webinar, experts from RightRev, Salesforce Revenue Cloud, and Norwest Venture Partners explore how leading companies are rethinking billing, revenue recognition, and financial reporting to support usage- and outcome-based models.
Learn how to:
- Navigate the shift from SaaS to usage- and outcome-based pricing
- Handle revenue forecasting for dynamic consumption models
- Build trust and security into AI-powered back-office systems
- Future-proof your quote-to-cash process for rapid innovation
Transcript
Let’s dive in.
AI and agentic models are here. I don’t know about you, but my inbox is flooded daily with subject lines about AI—everyone promising to help me work more efficiently. The pace of innovation is staggering, and that means back-office teams need to adapt just as quickly.
A Bain & Company report recently said that over “70% of leading GenAI companies have adopted usage-based pricing models—especially those offering APIs or infrastructure tools.” It’s not just AI-native companies like OpenAI, Anthropic, or Perplexity—enterprise SaaS providers are layering AI into existing models.
Gartner predicts that by 2028, “33% of enterprise software applications will include agentic AI.” These agentic models often rely on usage or outcome-based frameworks, which bring a host of challenges—managing large volumes of user data, decoupling billing from revenue, and more.
To unpack all of this, we’ve brought together three experts:
- Jagan Reddy, CEO and Founder of RightRev (known to some as the “Godfather of Revenue Recognition”)
- Mike Aaron, Product Leader at Salesforce Revenue Cloud, focused on billing
- Scott Beechuk, Partner at Norwest Venture Partners
Moderator (Alissa):
Scott, let’s start with you. You have a front-row seat to innovation. How are you seeing AI move beyond chat and assisted writing into the back office? And how are these products being priced and packaged?
Scott Beechuk (Norwest Venture Partners):
Thanks, Alyssa. It’s great to be here—this is such a timely topic.
AI has been on a rapid trajectory. Transformer models emerged around 2017, and by the end of 2022, tools like ChatGPT brought large language models into the mainstream. Very quickly, APIs became core to how developers began embedding AI into business workflows.
In 2023 and 2024, we saw a lot of copilots and chatbots integrated into business apps. That made sense given the conversational nature of early tools like ChatGPT.
Now in 2025, we’re entering a new phase: generative, agentic systems. These aren’t just assistants—they’re agents that can work in parallel or sequence, solve complex problems, and mimic human reasoning.
Many back-office processes rely on coordination between people and tasks. Agentic AI can automate that coordination. It’s not just about automating one step—it’s about completing entire workflows.
We’re moving toward a world where AI isn’t just augmenting work—it’s embedded into the fabric of how work gets done. And the way these tools are monetized is evolving just as fast.
Moderator (Alissa):
Mike, Salesforce recently launched AgentForce. Can you share the vision behind it and how it’s helping usher in the agentic era?
Mike Aaron (Salesforce Revenue Cloud):
Absolutely. To Scott’s point, AgentForce represents what we see as the third generation of AI—going beyond traditional copilots.
The goal is to deliver intelligent, low-hallucination agents that proactively drive customer success. These agents help guide users, execute tasks without human intervention, and serve as digital labor to augment human capabilities.
We’re not just talking about lead gen or opportunity management. These agents are working their way into quoting, billing, and all the way through revenue operations. And we’re seeing firsthand how these models help companies grow more efficiently.
Moderator (Alissa):
Right—so you’re not only building these tools at Salesforce, but also working directly with AI companies using Revenue Cloud for their quote-to-cash processes. What are you seeing in terms of monetization models?
Mike Aaron:
Exactly. It’s fascinating. We’re working with AI-native companies who are selling software that includes AI, to other AI companies. That creates a very meta scenario.
They’re not just charging $9.99/month for a subscription. AI value is in the interaction—it’s in the API call volume, the number of messages or actions. So the monetization strategy has to reflect that.
That means consumption models. You need quoting systems that can handle variable pricing. You need billing systems that can track usage, rate it correctly, invoice for it, and then recognize revenue accordingly.
AI companies are asking: how can our own AI products help run these systems more efficiently? That’s where we see this ecosystem evolving rapidly.
On AI Security, Trust, and Controls
Moderator (Alissa):
Jagan, as Scott and Mike described, sales teams are getting creative with how they go to market—layering usage on top of SaaS, experimenting with new pricing. What does that mean for back-office teams? From an accountant’s perspective, how is this impacting their day-to-day?
Jagan Reddy (RightRev):
We’re seeing a lot of variability—different pricing and accounting models depending on the AI company’s offering. There’s no one-size-fits-all.
Most companies start with a freemium model. It lets users try out the service before deciding if it’s valuable. From there, many move into subscriptions. But they quickly realize that subscriptions alone might not capture the full value—especially when usage varies widely. That’s when usage-based pricing becomes more attractive.
Within usage-based, we typically see two flavors:
Pay-as-you-go, like your phone bill—you pay at the end of the period for what you used.
Prepaid credits, where customers purchase a set amount upfront and draw down against it.
The complexity for accounting teams really begins when these models start to layer together—freemium, SaaS, usage, prepaid—and each has different rules for billing and revenue recognition.
Mike Aaron:
Yeah, we’re seeing a hybrid trend—80% of revenue as recurring SaaS, with 20% usage-based on the back end. That gives companies predictable revenue while still capturing value from high-consumption users.
Some AI companies are going pure usage, but most are blending.
Jagan Reddy:
Even within consumption-based models, it’s evolving. One of our customers is heavily into usage billing and now offering AI services. What they’re doing with revenue accounting is some of the most complex modeling I’ve ever seen—and I’ve seen a lot.
Moderator (Alissa):
Let’s pivot for a second. All this usage data and AI functionality introduces a huge trust and security component. Scott, if I’m a customer evaluating AI vendors, how do I know who to trust?
Scott Beechuk:
Great question. In the world of generative AI, trust is critical. You’re handing over sensitive, proprietary data to these models—so how do you ensure it’s protected?
It starts with the vendor. Do they understand the domain they’re operating in? Do they have a track record of handling sensitive data? Do they know how to apply the right security protocols and comply with regulations?
There are a lot of startups popping up in AI. Some don’t have domain experience, and that’s risky. The best vendors—like Salesforce and RightRev—build enterprise-grade solutions with security and compliance baked in.
We’ve chosen to partner with RightRev in part because they’ve proven they can handle sensitive financial data in complex environments.
Moderator (Alissa):
Right, and in usage-based models, the data you’re collecting is deeply tied to product usage. Jagan, what’s your take on safeguarding sensitive data, especially around accounting?
Jagan Reddy:
I’ve seen this story before. At my previous company, when we first moved RevRec to the cloud, finance teams were extremely resistant. Sensitive accounting data in the cloud? No way.
Over time, we earned their trust with certifications and controls. Now we’re seeing the same resistance with AI—only 10x stronger. Customers are asking in contracts: “Do you use AI? Do you share our data? We don’t want our data used in training models.”
At RightRev, we’re building with those concerns in mind. We’re working on customer-specific AI models—no cross-customer training. We want our customers to feel 100% confident that their data is protected and used only for their benefit.
Moderator (Alissa):
Let’s get back into usage-based billing and revenue recognition. Mike, we’ve talked about the many “flavors” of usage—pay-per-conversation, pay-per-action, etc. How is Salesforce Revenue Cloud helping customers navigate this complexity?
Mike Aaron:
It starts with flexible rating. When a usage record comes in, we need to know how to rate it—whether it’s postpaid, consuming from a balance, or drawing from virtual currency.
Then there’s pricing tiers, discount structures, and bundling. It gets complex quickly.
At Salesforce, we built a platform-level rating engine to handle that. Whether a company is shifting from SaaS to hybrid or going full usage-based, we help them quote, rate, bill, and send accurate data downstream.
That’s where our partnership with RightRev is key. We pass along accurate billing and usage info so they can recognize revenue properly. That handoff is critical.
Moderator (Alissa):
Jagan, that handoff—estimating what will come in—adds complexity for forecasting. What does revenue forecasting look like in usage-based models?
Jagan Reddy:
Forecasting becomes the biggest challenge.
In SaaS, forecasting is easy: take the contract value and spread it over the term. You can adjust for churn or uplifts and still project 10 years out.
In usage-based models, you don’t know how much the customer will use next month. You have to rely on historical usage patterns—but usage is rarely linear. It fluctuates. It can be seasonal.
One of our customers has an entire team of data engineers focused just on forecasting revenue from usage data. It’s expensive and difficult.
That’s why we’re investing heavily in AI models for usage-based forecasting. With the right historical data and seasonal trends, we can generate more accurate forecasts for both billing and revenue.
Moderator (Alissa):
That’s a game changer. Alleviates a ton of pain.
Scott, let’s talk about financial reporting. For 10+ years, ARR has been the North Star for SaaS. But with usage and outcome-based models, how do you, as an investor, think about ARR now?
Scott Beechuk:
ARR is still a helpful metric, but it doesn’t tell the whole story anymore.
We’re seeing companies experiment with blended models—some fixed SaaS fees, some consumption. That helps buyers (especially CFOs) feel confident in predictable costs while allowing upside with usage.
We’re also seeing high-water mark pricing—buy a set of credits for the year, top up when needed. That gives structure without wild month-to-month swings.
But from an investor’s lens, we’re increasingly focused on ROI. Is the product delivering measurable value? Are customers renewing? Are they expanding? That’s the real test.
On How Back-Office Systems Become Go-To-Market Enablers
Moderator (Alissa):
You mentioned experimentation, Scott—and I think that’s key. AI and usage-based companies are testing different pricing models, iterating quickly.
Sam Altman recently admitted that OpenAI is losing money on the $20/month ChatGPT Pro plan because usage is higher than expected. That puts pressure on infrastructure and back-office systems to support growth and agility.
It used to be that back office was just about invoicing and reporting. Now it’s a strategic enabler for go-to-market.
Jagan, can you share an example of how companies are using back-office systems as a growth enabler?
Jagan Reddy:
Sure. When we started RightRev, we built a general-purpose revenue engine to handle everything. But as we worked with customers who had complex usage models, especially in AI, we realized we needed something more specialized.
So, we built a purpose-built engine for consumption-based models. The complexity is enormous.
Take prepaid credits, for example. A customer might buy $1 million worth of credits upfront. They may consume those credits across multiple services—each with different pricing, discounts, and rules.
There are rollovers—unused credits from one contract carried into another. Then the question becomes: do those credits follow the old terms or the new ones?
What about bonus credits—say the customer buys $1M in credits, and we give them $100K extra. When usage comes in, do we consume the bonus credits first or the paid ones? It’s a debate.
The most common approach is to recognize revenue only on the paid portion. So if a $1 credit was used, and 10% was a free bonus, we only recognize 90 cents.
All of this is hard—impossible, really—to manage in spreadsheets. That’s why we embedded these capabilities into our engine with flexibility for customers to define rules that match their business.
Moderator (Alissa):
Love hearing you dive deep, Jagan. And you’re right—there’s no single way to do this. There are so many variations.
Mike, what are some “gotchas” you’ve seen? Where do companies get tripped up when implementing usage-based or outcome-based models?
Mike Aaron:
The biggest one? System fragmentation.
Quoting, billing, and revenue recognition are often managed in separate systems—and it’s those handoffs that create friction. If they’re not tightly integrated, it’s hard to support new pricing models or make changes quickly.
That’s why we deliver all of this on a single platform. It removes the friction and makes it easier to launch new models, like prepaid drawdowns, rollovers, or virtual currencies.
Moderator (Alissa):
Jagan, same question—what are some issues teams run into? What advice do you have?
Jagan Reddy:
One of the biggest pain points is contract modification.
Let’s say a customer buys prepaid credits in a contract. Then partway through, they want to renew, top up, or switch terms. Now you’re dealing with two contracts, multiple currencies, and changing exchange rates.
Rolling unused credits into the new contract requires precise linking. You have to track the value of those credits, apply new or old terms, and reallocate SSP (standalone selling price). If you get that wrong, revenue recognition can collapse.
Many companies don’t realize this and try to manage it manually—or think they can fix it in the RevRec system alone. But RevRec relies on clean upstream data. If you input garbage, you’ll get garbage out.
This is where Salesforce Revenue Cloud shines. The way they handle contract modifications—linking line items across old and new contracts—is a dream from a revenue accounting perspective. That linkage makes it much easier for RightRev to apply usage and recognize revenue correctly.
Moderator (Alissa):
Some companies still want to build in-house. They think their use case is unique, so no platform could possibly support them.
Mike, I’m sure you’ve seen both sides. What would you say to a company thinking of building billing and RevRec systems in-house?
Mike Aaron:
If you’re building in-house, you’re probably building to a single use case—and as soon as that changes, you’re back in the development cycle.
You’ll be stuck maintaining a homegrown system forever. Is that really your business model?
Salesforce and RightRev have seen thousands of customers and built in support for a wide range of use cases. Why spend years building what already exists—and miss your go-to-market window?
Moderator (Alissa):
Jagan, what about companies relying on legacy platforms like ERPs?
Jagan Reddy:
From a revenue recognition standpoint, you’re either very simple or very complex—there’s no in-between.
Legacy ERPs handle traditional models fine, but they can’t support the complexity we’re seeing today: rollovers, multi-element arrangements, prepaid credits across multiple services, etc.
We’ve seen companies invest millions to build something in-house, only to realize we already support it—and they could’ve launched months faster.
It’s all about time-to-market. Let us handle the complexity, so your team can focus on launching and scaling your product.
Moderator (Alissa):
Speed is key. Many companies are launching new AI features in 90-day windows. Mike, what does quote-to-cash readiness look like in a short timeframe?
Mike Aaron:
Instrument everything. Don’t just track one metric—track more usage data than you think you’ll need.
That way, if you want to change pricing, shift to a different model, or experiment with new billing structures, you’re ready.
If the data is there, you can rerate it, reuse it, and support new models faster.
Moderator (Alissa):
Jagan, anything you’d add?
Jagan Reddy:
Yes—capture your usage data cleanly and thoroughly. It’s the foundation.
We’ve seen companies struggle to invoice or recognize revenue correctly because their usage data isn’t properly classified or tied to contracts.
If your usage data is clean, your billing and revenue will fall into place. Otherwise, you’re setting yourself up for pain.
Moderator (Alissa):
This has been super insightful—thank you all.
Any final thoughts before we wrap?
Scott Beechuk:
Just a big thank you to RightRev and Salesforce. This is going to be an exciting year.
We’ve unlocked the ability to model complex human work with AI. That’s going to accelerate business processes across the office of the CFO—from close to forecasting to strategic planning.
And I think we’ve only seen the snowflakes on the surface—there’s a lot more coming.
Moderator (Alissa):
Thank you, panelists. And thank you to everyone who joined us.
Check out the handouts and resources, including a short demo on how RightRev supports usage-based billing. There’s also a “Better Together” story featuring RightRev and Salesforce Revenue Cloud.
We’ll send out the recording shortly.
Have a great day!
