SaaS Metrics School · 2026-06-24 · 4 min
Key moments - from our scoring
Substance score
32 / 100
Five dimensions, 20 points each
Ben Murray examines how AI ARR alone no longer justifies software valuations, drawing on a Guggenheim analyst's critique of Salesforce's $1B AI revenue claim - "easy to announce, hard to prove." The episode charts the evolution of investor expectations: 2024 saw companies launch AI features, 2025 brought AI ARR announcements, but now the market demands financial transparency showing where AI revenue actually appears in margins, customer outcomes, and profitability. For SaaS founders and CFOs preparing for budget season, Murray outlines the operational shift required: granular revenue tracking by AI type (native AI vs. influenced vs. upsell), detailed cost accounting in COGS (inference costs, infrastructure, observability), and instrumentation of subscription-based LLM products to segment user cohorts (heavy, medium, light) and measure their impact on LTV and CAC. Without this infrastructure in place, companies risk difficult board conversations and equity undervaluation.
Investors now demand proof that AI revenue flows through to the bottom line. A Guggenheim analyst noted that while Salesforce claimed $1B in AI ARR, it was hard to see the revenue impact in actual financial statements - investors want to see margins, trended margins, and customer outcomes, not just revenue announcements.
Companies must segment AI revenue into native AI revenue, AI-influenced revenue, and AI upsell revenue, tracked by specific SKUs and product IDs. This granularity helps board members and investors understand which AI initiatives are actually driving margin-accretive growth.
SaaS companies need to track inference costs, infrastructure costs, and observability costs in COGS to understand the true margin impact of AI-powered revenue. This gives visibility into whether AI revenue is actually profitable and dropping to the bottom line.
Companies should build instrumentation to segment users into heavy, medium, and light categories, then track how these usage patterns affect margins, LTV, and CAC for each pricing tier. This reveals whether subscription models are sustainable or losing money on certain user segments.
Board members will ask: How much AI revenue are we generating? How much is AI costing us internally and in product delivery? What margins are we achieving? Are we improving customer outcomes? Companies without this breakdown risk difficult conversations during budgeting.
Our reviewer’s read on each dimension, with quotes from the episode.
The episode surfaces a few genuinely actionable distinctions for SaaS CFOs - parsing native AI revenue from influenced or upsell revenue, and capturing inference/infrastructure costs inside COGS - but most points are stated at a high level without elaboration, and a 4-minute monologue leaves almost no time for depth.
we have to answer how much AI revenue we're generating or is it AI influenced? Is it AI upsell revenue? So not just AI arc...but what type of AI revenue? Pure native AI revenue, influence revenue upsell revenue. Is this a standalone AI sku?
understanding do we have heavy users, medium users, light users, how are those affecting our margins, our LTV to cac
The core observation - that capital markets now demand margin proof behind AI revenue claims - is correct and timely, but it is fundamentally standard finance logic applied to a new label; no contrarian or first-principles argument is made, and the framing mirrors what analysts and CFO Twitter have been saying since early 2025.
It's easy to say that AI annual occurring revenue grew to over 1 billion, but it remains difficult to see it anywhere in the numbers
that missing link today now is that AI ROI that shows AI revenue, AI margins. How's that dropping to the bottom line?
This is a solo monologue by the host; there is no guest at all, eliminating any opportunity to hear a practitioner who has actually built and sold AI products at scale; Murray appears credible as a SaaS finance educator but his own practitioner depth is undemonstrated in this episode.
Welcome. My name is Ben Murray.
I've got some blog posts on this, I'll put that in the show notes
The named Guggenheim analyst and the $1B Salesforce AI ARR figure are the only hard data points; the cost categories mentioned (inference, infrastructure, observability) are real but generic, and no concrete company case study, margin percentage, or dollar benchmark is provided.
the Guggenheim analyst John DeFucci wrote, It's easy to say that AI annual occurring revenue grew to over 1 billion, but it remains difficult to see it anywhere in the numbers
inference cost, infrastructure, cost, observ, observability, um, costs, all that great stuff that goes into delivering revenue
The episode is an uninterrupted solo monologue with no interview dynamic, no follow-up questions, and no pushback on any claim; delivery is noticeably rambling with repeated filler phrases, undermining even the modest structural clarity of the points.
So love the news that's coming through just on all this AI stuff and that comment by the analyst, so true that expectations are ramping up on our AI financial transparency. So I hope you enjoyed today's edition of SAS Metric School.
Computed from the transcript - who did the talking, and the words that came up most.
AI ARR is easy to announce. Proving it is where most SaaS finance teams are about to get exposed. In episode #379, Ben Murray tackles the new bar for AI financial transparency and what it means for your next budget season. The public markets have already moved the goalposts. Launching AI was the 2024 story. Reporting AI ARR was the 2025 story. Now investors and boards want to see AI margins, customer outcomes, and proof that AI revenue is actually dropping to the bottom line. That same pressure is heading straight for private SaaS, and your board will bring it to budget season whether you are ready or not. Understand why AI ARR by itself no longer satisfies boards or investors, and what they now demand to see in the numbers. Separate pure AI revenue, AI-influenced revenue, and AI upsell so your reporting survives scrutiny, using clean SKUs, product IDs, and chart of accounts. Know which AI costs belong in COGS, including inference, infrastructure, and observability, so you can show your real AI margins. Walk into budget season ready for the board questions on AI revenue, AI cost, and margin by revenue stream.
Transcribed and scored by The B2B Podcast Index.
Speaker A: ARR is easy to announce, but getting harder to prove. Let's find out why in today's edition of SAS Venture School. Welcome. My name is Ben Murray. So saw a finance news article come across, uh, my notifications this morning, and there was an interesting comment from one equity analyst. And this was about Salesforce's stock is down something like 40% year to date, so getting hammered. And in this comment, uh, the Guggenheim analyst John DeFucci wrote, It's easy to say that AI annual occurring revenue grew to over 1 billion, but it remains difficult to see it anywhere in the numbers. And I focused on that comment. Amazing comment. Because we've seen an evolution in the public markets. This is coming to private SaaS. So we've got to be prepared, especially for this upcoming budget season. But there are a couple themes here. In 2024, we could say we launched AI Great in 2025. Last year we have AI ARR and that was big in the public market, showing that we are producing AI ARR to help with their, um, valuation. And this year, now you can see from that comment, show me where it shows up in the financial statements. I want to see margins, I want to see trended margins, I want to see customer outcomes. So those expectations are ramping up and the public tech markets always a little bit faster than us or ahead of us as far as reporting and just the, you know, the street wants to see. But this is definitely coming to private SaaS. So a couple things here. First, AI AR is not enough anymore. Now this is really important. We have to answer how much AI revenue we're generating or is it AI influenced? Is it AI upsell revenue? So not just AI arc, and that's always so hard to say, but what type of AI revenue? Pure native AI revenue, influence revenue upsell revenue. Is this a standalone AI sku? So we've got to get very detailed in our sku, our product id, our chart of accounts as far as revenue tracking. So that's really important. That's that first step. Second thing is we've got to be prepared for those board questions that are coming, especially this budget season. So how much AI revenue are we generating? The big thing, how much is AI costing us, both in a product line delivery sense and also just internal AI use? What margins are we generating? So that's what, what this analyst wanted to see, that. All right, all right, we're generating these mar or this revenue, but is it dropping to the bottom line? So that's where now we think about our chart of accounts capturing those AI costs. In our COGS area and I've got some blog posts on this, I'll put that in the show notes, inference cost, infrastructure, cost, observ, observability, um, costs, all that great stuff that goes into delivering revenue. And then is it actually improving customer outcomes? What kind of pricing model are we offering? Is it an outcome base? Is it hybrid platform plus usage? What are the margins on these different pricing plans? So that missing link today now is that AI ROI that shows AI revenue, AI margins. How's that dropping to the bottom line? Are we improving customer outcomes? So a lot of companies, of course we see we're uh, tracking token costs. But now we've got to get a little bit more sophisticated there, put the instrument play, instrumentation. Now especially if we have subscription based AI product lines that are LLM based and understanding do we have heavy users, medium users, light users, how are those affecting our margins, our LTV to cac. So we've got to think about this progression and this is new instrumentation for the back office for CFOs. So this is now we're seeing this evolve real time and we've got to put this in place because if we don't put this in place, we're starting to put that infrastructure in place. And if we're a decent sized SaaS company, that budget season could get a little messy if we can't break out these revenue streams, margins by revenue streams, AI influence and also internal AI productivity. So love the news that's coming through just on all this AI stuff and that comment by the analyst, so true that expectations are ramping up on our AI financial transparency. So I hope you enjoyed today's edition of SAS Metric School.
Other episodes covering the same guests and topics, from across The B2B Podcast Index.
How Fortune 500s Use Procurement to Manage Vendor AI Training Data RightsEnterprise Tech with Fexingo · on Salesforce90 / 100
Why revenue growth breaks down (and how great companies fix it) with Dan Bernoske If Prices Could Talk · on Salesforce89 / 100
Why B2B Brands Fail at Account Based Marketing AttributionThe Marketing Operator Podcast with Fexingo · on Salesforce88 / 100
#410 - How Mazy Dar found room in Google and Microsoft's market - and won the world's biggest banksThe Remarkable SaaS Podcast · on Salesforce87 / 100
How to Sell Against a Competitor Already in the BuildingSales Leadership with Fexingo · on Salesforce85 / 100
Why B2B Brands Are Using AI for Account PrioritizationThe Growth Operator with Fexingo · on Salesforce84 / 100
How Much Are Tech CFOs Actually Making in 2026?50 / 100The Latest GRR Benchmarks66 / 100Here's What Separates the 9 Public SaaS Companies that Trade Above 10x63 / 10012 Steps to Creating an Outcome-based Pricing Plan63 / 1005 Takeaways for CFOs from the 2026 AI Pricing Report