Jul 18, 2025 · 41m · a16z
Tech Executives: AI Has Changed SaaS Forever (Don't Fall Behind)
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of a16z, host Martin Casado and Metronome CEO Scott Wu discuss how artificial intelligence is transforming SaaS business models from traditional seat-based subscriptions to agile, usage-based monetization systems. They provide actionable frameworks for tech executives to navigate the technical, organizational, and strategic shifts required to maintain competitive pricing in the AI era.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. How this is scored →
The host as informed peer 5.2 Guest teaching 5.5 Guest disagreement 1.1 The host pushing back 1.3
speaking balance: gold is the host, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 28:05 Rejecting Pricing Councils for Pricing Dictators
Scott explicitly rejects the standard corporate practice of relying on cross-functional pricing councils, insisting executives appoint a single pricing dictator to push changes past internal drag.
Hardest push from the host ▶ 16:36 Interjecting on Sales Comp PitfallsThe host interrupts Scott's point on sales incentives to firmly correct the premise that companies design reasonable comp plans, emphasizing how often leadership misaligns sales commission structures.
Biggest teaching moment ▶ 11:38 Real-Time Unbounded Spend MechanicsScott educates the host on the subtle infrastructure demands of usage billing, showing why real-time monitoring is needed to stop unbounded runaway processes like Segment's accidental $80k customer bill.
The host holds their own ▶ 11:02 Distributed Systems Real-World ExperienceThe host demonstrates deep domain expertise by citing personal experience leading infrastructure projects, noting that even world-class distributed systems engineers struggled to build usage billing in-house.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The host as informed peer | Guest teaching | Guest disagreement | The host pushing back | Why |
|---|---|---|---|---|---|---|
| Title Card: AI Is Upending SaaS Pricing | 0 | 5 | 1 | 0 | Scott Wu delivers an opening overview of billing technical debt and Metronome's origins at Dropbox. The host does not actively participate in this opening context, keeping host-side metrics at zero while Scott outlines experimental lag and infrequent billing cycles. | |
| Why Legacy Billing Fails and Usage Data Infrastructure | 2 | 5 | 1 | 0 | The host asks a straightforward question regarding the inflection point for founding Metronome. Scott explains why billing is notoriously where engineering careers go to die and how usage-based models transformed billing into a data infrastructure challenge. | |
| The Three Historical Eras of Software Monetization | 7 | 5 | 1 | 1 | The host establishes high context by referencing personal board seats and past experience selling perpetual software licenses. Scott expands on this framing by categorizing software monetization into on-prem, cloud seat-based, and AI value eras. | |
| Application vs Infrastructure Pricing and Hybrid SaaS Models | 8 | 6 | 1 | 1 | The host highlights personal experience trying and failing to build usage billing with a top-tier distributed systems team. Scott educates the host on the hidden mechanics of real-time unbounded spend, dynamic contract edge cases, and golden financial data pipelines. | |
| Aligning Company Strategy and Incentives Around Usage | 7 | 5 | 1 | 4 | The host interjects to challenge common assumptions around sales comp plans, pointing out that companies frequently get commission structures wrong. Scott agrees and elaborates on how usage models align company-wide incentives across sales, product, and finance. | |
| Restructuring Sales Compensation and Customer Success Roles | 7 | 5 | 1 | 1 | The host prompts Scott for a CEO action list and contributes an industry observation about Customer Success KPIs shifting toward gross churn over expansion. Scott outlines changes in sales territory, technical pre-sales, and solution architect roles. | |
| Evolving Product Managers and Finance into Strategic Data Teams | 2 | 6 | 1 | 0 | Scott walks through how product managers and finance teams must evolve into real-time data operators under usage models. The host largely listens as Scott details how product feature metrics directly determine corporate revenue stream velocity. | |
| Mapping Software Segments: Pure Usage, Hybrid, and Enterprise | 5 | 6 | 1 | 2 | The host asks whether all software categories will adopt usage billing or if seat models will persist. Scott breaks down market segments, arguing infrastructure goes pure usage, human-centric SaaS goes hybrid, and B2C remains subscription to avoid consumer cognitive load. | |
| Pricing Governance and the Need for a Pricing Dictator | 3 | 6 | 2 | 1 | Scott presents a mild contrarian stance against corporate consensus, arguing CEOs must appoint a pricing dictator rather than relying on slow pricing councils. The host prompts the organizational structure discussion, letting Scott highlight Salesforce's rapid pricing changes. | |
| Navigating the AI Cost Shock with Pricing Agility | 7 | 6 | 1 | 3 | The host challenges the traditional software rule that prices cannot be raised, pointing to modern product-led companies using pricing as a distribution weapon. Scott agrees that AI value changes permit dynamic pricing, cost-plus margins, and capital-backed market dominance. | |
| Early Internet Parallels and De-Risking AI Consumption Models | 8 | 5 | 1 | 2 | The host draws explicit historical parallels to the dot-com era brand moats and distribution dynamics, inviting Scott to keep them honest on monetization differences. Scott validates the parallel and provides examples like Intercom offering $1M guarantees on AI agent resolutions. | |
| Actionable Guidance for Engineers and Executives in AI | 6 | 6 | 1 | 1 | The host synthesizes the episode into a core takeaway regarding the AI supercycle and strategic billing. Scott closes with a concrete engineering example where Snowflake engineers must throttle query optimizations because instant efficiency drops company revenue. |
Statements from this episode (29)
Wu: Taking 18 months to adjust SaaS pricing will leave companies outmoded
“And so if you're the one sitting there, startups on the one hand are kind of like Cannibalizing or rewriting your market, and you're sitting on a pricing change that you're gonna sit, and you're gonna do a nine month case study, and then roll it out after six …”
Scott Wu: Billing is a product surface, not a monthly workflow
“Billing was really honestly more of a product surface than a, like a once a month workflow.”
Wu: Internal billing systems are where engineering careers go to die
“I think the kind of, I think billing is one of those, if you like talk to people who've worked at companies, billing is always like where engineer careers go to die.”
Wu: Dropbox's billing system was continuously handed off to summer interns
“Like at Dropbox, for instance, our billing system was built by interns every summer. Like it would get updated every summer because the only people we could convince to work on it were like interns.”
Wu: Usage-based pricing converts software billing into a data infrastructure problem
“In usage, it's like now you're like, okay, I, you know, we have a trillion queries on ChatGPT, and what we're doing is we're pricing per token, but if it's like above this token amount, it's priced totally differently. And so actually now the complexity of the…”
Scott Wu: AI output-based value breaks traditional seat-based SaaS monetization
“AI is changing the value prop of software. You know, it's kind of doing two things at once. One, it's like the software is not doing work on your behalf. And so the core value of the software is not who in my org has access to the data. It's like, what can thi…”
Scott Wu: Pure usage-based pricing is difficult due to unpredictability
“Moving to pure consumption is actually kind of a challenging thing to do because seats and subscriptions is really simple and easy to understand. Usage is not nearly as simple and it's really, it's much harder to predict for good reasons.”
Scott Wu: Salesforce combines seat fees with usage pricing for Agentforce
“Salesforce, for instance, you know, they went and like added this agent for stuff. They still have per user access fees, but now the agent forces is variable cost on top that allows them to like, if the agents actually are providing value, now my business will…”
Wu: Segment lost $80K in uncollectible fees from customer misconfiguration
“We talked to segment five years ago. And they were one of the first pioneers of this stuff. And they had some small startup in India with like, 3000 dollars in their business, in their account. They turned on Segment, and then they sent them a bill at the end …”
Wu: Public company manages $1B in usage billing entirely by hand
“We're talking to a company last week, a public company, they're in like a billion dollars in usage-based spend through enterprise contracts. All of that billion dollars is handled by hand.”
Wu: 99% accuracy in financial billing data is equivalent to fraud
“The 99% accuracy is fraud in a financial sense, and so you actually have to have really golden data pipelines in order to solve this problem, especially at public company scale.”
Wu: Most usage-based SaaS companies misalign sales commissions and revenue
“Most people just, like, they get the worst of both worlds. They basically pay you out for the million dollar contract. And then they only get to recognize revenue when the usage is accrued.”
Wu: Dropbox's top 1% usage customers were margin negative
“The commercial model of Dropbox incentivized, it was a team-based thing, number of users. It incentivized us to get really good at getting a wall-to-wall deployment. Did not incentivize us at all to drive incremental usage. In fact, in a very literal sense, yo…”
Wu: Metronome requires CEO involvement for sales deals to proceed
“When we're talking to a sales deal, if I can't talk to the CEO of that company, it's not a real deal because the CEO has to be like, I'm going to take the hard path.”
Wu: Traditional Customer Success roles fail to justify cost in usage SaaS
“I think in a lot of cases, the role of CS is really much thinner than you would otherwise have. I think you really do want a team who's deeply embedded in the success of the company, but most CS folks on the label Aren't doing enough to justify that.”
Scott Wu: Product teams directly drive revenue in usage-based SaaS
“And so actually in a lot of ways, your product team becomes almost like a revenue driving team in a way that in seat subscription, you kind of had enough buffer that you could kind of like be like, eh, okay, product's probably providing value in these usage ba…”
Scott Wu: Monthly finance reporting is too slow for usage-based SaaS
“Like a finance team that's used to working quarterly monthly is like orders of magnitude too slow for this world.”
Scott Wu: Human-Facing SaaS Models Will Be Hybrid Near-Term
“I would say the layer where humans are actually interacting. I think the dominant model, at least for the next couple of years, is going to be hybrid.”
Scott Wu: Enterprise SaaS Will Shift Toward Pure Usage Pricing
“And then enterprise, I think will just become closer to pure usage because basically a scaling fee that grows with your headcount is kind of like incentive disaligned with like most CFOs.”
Scott Wu: Companies need a pricing dictator, not a pricing council
“My main thing is like, you don't want a pricing council. You want a pricing dictator. You want someone who is telling you what the right pricing is.”
Wu: Salesforce changed its fundamental pricing structure three times in 12 months
“Agent force, like literally Salesforce, one of the biggest companies in the world changes fundamental pricing structure three times in the past 12 months.”
Wu: Software companies ignoring AI infrastructure costs are falling behind
“I definitely think you're behind because you just inserted an insanely expensive line item into your bottom line in the form of AI. And if your CFO isn't screaming at you now, then probably either, either the AI is useless for you, which, okay, interesting. Or…”
Wu: Salesforce tests roughly a thousand pricing variants behind the scenes
“You even see exploration within, take a large company, take a Salesforce. Behind the scenes, they have their like list price. They have like a thousand variants in production, trying different things out, looking for the one that's going to stick and drive the…”
Wu: Seeking one fixed AI pricing model is a category error
“Basically, sometimes people are like, what's the right pricing model? And I'm like, that's a category error. Like there is no right pricing model. The only thing I know is agility is the only thing that's rewarded in this environment.”
Wu: AI startups are capping gross margins at 20% via cost-plus pricing
“Actually, one of the things that's really fascinating is you'll see companies that are actually, their pricing model is like a cost plus model in this moment, where basically they just fix their margin. And then they basically say, look, we just, we're going t…”
Wu: Using venture capital to undercut AI rivals is a smart bet
“And so fundamentally they sh they're using capital as a weapon to kill competitors. To gain market share and they'll figure it out down the line. And I think that's like, honestly, it's a pretty good bet in this moment.”
Scott Wu: Intercom offers a $1M refund guarantee for Fin AI agent
“Actually, I was literally, as I was driving down here, I saw the intercom. They have this fin product, which is an AI agent in it. It does auto resolution of things like the auto resolution of customer support. They're like, cool. We have benchmarked this like…”
Scott Wu: Autonomous AI agents will prefer usage-based pricing
“And in the agent world, Agents actually don't care about complexity. They just care about optimal price. And so that they're gonna favor that, that, that kind of model.”
Wu: Usage-based software companies must meter out performance optimizations to protect revenue
“Add a company that's a user-based company, take like a snowflake. That engineer is not allowed to ship that optimization. Why? Because basically if you decrease the cost of computation by 50% overnight, suddenly your revenue on your like one of two, you know, …”