Marcos Rivera

What fails is charging for seats without examining what the product does.

• If the product works fine with no human involved, consumption or usage pricing fits.
• If a person uses, accesses, or collaborates with it, user-based pricing or a hybrid of seats plus usage fits.
• Rivera warns against swinging the pendulum from everyone charges for seats to everyone abandons seats.

At an executive dinner during the Pavilion GTM 2026 Conference, hosted by Sherri Sklar, about 20 venture capital and private equity operating partners, CEOs, and founders discussed the topic. Rivera's takeaway was that adoption is not running as fast as headlines claim, especially in regulated sectors like financial services and health tech, and some companies have pulled back on AI after token spend exceeded expectations.

Read original · Original 2

The answer is usually silence.

• Rivera cites a Revenera survey of 415 software leaders: 96% had shipped AI, almost half said customers barely use it, and only 7% reported losing seats to AI.
• A Benchmarkit survey of 296 buyers found 89% went over their AI budget, mostly because AI features drove more usage than planned.
• Rivera's conclusion is that vendors cannot see usage and buyers cannot forecast it, so no pricing metric fixes the blind spot.
• Measurement comes first; otherwise a team cannot tell a feature nobody values from one it packaged badly.

Rivera says the same question applies in private equity diligence: is AI adding revenue, protecting it, or just raising cost of goods sold?

Read original

Rivera uses Atlassian's expansion of usage-based pricing around AI to argue that leading with visibility and spend controls is the right instinct, but visibility is not predictability. Rivera says any team shipping a usage-based model should answer three questions: can a buyer estimate their bill before signing; what happens when usage spikes, whether a hard cutoff, soft cap with alerts, or silence until invoice; and does the metric track customer value or just infrastructure cost?

Rivera describes a model that cannot answer those questions as a variable invoice with a fancy interface.

Read original · Original 2

Rivera quotes Emily Sanz, who argues a well-run credit model is a flavor of outcome-based pricing that tracks multiple outcomes instead of one.

• Rivera also quotes Steve Inman, who compares it to a hot shower: the outcome is the completed shower, while usage is the kilowatts to get there.
• Sanz extends the analogy to a faucet, asking what it means to charge every time someone turns it on when nobody knows how many times a day they do.
• Rivera cites buyer research in which 55% said credit and token pricing is harder to evaluate for AI than for traditional SaaS.
• The fix Rivera endorses is a credit model with clear definitions and fewer definitions that look more like outcomes, since fewer actions make consumption more predictable.

The hot-shower and faucet comparisons are used to show that metering raw consumption does not match what buyers think they are buying.

Read original

First, adoption volatility: AI delivers swings between high value and no value, far less predictable than SaaS.

• Second, real cost structure: LLM token costs are significant and constantly shifting, unlike the near-zero marginal cost of adding a SaaS user.
• Third, blurred attribution: the line between human and system contribution is unclear, complicating value capture.
• Rivera says these break at the same time, but repeats that seats can still be part of the answer for many companies; they just cannot carry the whole model once AI does real work inside the product.

Rivera summarizes a dinner co-hosted with Camino Search and EisnerAmper on moving from seats to outcomes.

Read original

At companies under $50M in revenue, 77% rank predictable total cost as a top-three priority when evaluating an AI vendor; that drops to 63% above $1B in revenue.

• Meanwhile, ease of justifying spend to finance and the CFO climbs from 44% to 55% across the same range.
• Rivera reads this as smaller teams trying to guess the bill while larger teams already know they will spend and need help defending that spend internally.
• Rivera connects this to more than 480 pricing engagements and says the budget conversation never disappears; it moves from what will this cost us to can we justify what we are already spending.

Rivera warns that a pricing motion answering only one of those questions loses deals at the other end of the market.

Read original

Rivera co-produces an AI Pricing Report with Benchmarkit and cites its findings across posts on predictable cost, credit models, buyer-size differences, and usage measurement.

• Rivera also references a Revenera survey of 415 software leaders and a base of more than 480 pricing engagements.
• Several posts end with an invitation to download the full report.
• This makes research a recurring priority: the argument is expected to come with buyer data, not just a framework.

The AI Pricing Report figures appear in posts about predictable total cost, credit and token pricing, and how buyer priorities change with company size.

Read original · Original 2 · Original 3 · Original 4

Rivera uses AI tools but draws a distinction: a prompt can produce an answer in seconds, while conviction only shows up after wrestling with the problem.

• Paper slows Rivera down enough to work out what Rivera actually believes, and the crossed-out parts are where Rivera is wrong on paper so as not to be wrong in front of a client.
• Rivera closes by calling pricing the ultimate expression of business strategy and says you cannot copycat or prompt your way into a strategy.

The post shows the handwritten pricing notes as the starting point for pricing strategies Rivera has shipped.

Read original

Start here if you only open one post.

Rivera opens with survey numbers: 68% of AI buyers rank predictable total cost among their top three pricing priorities, and only 19% care about a lower entry price.

• He then turns to Atlassian's expansion of usage-based pricing around its AI features, noting that Atlassian led with visibility and spend controls rather than just a new meter, and calls that the right instinct.
• The rest of the post separates two ideas that vendors treat as one: making usage visible is not the same as making it predictable.
• He lists three questions a vendor should answer before shipping a usage model.
• Can a buyer estimate their bill before they sign, not after?
• What happens when usage spikes: a hard cutoff, a soft cap with alerts, or silence until the invoice?
• Is the metric tracking customer value, or is it just tracking infrastructure cost?
• His stated failure mode is a model that answers none of them, which he describes as a variable invoice with a fancy interface.

For brands planning a campaign

Use the three questions as a launch checklist. The spike question is the one that becomes a customer-facing control, such as spend alerts, a cap, or a live bill estimate, and that control gives a pricing message something concrete to point at instead of a general claim about transparency.

It also shows how Rivera handles the pressure executives feel from a board that wants an AI revenue line.

Rivera takes the executive line, that a company needs to start charging for its AI features, and answers it with a question: who is using them?

• The answer, in his experience, is usually silence.
• He cites a Revenera survey of 415 software leaders reporting that 96% have shipped AI while almost half say customers barely use it, and that only 7% are losing seats to AI even as nearly half feel pricing pressure from it.
• A second study with Benchmarkit, covering 296 buyers, found 89% went over their AI budget, mostly because AI features drove more usage than they planned for.
• His conclusion is that vendors cannot see usage and buyers cannot forecast it, and that no pricing metric fixes that gap; measurement comes first.
• Without it, he writes, you cannot tell a feature nobody values from one you packaged badly.
• For investors in diligence he offers a blunt question about whether AI is adding revenue, protecting it, or just raising cost of goods sold.

For brands planning a campaign

Pair this with any proposed metric change. If a pricing team cannot show a usage baseline, the post supplies the argument for a measurement phase before the new model and the language for explaining that phase to a board.

Rivera calls many AI credit models electric bills with a nicer name, because they turn into pure pay-per-use, pay-per-click billing.

• He quotes two members of his delivery team.
• Emily Sanz argues that a well-run credit model is a flavor of outcome-based pricing tracking multiple outcomes instead of one.
• Steve Inman compares it to a hot shower: the outcome is the completed shower, the usage is the kilowatts it takes to get there.
• Sanz extends the analogy to a faucet, asking what it means to charge every time someone turns it on when nobody knows how many times a day they do.
• Rivera backs the argument with data from an AI pricing report built with Benchmarkit: 55% of buyers said credit and token pricing is harder to evaluate for AI than for traditional SaaS, and 68% rank predictable total cost in their top three priorities.
• Sanz's fix, which Rivera endorses, is a credit model with very clear definitions and fewer of them, each sitting closer to an outcome.

For brands planning a campaign

The actionable part is the rate card, not the model name. Cutting the number of billable actions makes consumption easier for a buyer to estimate, and a smaller set of units is easier to explain in sales material.

Rivera summarizes a dinner he co-hosted with Camino Search and EisnerAmper on moving from seats to outcomes, and lists three reasons old models break in the AI era.

• The first is adoption volatility, where AI delivers swings between high value and no value that are far less predictable than traditional SaaS.
• The second is real cost structure, where LLM token costs are significant and constantly shifting, unlike the near-zero marginal cost of adding a SaaS user.
• The third is blurred attribution, where the line between human contribution and system contribution is unclear, which complicates capturing value.
• He is explicit that seats can still be part of the answer for many companies and are not dead, but that they cannot carry the whole model once AI is doing real work inside the product.

For brands planning a campaign

Useful as the framing document for an internal discussion. Each of the three reasons maps to a different part of a monetization plan: forecasting, gross margin, and how the value story gets attributed between the product and the user.

Rivera writes from an executive dinner he attended during the Pavilion GTM 2026 Conference, hosted by Sherri Sklar, with about 20 venture capital and private equity operating partners, CEOs, and founders.

• His takeaway is that adoption is not running as fast as the headlines claim.
• Financial services, health tech, and other heavily regulated sectors are moving slower than press coverage suggests, and he notes that these sectors make up a large share of large enterprises.
• Some companies have already pulled back on AI after token spend exceeded expectations, and now ration usage through allowances.
• He cites his own market study with The AI Report, which found seats are not as dead as people claim.
• What is dead, in his framing, is lazy seat-based pricing, not seats themselves.
• If the product works fine without a human involved, consumption pricing makes sense; if a person uses, accesses, or collaborates with it, user-based pricing or a hybrid of seats plus usage fits better.
• He warns against swinging the pendulum from everyone must charge for seats to everyone must abandon seats.

For brands planning a campaign

This is the post to cite when a pricing change is being justified by market fashion. Rivera's decision test, whether a human is in the loop, gives a team a defensible reason to keep seats in the model or to move away from them.

Drawing on a survey of 292 B2B software buyers for an AI pricing report, Rivera contrasts buyer priorities by company size.

• At companies under $50M in revenue, 77% rank predictable total cost as a top-three priority when evaluating an AI vendor; that drops to 63% once a company crosses $1B.
• Meanwhile, ease of justifying spend to finance and the CFO climbs from 44% to 55% across the same range.
• His reading is that smaller teams are trying to guess the bill while larger teams already know they will spend and need to defend that spend internally.
• He connects this to more than 480 pricing engagements and describes the budget conversation as never disappearing, only moving from what will this cost us to can we justify what we are already spending.
• His warning is that a pricing motion answering only one of those questions loses deals at the other end of the market.

For brands planning a campaign

A direct input for segmenting buyer communication. Predictability claims and bill estimators speak to smaller accounts; documentation, benchmarking, and CFO-ready justification speak to larger ones.

Rivera describes a personal habit rather than a pricing technique.

• He keeps pages of handwritten pricing notes, and says each one started as a question he could not answer yet.
• He uses AI tools, but draws a distinction: a prompt can hand you an answer in seconds, while conviction only shows up after you have wrestled with the problem yourself.
• Paper slows him down enough to work out what he actually believes, and the crossed-out sections are him being wrong on paper so he is not wrong in front of a client.
• He closes with a line about his own practice: pricing is the ultimate expression of business strategy, and you cannot copycat or prompt your way into a strategy.

For brands planning a campaign

Primarily a signal about working style for anyone evaluating Rivera. A brand or agency looking for an advisor who builds positions from a client's own numbers will read this differently from one looking for a rapid template.

What fails is charging for seats without looking at what the product does.

• He draws the line at the work: if the product runs fine with no human involved, consumption or usage pricing fits; if a person uses, accesses, or collaborates with it, user-based pricing or a hybrid of seats plus usage fits.
• He describes an executive dinner during a 2026 go-to-market conference where about 20 operators, investors, CEOs, and founders dug into the question, and where his takeaway was that the pendulum should not swing from everyone charges for seats to everyone abandons seats.

Rivera frames the decision as a pendulum and ends by asking readers where their own model sits on it.

For brands planning a campaign

Useful for a software company deciding whether AI features need a new pricing metric at all. Rivera's framing gives a brand a defensible middle position for a campaign: keep seats where humans still work alongside the product, and add usage where the product delivers value without them.

Examples: Source 1 · Source 2

He points to a Revenera survey of 415 software leaders in which 96% had shipped AI while almost half said customers barely use it, and only 7% reported losing seats to AI.

• He pairs that with a buyer survey of 296 respondents run with Benchmarkit, where 89% went over their AI budget mainly because AI features drove more usage than planned.
• His conclusion is that vendors cannot see usage and buyers cannot forecast it, and that no pricing metric fixes that blind spot.

For brands planning a campaign

Gives a brand a concrete, evidence-led hook for pricing content and campaigns: measurement before monetization. The three questions he says he answers before touching a metric can be offered to prospects as a practical self-assessment.

Examples: Source 1

He uses Atlassian's expansion of usage-based pricing around AI to press the point that making usage visible is not the same as making it predictable, and he gives three questions any team shipping a usage-based model should answer: can a buyer estimate their bill before signing, what happens when usage spikes, and does the metric track customer value or just infrastructure cost.

For brands planning a campaign

A brand selling AI features can borrow these three questions as a diagnostic or lead magnet, and can lead with predictability and spend controls rather than a headline price cut.

Examples: Source 1 · Source 2

He cites the same buyer research showing 55% find credit and token pricing harder to evaluate for AI than for traditional SaaS. His suggested fix is a credit model with very clear definitions and fewer definitions that look more like outcomes, since fewer actions make consumption more predictable.

He develops the argument with team riff sessions and everyday analogies, then closes with a direct question about how many actions sit on the reader's rate card.

For brands planning a campaign

Directly relevant to a company designing usage-based pricing for AI features: it argues for simplifying the rate card and moving credit definitions toward outcomes, a concrete change a brand can pitch or adopt.

Examples: Source 1

At companies under $50M, 77% rank predictable total cost as a top-three priority when evaluating an AI vendor, and that falls to 63% above $1B in revenue.

• At the same time, easy to justify to Finance and the CFO rises from 44% to 55% across that range.
• His read is that smaller teams are trying to guess the bill while larger teams already expect to spend and need help defending it, a pattern he says he has seen across 480-plus pricing engagements.

For brands planning a campaign

Lets a brand split its message by target segment: cost predictability and estimation tools for smaller buyers, CFO-ready justification and reporting for enterprise buyers.

Examples: Source 1

He describes these as breaking at the same time, and repeats that seats can remain part of the answer but cannot carry the whole model once AI does real work inside the product.

For brands planning a campaign

A compact framework a brand can use in a webinar, article, or sales conversation to explain why AI monetization needs a different design process rather than a renamed seat plan.

Examples: Source 1

He co-produces an AI Pricing Report with Benchmarkit and cites its findings repeatedly across posts on predictable cost, credit models, buyer-size differences, and usage measurement.

• He also references a Revenera survey of 415 software leaders and his own base of 480-plus pricing engagements.
• Several posts end with an invitation to download the full report.

For brands planning a campaign

Suggests a data co-marketing or research-collaboration angle: a brand with access to pricing or buyer data could partner on a study and reach an audience already primed to share results.

Examples: Source 1 · Source 2 · Source 3 · Source 4

Posts combine a provocative opening line, numbered or bulleted reasoning, survey figures, and a closing question to the reader.

• He also posts personal material, including a note about turning down four New York event dinners to see his brother and a post thanking a colleague for a shoutout alongside a pricing thought.
• One post describes his practice of keeping handwritten pages of pricing notes and wrestling with a problem on paper before letting AI join the conversation.

He treats questions he cannot yet answer as the starting point, writes them out by hand, and works through them before publishing.

Not every post is about pricing. Family and relationship posts sit alongside the commercial material and show a broader register a brand would be buying into.

For brands planning a campaign

A brand can expect essay-style posts tied to real events and published research, with room for human-interest material. The handwritten-notes habit supports a credible craft narrative about how his views are formed.

Examples: Source 1 · Source 2 · Source 3 · Source 4 · Source 5 · Source 6

The post reflects the company back to readers without stating Rivera's formal role or title.

Original disclosure

The post does not state whether Rivera owns, is employed by, or contracts with Pricing I/O; it describes the company as "we" and "our."

View example

The post does not say what Rivera's connection to Currv is or whether the company owns it.

Original disclosure

Relationship unstated in the post.

View example

The research underpins Rivera's pricing arguments rather than serving as a one-off name-drop.

View example

View example

View example

View example

The posts show participation in the community, not a stated commercial agreement.

View example

Both treat it as a cross-functional issue spanning product, sales, finance, and legal, and both address an executive audience rather than a technical one.

Moogimane's posts read as advisory and media commentary, including a CIO Dive interview and a panel appearance alongside enterprise go-to-market speakers, and lay out parallel lists of provider and buyer challenges. Rivera's posts are built around his own survey work with Benchmarkit and a base of pricing engagements, and close by asking readers a direct question about their own model.

Read original · Original 2

Both make a timing argument: information that arrives after the decision has been made cannot change the decision.

Martins writes from the procurement side, about how much mid-market companies overspend and how benchmarks can reach buyers inside a procurement workflow, and her retained posts are short product and event announcements. Rivera writes from the vendor side about designing the pricing model itself, in longer analytical posts that walk through the reasoning.

Read original · Original 2

Rivera's post about the narrative running past reality works the same way.

Howe writes about finance and accounts-payable automation from a product-building perspective, covering invoice capture, coding, and approval routing at a specific company. Rivera writes about pricing and monetization strategy for software vendors, and reaches for survey data and executive-dinner discussions where Howe reaches for operational detail.

Read original · Original 2

His reviewed posts frame AI pricing as a cross-functional problem: providers have to weigh outcome-based pricing against margin risk, build guardrails on AI-related costs, and treat pricing as a product, sales, finance, and legal issue, while buyers are managing tool sprawl, chasing transparency, and trying to keep flexibility.

• A second post promotes a session with ZoomInfo and Google on why go-to-market AI pilots stall, why data quality and context matter, and how teams move from experimentation to business outcomes.
• For a B2B software company launching usage-based AI pricing into enterprise accounts, that material speaks to the conditions that decide whether the product is adopted at all: who signs off, whether the data underneath is ready, and how buyers govern spend across overlapping tools.

You give up metric-level pricing design.

• The reviewed Moogimane posts do not work through credit models, rate cards, spike policies, or how a buyer estimates a bill before signing, which is where Rivera's three-question test and credit-model fix sit.
• Rivera's reviewed material is also more specific about the pricing decision itself, including the seat-versus-usage split and the buyer-size difference in what matters.
• Moogimane's reviewed evidence is two posts, one of them an event announcement, so the enterprise framing is broader but thinner on pricing method than Rivera's output.

Examples: Source 1 · Source 2

Her reviewed posts describe SpendHound benchmarks drawn from a 10,000+ vendor dataset based on what more than 1,500 companies pay, now available inside Zip's procurement workflow, and a ProcureCon East push inviting procurement teams to compare their SaaS, AI, and cloud pricing against those benchmarks.

• For a B2B software company rolling out usage-based AI pricing, that is the purchasing-side evidence a prospect asks for when they want to know whether the number on the contract is normal.
• It answers a different part of the same campaign decision: not how to design the meter, but what the market pays and how procurement will test the price.

The tradeoff is positioning and depth.

• This is spend benchmarking and negotiation leverage aimed at buyers, not pricing-model design aimed at vendors.
• The reviewed posts do not address how to structure a usage or credit model, how to reduce the number of rated actions, or how to explain a new meter to a prospect; they support a claim about price levels and procurement workflow.
• The two reviewed posts are short product and event updates, so they establish the benchmark offer but not a full method, and they leave Rivera's pricing-design questions unanswered.

Examples: Source 1 · Source 2

His reviewed posts argue there is a wide gap between how AI in finance is discussed and how it is used: named agents ship in every major ledger and most CFOs report some agentic use, yet around two-thirds of AP teams still key invoices by hand, and most finance teams have caught AI fabricating data.

• The pattern Howe describes is deterministic systems carrying repeatable volume, AI handling judgment calls, and a human approving before money moves.
• A second post covers MakersHub's rebuilt approval workflow, where rules combine conditions such as vendor, project, entity, and characters inside an invoice number.
• For a software company selling AI into finance, that is the operational argument that decides whether finance will let the system touch the ledger.

You give up pricing content entirely.

• The reviewed Howe posts do not cover pricing metrics, buyer cost predictability, or vendor pricing strategy; they cover finance operations, approval routing, and AP automation.
• The audience is controllers, AP teams, and finance leaders rather than pricing and product leaders.
• The material is also tied to MakersHub's product, so it carries a vendor's perspective on the workflow it sells.
• A campaign built on Howe answers a control-and-trust brief, while Rivera's reviewed material answers a pricing-and-packaging brief.

Examples: Source 1 · Source 2