Phil Crumm

SaaS-Margin Agencies Have a Half-Life

phil@fulcrumm.com (Phil Crumm) — Sun, 14 Jun 2026 12:00:00 GMT

When the example is the indictment

Y Combinator's Spring 2026 Request for Startups gave three examples of what an AI-powered agency might look like. Read them straight:

Think of a design firm that uses AI to produce custom design work for clients upfront, to win the business before the contract is even signed. Or an ad agency that uses AI to create stunning video ads without the time and expense of setting up a physical shoot. Or a law firm that uses AI to write legal docs in minutes, rather than weeks.

Read them again and ask the question they invite.

Why are those not just SaaS businesses?

A design tool that produces custom design upfront is a design tool. An ad-creation tool that produces video without a physical shoot is an ad-creation tool. A legal-document tool that writes contracts in minutes is a legal-document tool. The framing dresses each one in agency clothing — a firm, a roster, a relationship — but the unit of value is the SaaS underneath. The agency layer is decorative.

This essay is about that decoration. About why dressing a SaaS up as an agency creates a business that prints money for a window and then doesn't compound. About what the RFS got right that nobody is saying out loud, and what it got wrong that everybody is repeating.

The thesis comes in two pieces.

The first is an indictment. The agency that builds itself around a productized AI workflow is not a competitor to SaaS. It is a feeder to it. Every contract such a shop signs is free demand validation and free workflow standardization for the SaaS company that will eventually ship the same thing and own the category. The productized-services shop is the R&D arm of its eventual replacement, paid for by clients who don't yet know they're funding it.

The second is what survives. The agencies that compound from here are the ones flipping the unit of sale entirely — selling judgment, opinion, and accountability under ambiguity. The work that doesn't survive a price card. The position that sits above where AI keeps climbing, because the climb is structurally uneven.

Both are true at the same time. The indictment is the engine of the argument. The alternative is the way out.

The feeder thesis

Picture the design firm from the RFS. Founders have built a tool that produces high-quality custom design output for a prospect upfront. The pitch is the product. The product wins the contract. Then the firm uses the same tool to deliver more design across the engagement, billed by the deliverable rather than the hour, with humans hovering over the AI to keep it on-brand and on-spec.

This works. It might work very well for a window.

Then watch what happens next.

The firm signs forty clients. Each client is a different brand context — automotive, healthcare, consumer fintech, B2B SaaS. Each engagement teaches the firm's tool how to handle a different vertical. Each set of feedback rounds teaches it what "good" looks like in a different domain. Each contract is, mechanically, a labeled training set for the firm's workflow — what works, what fails, what clients reject, what they sign off on.

The firm now has something valuable. A tool, plus the operational knowledge to run it, plus contractual relationships that prove it produces commercial outcomes.

The firm also has a problem. The same value is visible to Figma, to Canva, to Adobe, to Framer, to Anthropic, to OpenAI, and to a hundred well-funded startups whose business it is to be visible to that exact opportunity. The firm has, by virtue of doing the work, shown all of them where the demand is, what the workflow looks like, and how much customers will pay for it. The firm has performed the most expensive parts of SaaS product discovery — at its own cost, with its own capital, on its own roster of clients — and published the answer.

The diagram makes the loop visible. Demand flows from clients to the shop. The shop's existence flows, as legible market data, to every SaaS company looking at the category. The SaaS company ships parity. The customers route to the SaaS company. The shop is left holding workflow expertise that the SaaS company has now generalized and embedded.

This is roughly the history of the no-code consulting market between 2018 and 2023. It is roughly the history of the WordPress build-and-run market between 2010 and 2020. It is the shape of every productized-services arbitrage on a maturing platform. The pattern repeats because the economics are structural.

The naive read of the RFS is that productized services lets a firm capture SaaS-shaped margin by removing humans from delivery. The structural read is that productized services lets a firm transfer the value it creates — demand, workflow, willingness-to-pay — out of its own books and into the books of whichever SaaS company is paying attention. The arbitrage is real, and it is routinely mistaken for a business model.

Why the moat moved upstack

The historical agency moat sat in a particular slice of work: standardized output produced repeatedly for paying clients. Logos, websites, landing pages, video edits, simple legal documents, monthly content calendars, basic media buys. This was the work that filled the calendar and paid the salaries. The moat was the relationship plus the production capacity, and both held for decades.

SaaS ate that slice. Logo design went to Looka and a thousand AI-image variants. Landing pages went to Framer and Webflow templates. Simple copy went to Jasper and a parade of GPT wrappers. Basic media buys went to programmatic platforms with self-serve UIs. Each new SaaS that won a category did the same thing: took the standardized output, automated the production, and priced against the SaaS curve rather than the agency curve.

The moat moved upstack.

What survived in agency form is the work above where SaaS could reach. Brand systems that have to hold across years of campaigns. Monetization strategy for publishers with idiosyncratic audiences. Discovery work where the client doesn't yet know what they're buying. Negotiated outcomes that require accountability for ambiguous deliverables. Senior judgment applied to messy organizational situations. These categories share a property: they resist being priced against a deliverable. The value sits in the framing, the choice, the bet on what to do. The artifact at the end is incidental.

SaaS bounced off this work because price-cardable means commoditizable. The moment a clean number can be put next to "brand strategy for a category-creating fintech," somebody has either already productized it or is about to, and the productized version trades at productized prices.

The productized-services shop operates in the layer SaaS already eats. It automates inside that slice and sells the automation as if it were upstack work. The slice below the moat, dressed in moat clothing, priced as if the moat were still there. The pricing power lasts as long as customers don't notice the dress.

The ladder AI climbs unevenly

AI capability climbs unevenly. The unevenness is structural, built into who builds the models and which pain those builders feel acutely.

Code has been eaten fastest. Front-end implementation, scaffolded back-end work, infrastructure as code, test generation, code review on familiar patterns — all of it has been pulled forward dramatically in the last eighteen months. Junior engineers shipping at mid-level pace on familiar stacks is common. The productivity step-change in code is the largest in the AI era so far.

Design has moved more slowly. AI produces competent design output and stops short of design judgment — when to break a pattern, why a brand should resist a trend, what a piece of work means in the context of a client's category position. The handoff between AI design output and a usable production artifact still requires a human in the middle, and the middle is the work.

Strategy and monetization have moved slowest of all. There are reasons, but the most useful one is mechanical. AI models are built by engineers, for engineers, against problems engineers feel acutely. The model labs are full of people who write code every day and feel the pain of code being slow. The pain of running pricing experiments on subscription products, negotiating brand-coherence concessions across multi-year client relationships, or sitting with publishers as programmatic CPMs decline — that pain lives somewhere else, in a population the labs are not training against.

The ladder will move. The order will reshuffle. Strategy will be eaten too, eventually. "Eventually" matters here. A productized-services shop betting on AI's current capability has eighteen to thirty-six months before the layer it's operating in is fully absorbed. A POV-led firm betting on judgment and accountability under ambiguity is betting on a layer the model labs are not currently pointed at. The latter is a longer bet on a more defensible position.

The productized-services shop is taking the short bet on the layer where the climb is fastest. The math is brutal.

Distribution economics and the worst of both worlds

The case for SaaS-margin agencies leans implicitly on a misreading of where SaaS margins come from.

SaaS margins come from distribution economics. One product, sold to many buyers, with marginal cost approximately zero on the next sale. The same code that serves the first customer serves the thousandth. Customer acquisition cost amortizes across an account that auto-renews for years. Gross margin compounds because the cost structure scales sub-linearly with revenue.

The SaaS-margin agency runs on agency distribution.

Each engagement is sold individually. Each engagement involves a sales cycle measured in weeks or months. Each engagement requires account management, scope negotiation, and the kind of senior-led pitch process that doesn't scale. The shop's customer acquisition cost looks like an agency's — proposals, decks, conference attendance, network maintenance, senior partner time burned on pitches.

What the productized-services shop has is SaaS-shaped delivery costs. Software, compute, model calls, the engineering to keep the workflow integrated. These costs are real and scale, just not as steeply as labor.

Stack the two together and the picture clarifies. Delivery costs that look like SaaS. Distribution costs that look like services. The combination is worse than either pure model, because neither curve is the one that produces SaaS margins. The shop is paying SaaS infrastructure costs to deliver while paying services-business GTM costs to sell. The margin window between those two cost structures is the arbitrage. It is real. It is also narrow and closing.

The SaaS company that ships parity has the right distribution economics for its delivery cost structure. The productized-services shop has neither.

Roadmap debt

Year one looks good. Margins improve. Headcount stays roughly flat against revenue. The pitch lands because the proof is in the work.

Year two is where the structural cost arrives.

The SaaS company that owns the category — or the well-funded startup positioning to own it — observes the productized-services shop and ships parity. The SaaS company ships it embedded in the platform clients already use, with integrations the shop never had, with a roadmap of adjacent features the shop cannot afford to build, and at a price that reflects SaaS economics rather than services economics.

The productized-services shop has defenses. They are worth steel-manning.

Switching costs are real. A client who has integrated the shop's workflow into procurement, creative approval, and project management does not switch on a dime. Embedded process is sticky.

Brand matters. A client who trusts a specific firm with a specific partner does not always re-evaluate when a cheaper option ships.

Integrations are non-trivial. The shop has built bespoke pipes into client systems that the SaaS company has to reconstruct.

These defenses buy time without changing direction. The SaaS company closes the gap on each one — switching cost erodes the moment the SaaS company's parity feature gets cheap enough; brand erodes the moment a procurement team's CFO asks why the line item is twice the alternative; integration debt is exactly the kind of work AI is best at compressing. The defenses degrade on the same curve that SaaS adoption follows everywhere else.

Customers, asked to choose between an agency that wraps a workflow and a SaaS platform that owns it natively, choose the platform. They get more from it. They pay less for it. They absorb the switching cost once the gap is large enough to notice.

The roadmap debt is paid in irrelevance.

The eighteen-to-thirty-six month window

The arbitrage window has a shape, and the shape is observable.

Clients buying AI-augmented services today are eighteen months into a particular conversation. The conversation began with "should we use AI to make this work faster" and has matured into "the work happens with AI now, and we are paying for the result rather than the process." That maturation took roughly eighteen months from first-meaningful-deployment to baseline expectation.

The next eighteen months follow the same arc, on the same kind of conversation, with a different question. Clients now paying for AI-augmented services will start asking the second question. The second question is: if the AI is doing the work, why are we paying you to operate it rather than operating it ourselves. That question is already being asked in private. It will be weight on every contract negotiation soon.

The math of the window is straightforward. The arbitrage opens when the SaaS doesn't yet exist for a given workflow and the client is not yet sophisticated enough to operate the AI directly. It closes when either condition flips. SaaS coverage is expanding monthly. Client sophistication is expanding monthly. The window is open in both directions and closing in both directions.

Eighteen to thirty-six months is a generous read. A specific shop in a specific category may have less. A shop in a less mature category may have more. The shape of the window — open now, closing on two sides — is the structural fact. The exact length is the variable.

Take the arbitrage when it appears. Refuse to mistake it for a business model. Capture the window, take the cash, do not build an organization that assumes the window stays open. The shops that compound do something different.

What survives

The agencies that compound from here are the ones flipping the unit of sale.

The unit of sale that survives is judgment, applied to ambiguous problems, with accountability for outcomes. What clients buy is a point of view on what they should do, why, and at what cost — and the willingness of the firm to stake its name on the answer being right. The deliverable is a decision, supported by artifacts, taken on the client's behalf.

The shops still selling discovery-as-billable-hours and calling it strategy will not survive the next thirty-six months. The shops building a public point of view — pricing it, publishing it, betting compensation structures on it — will. The first kind operates below the moat that AI is steadily relocating upward. The second kind operates inside the moat and will keep operating there because the moat is not, structurally, where the next generation of foundation models is being pointed.

Three things mark the difference between a shop with a point of view and a shop performing the trend deck.

The first is pricing. A shop with a point of view prices against the decision. The price reflects the value of being right rather than the cost of the hours or the artifacts that produced the work. A shop performing strategy prices the way it has always priced and labels the result strategic.

The second is publication. A shop with a point of view publishes it. Essays. Talks. Public positions on what its category is getting wrong. A shop performing strategy circulates trend decks privately and keeps the position vague enough to retreat from when a client disagrees.

The third is compensation. A shop with a point of view structures compensation around the strategic call — partners and senior operators paid against the decisions they made, the bets that paid off, the clients they walked away from. A shop performing strategy pays on utilization and revenue per head, on the metrics that grade execution.

A shop hitting all three has a moat. A shop hitting one or two is in transition. A shop hitting zero is operating below the line where AI is climbing, and the climb is faster than the lease on the office.

The right read of the RFS

Y Combinator was looking at something real. The delivery model is broken. Hours are the wrong unit. AI changes the labor math underneath every services category in a way that hasn't fully landed in P&Ls yet. The RFS is correct in diagnosis and the funding will produce companies.

The RFS is wrong in prescription. The companies it will produce are the feeders. They will validate categories. They will standardize workflows. They will publish, by virtue of doing the work, the answers the next generation of SaaS companies need to build inside their core platforms. Some will be acquired by those SaaS companies in the kind of bolt-on deal that pays the founders well and ends the firm. Most will simply run out of distance.

The agencies that compound — that grow into durable services businesses operating at margins approaching but never matching SaaS — are doing the opposite of what the RFS asked for. They concentrate senior judgment in the firm. They price judgment explicitly. They refuse the work that doesn't carry strategic stakes. They operate above the climbing line. They publish a position the founder is willing to defend. They build compensation around the decisions they make rather than the hours they bill.

The meter is the bug. The position below the moat is the other bug. SaaS-margin agencies fix neither; the agencies that compound from here fix both.

Storytelling Is the Whole Job Now

phil@fulcrumm.com (Phil Crumm) — Wed, 03 Jun 2026 12:00:00 GMT

An imposter cat

In July 2023 my wife posted a video about a cat named Oyen.

Oyen is an orange tabby. During the pandemic he wandered into the capybara habitat at Malaysia's Zoo Negara as a kitten, presumed orphaned. The capybaras adopted him on sight. He eats with them, sleeps with them, cuddles with them. In 2023 the zoo gave up pretending he wasn't theirs and put his own signage on the habitat next to the capybaras'. The deputy president has said something close to ninety percent of visitors now come hoping to see him.

The wire-service version of this story is forgettable. Cat lives with capybaras at zoo. End of beat.

Jenn's version opened with a deadpan reframe: "An imposter cat has seemingly been accepted by a herd of capybaras." She let the absurdity sit, then turned tender — orphaned kitten, found a family, the capybaras "really don't seem to mind at all." The tension was the quiet wait, this is actually real? The payoff arrived with the signage — the zoo, the institution, officially recognizing what the animals had already decided.

The post pulled 31.2 million views. She's revisited the beat twice since — a year-end recap hit 7.2M, an earlier version 4.2M.

Same story. Different craft. Three orders of magnitude in reach.

That delta is the whole essay. Storytelling is the most important capability a company can build, because engagement and distribution are the only two non-substitutable dependencies of a business, and both run on it.

The substitution test

Run the substitution test against everything a company does.

Product can be substituted. Better versions get built; worse versions get replaced. The product you sell in five years won't be the product you sell today, and your company can survive the swap if everything else is intact.

Operations can be substituted. Outsource the back office. Buy a different ERP. Restructure the team. Painful, recoverable.

People can be substituted. Operators don't like saying this out loud, but the proof is in every acquisition that retains sixty percent of headcount and ships better in year two. Roles matter; specific occupants are mostly fungible at the org-design level. Even founders get replaced.

Capital can be substituted. Different round, different lender, different ratio of debt to equity. Money is the most substitutable input in the entire business.

Brand assets — logo, palette, tagline — can be substituted. The number of rebrands that produced no measurable business effect is roughly the number of rebrands that have ever happened.

What can't you substitute?

You can't substitute the fact that someone cares about your company. You can't substitute the channels through which they encounter you. Strip the rest away and these are the two dependencies that hold the rest of the business up. Engagement and distribution. People have to care, and people have to find you.

Almost every other failure mode is recoverable. Lose engagement and distribution and you are no longer running a company. You are operating a service no one needs.

Both run on story.

Brains run on stories

There's a 2010 study out of Princeton — Stephens, Silbert, and Hasson, in PNAS — where researchers put speakers and listeners in fMRI scanners and watched what happened during storytelling. The listener's brain activity began to mirror the speaker's. The coupling was measurable, and stronger coupling predicted better comprehension.

This gets misquoted into a stock photo of a brain with arrows. The actual finding is sharper: story is a transmission protocol. It is the mechanism by which one nervous system's state is loaded into another. The pattern recognition, the anticipation, the resolution — these are the structure that makes neural coupling possible at all.

When a story works, the craft is doing the same thing every time. Stakes you can recognize. Tension you carry across the gap. A payoff that resolves what the opening set up. Imposter cat → capybaras don't mind → zoo makes it official. Setup, tension, resolution. The form is older than the species.

Decks, dashboards, and bullet lists are data dumps. They hand the listener a pile of facts and ask them to assemble the connective tissue themselves, and most listeners won't.

The steel-man: distribution mechanics

The strongest case against this thesis is the infrastructure-platform case.

Stripe scaled with almost no narrative voice. So did Linear. So did AWS in its early decade. Their growth came from API surface area, developer experience, integration availability, and the boring discipline of being reliable. The customer-facing narrative was something like we exist, the docs are good, the SDKs work in your language. No hero's journey. No founder mythology in the launch posts.

The argument follows: if these companies built generation-defining businesses with vestigial storytelling, then story is decorative for serious operators, and what actually scales is distribution mechanics — paid acquisition, product loops, integrations, time-to-first-value.

The argument is mostly right inside its frame and almost entirely wrong outside it.

Inside the frame: technical-buyer audiences process credibility through artifacts. The README carries the story. The latency graph is the hook. The status page is the trust signal. Peer reputation among engineers — what the senior backend at Shopify says about your SDK at the bar at re:Invent — is itself a story, transmitted in a register that doesn't read as narrative to anyone outside the room. Hook, tension, payoff. The hook is I tried this on a side project. The tension is the docs were not lying. The payoff is we moved the whole company onto it.

Outside the frame — i.e., for the ninety-nine percent of businesses that don't sell infrastructure to engineers — the substitution test still holds and the steel-man dissolves. Most companies don't have a Stripe-shaped audience. They have a people who don't yet care audience, and the only way to convert that audience is to be the kind of company someone tells a story about.

The end of volume strategy

For two decades the dominant content posture was more at-bats. Publish more, post more, optimize for the algorithm's appetite. The implicit math: production is the constraint, distribution is the lever, some percentage of attempts go viral, scale beats craft.

This is over.

Production cost has collapsed. A competent model produces an em-dash-heavy LinkedIn post in eight seconds. Three of those, a header image, a carousel — call it forty seconds for an entire content unit indistinguishable from the median post in the feed. The marginal cost of corporate content has gone from one person-hour to one prompt.

Attention is the constraint now.

Open LinkedIn. Scroll. You can feel the homogenization in your jaw. The same cadence on every fourth post — a punchy first line, three bullets, a one-line bow. The same em-dashes. The same fake-vulnerable opener. The same imitation of the format that worked for the last person who tried it. The feed has the texture of a single model producing a single voice from ten thousand accounts, because in some functional sense that is what is happening.

Story survives this — specifically the part of story the model cannot synthesize because it requires the specificity of having been somewhere and seen something. The capybara cat with the official signage. The twenty-two-year-old social coordinator who put a twerking owl on TikTok. The thing that happened to your team last Thursday.

The volume era rewarded competence at producing the average. The story era punishes it.

The Duolingo flip

In early 2021, Duolingo's TikTok account had about fifty thousand followers and was producing the kind of content you'd expect — language-learning tips, mascot waves, faintly aspirational copy. The default corporate social output of a company that took itself appropriately seriously.

That year a twenty-two-year-old junior social media coordinator named Zaria Parvez was handed the account.

In September of 2021 Parvez had the Duo owl twerk on an intern in response to a comment about Dua Lipa. The video pulled roughly three million views in days. The unhinged Duo persona — horny, threatening, chronically online, weighing in on Taylor Swift and The Bachelor and ghosting users who skipped lessons — locked in. The account crossed five million followers inside eighteen months. It now sits north of sixteen million.

The owl is the trailing indicator. The interesting change happened upstream of the content.

To make the unhinged Duo work, Duolingo decommissioned the approval pipeline for that surface. One person ran the account in close-to-real-time. The legal-marketing-comms chain that produces median-quality corporate content was switched off. The company decided that moving fast and occasionally producing something cringe-inducing cost less than producing nothing memorable, ever.

Duolingo had made a real decision about what kind of company it was going to be. The owl was downstream of that decision; so was the operating model around the social team; so was the willingness to live with the occasional bad-take that comes with running an account in close-to-real-time. The content was just the visible part of a much bigger commitment.

Most companies will continue running their social channels through the same approval pipeline that produces median content, then hire a "head of content" to fix it, and the hire will fail because the hire has no authority to change the pipeline. The pipeline is what produces the content. The content can't be fixed without fixing the pipeline.

The masthead signal

Look at who's hiring for editorial leadership right now.

Visa is hiring an Editor in Chief at $153K–$245K. Pottery Barn, a Managing Editor at $125K–$130K. Toast, a Senior Director of Content & Audience. BNY, an SVP for Markets Content. Guy Carpenter, a Director of Content & Thought Leadership at up to $244K. Bunkerhill Health. Profound. Prenuvo. The Anti-Defamation League. The list spans payments, retail, reinsurance, healthcare, AI infrastructure, banking, and advocacy.

Two things are interesting about the list.

The first is the titling. Editor in Chief. SVP Markets Content. These are publishing-house framings being adopted by operating companies that don't publish anything. The work is being modeled as masthead work — singular editorial authority, durable seniority, organizational permanence — rather than as content marketing.

The second is the comp. $244K at the top of the band is director-VP territory. Companies do not pay that to fill a tactical seat. They pay that when they believe the function decides whether the company works.

What this list shows is companies waking up to where competitive advantage now lives. Storytelling capability is being moved from a marketing line item with a quarterly budget to a senior, durable organizational function with multi-year commitment. The trend is real, and the trend is right. Whether each individual hire succeeds will depend on whether the company rebuilds the operating model around the function — the way Duolingo did — or whether they import the title without making the structural changes that let the work do anything.

Internal storytelling is the same craft

The external version of this argument is the easier version. Engagement and distribution sit outside the company; story unlocks them; therefore story is a function the company must build. Fine.

The internal version is harder to see and does more of the work.

Every meaningful change inside a company requires the same craft. There is a current state someone has to understand. There is a tension — a thing that's not working, a thing that's coming that we're not ready for, a thing we said we'd do that we haven't done. There is a proposed resolution and a set of stakes if we don't get there. Setup, tension, payoff. The same form, transmitted internally instead of externally.

Retention runs on this. The reason people stay at companies for a decade is rarely the comp. They stay because they have a coherent story about why they are there, what the work is for, and where it's going. When that story stops making sense — when the company can't articulate it, or articulates a new one every quarter, or articulates one thing and operates by another — retention erodes from the senior end first, which is the most expensive end.

Alignment runs on this. OKRs are the artifact. The story underneath the OKRs is what makes them sortable when reality forces tradeoffs. Without a coherent internal story, every priority conflict becomes a political negotiation rather than a clarifying choice.

Recruiting runs on this. Great senior people take a meeting because of the story they've heard about the company from someone they trust. They sign because of the story you tell them in the room. Compensation closes the deal; story decides whether you ever got the chance to make the offer.

The companies that do internal storytelling well treat it as one operating capability with two surfaces, not two unrelated marketing functions. The external story and the internal story are the same story, told in different registers, by people across the organization who are fluent in both.

What companies that take this seriously actually do

The companies that treat storytelling as a real capability do four things differently.

They invest in it durably. Senior ownership, real headcount, multi-year horizon — the same operating commitment a company makes to product engineering or finance. The masthead-hiring trend is the trailing indicator. Companies that meant it built a permanent function at the table where capital and people get allocated. Companies that didn't built a content-marketing pod reporting up through demand-gen, and the work continues to do roughly what a content-marketing pod can do.

They rebuild the operating model around it. Approval chains shorten or get removed for the surfaces where story craft matters. The legal-marketing-comms gauntlet that produces median content gets reorganized. Duolingo decommissioned the TikTok approval pipeline. They did not do this lightly. They did it because the alternative was producing nothing memorable, ever, and they understood that the cost of the alternative was the entire point of having a social channel.

They get fluent across the organization. The story isn't sequestered in the editorial team. Every leader can tell it. Every salesperson can tell a tighter version. Every recruiter can tell it to a candidate. The narrative is internal and external in parallel — same thesis, different registers — and it stays consistent because every person retelling it knows what the company is for in a way that survives translation.

They commit to specificity. The model can produce competent average; companies treating storytelling as real capability know their advantage lives in the things only they have seen, only they have done, only their customers have lived. They invest in capturing those specifics — in operator interviews, in customer journalism, in the unglamorous editorial work of finding the particular thing worth telling. Then they tell it.

The companies that do these four things own the only two business inputs their competitors cannot substitute. The companies that don't will continue posting median-quality content into an AI-flooded feed and wondering why the volume math stopped working.

Storytelling is the work, not the wrapper

For most of the past two decades, businesses have treated storytelling as the wrapper around the work — the thing you do after you've built the product, run the service, hired the team. Marketing's problem. Comms's problem. The agency's problem. A function downstream of the real business.

That model assumed two things that have stopped being true. It assumed that production was the constraint on reaching people. And it assumed that the default state of the channels — search, social, the open web — was that competent corporate content could occupy them by showing up consistently.

Production is free now. The channels are saturated with model-grade average. The default state has flipped: showing up consistently with competent content reaches almost no one, because everyone is showing up consistently with competent content.

What survives in the new default is craft. Specificity. The thing a model cannot synthesize because it requires having been somewhere and seen something — a capybara habitat with a cat in it, a social coordinator with the freedom to post the bit, a company willing to look slightly cringe-inducing in pursuit of being memorable instead of fine. The companies that treat storytelling as the work itself — as what the business is actually producing, alongside the product or the service — are the companies that will own engagement and distribution. Everyone else is competing for what's left.

Storytelling is the most important capability a company can build now. Run the substitution test on your own business and see what it tells you.

T&M Punishes the Efficient

phil@fulcrumm.com (Phil Crumm) — Mon, 18 May 2026 12:00:00 GMT

Whose hour is it?

Imagine one engineer, three Claude agents, three clients, one wall-clock hour.

Each task is real work. Each task takes its full estimated time. The agents are not doing trivial things in parallel while a human does the real one. They are doing three substantive pieces of work at once, with the engineer steering, reviewing, intervening — the way the best people in my practice already work, and the way most of us will be working soon enough.

Whose hour is it?

Time-and-materials billing has one answer to that question and it is not a good one. Pick one. Bill the elapsed wall-clock to whichever client your conscience nominates. Don't think too hard about it.

There is no honest answer inside the model. T&M was built on an assumption — one body, one task, one timer — and that assumption is failing in front of us, on Tuesday afternoons, in shared screens, in commit histories, in the part of the timesheet your senior engineers fill out last.

I spent the last four months in conference rooms with my practice leadership trying to figure out how to bill three agents at once. I did not get a clean answer. I got something better, which was clarity about why the question is unanswerable inside T&M, and what has to change for the question to stop being asked at all.

The model T&M was built for

Hour-of-labor was a defensible proxy for value-delivered when human throughput was roughly linear and there was no other way to fake it.

If a developer billed forty hours, you got roughly forty hours of developer-grade output. If they billed sixty, you got roughly sixty. The mapping was lossy but monotonic, and the proxy survived because it had to. Nothing else was countable. Story points were a planning fiction, not a billing artifact. Outcomes were too fuzzy. Hours were what the courts understood, what procurement understood, what your CFO understood. Hours were the lingua franca and they earned the role.

The proxy held because the world held. The world stopped holding.

This is a structural problem that represents itself as an individual behavior problem. When utilization slips, when a senior engineer "loses" a billable hour to mentoring, when an estimate gets revised down mid-project, when somebody quietly bills elapsed wall-clock for two agents running in tmux — none of these are character flaws. They are the structure failing in the only way it knows how to fail, which is by routing the failure through individual people and asking them to absorb it.

The structure made sense in 1998. It made sense in 2008. It made sense in 2018. It is failing in 2026. Hours stopped being the unit of value, and the meter that measures hours is now measuring the wrong thing while everybody inside the building gets graded on it.

The four ways T&M is now broken

There are four. Each is a no, but.

It punishes the efficient

Every hour you save with AI is an hour of revenue you didn't bill.

Pick your best engineer. Give them Claude Code, Cursor, a tmux setup, and the kind of taste it takes to use those tools well. Now watch what happens to their billing. Better team, worse P&L. The incentive runs the wrong direction.

Pull the data from any T&M shop with a tenured practice and you'll see the tell. Variance between estimates and actuals is high for junior people and low — strangely, suspiciously, structurally low — for senior people. I just fundamentally don't believe that variance compression is a product of senior engineers being really good at guessing how long things take. It is a product of them being really good at managing the meter.

It hides AI cost in the wrong column

Agency billing rates were set against a labor cost curve. They were set against salaries and benefits and overhead and a margin target that had two decimal places.

They were not set against a two-hundred-dollar-per-seat Claude bill and a Cursor org plan and a compute line that grows with adoption. Those costs are real. They show up monthly. They scale with the engineers who are best at using them — which means they scale exactly with the people whose hours are simultaneously falling.

You can do the math two ways and neither one ends well. Charge a premium for AI-augmented work and you're telling the client the meter is even more arbitrary than they suspected. Don't charge a premium and you eat the cost on the margin line nobody re-priced. I have not met an agency doing otherwise. The margin gets eaten from a column nobody priced, and the people who priced the labor line aren't the people running the AI line.

Estimates get more wrong, not less, mid-project

The team that delivers in week eight is materially faster than the team that estimated in week zero.

This was always somewhat true — teams learn the domain mid-project. AI has made it dramatically true. The statement of work described a slower team than the team executing the work, and the gap widens every sprint. The estimate is a fixed point. The team is not.

What happens next is predictable and human. People smooth hours. Not in a dishonest way — in a make-the-meter-make-sense way. The engineer who finishes a five-hour task in two hours bills four, because that's what the SOW expected, because billing two would invite questions about the estimate, because the next task on the backlog needs to be available for them to roll into and the meter doesn't know how to express "rolled in early." The hours arrive at the line item the SOW called for them to arrive at. The work didn't actually happen that way. You have trained your best people to be quietly dishonest with you, and they know they're doing it, and you know they're doing it, and the meter keeps running.

It can't price parallelism

Return to the cold open. One engineer. Three agents. Three clients. One hour.

One body, one task, one timer is the assumption underneath every T&M invoice ever written. Strip it and the bookkeeping collapses. Bill elapsed to one client and you've stolen from the other two. Bill task-time to all three and you've billed three hours of wall-clock for one. The contractor you hired to do the same kind of work for you is contractually forbidden from billing this way. Pull a standard agency-to-subcontractor agreement and you will find a clause — sometimes two — that says the sub may not bill the same hour to multiple engagements, may not bill in parallel, may not invoice for unattended automation. Those clauses are in there because the agency wrote them, because the agency understood, in the specific case where it was the buyer, that the meter cannot survive parallelism without becoming a fraud vector. The same agency turns around and invoices its own clients on a model that has no such clause, because there is no counterparty in that direction with the leverage to demand one. The asymmetry is its own confession.

The cultural cost

"I'm out of hours" has become a collaboration killer.

People won't help across projects because their meter doesn't tick. Two-week retros pivot on whether somebody can absorb an unplanned hour without their utilization slipping. A senior engineer who notices a problem in another team's code will sometimes pretend not to, because flagging it means owning it, and owning it means a billing conversation with two practice leads who will both lose.

The people doing these things are not bad. They are not lazy. They are not gaming. They are responding rationally to a structure that grades them on a number, and the number rewards a specific kind of behavior, and the behavior is corrosive to the company that depends on them not behaving that way.

This is a structural problem that represents itself as an individual behavior problem. The structure makes good people look like they're lying. They aren't. They're answering the question the structure asked them, which is how do I make my line on the spreadsheet add up — and the only way to make it add up under T&M with AI in the toolchain is to bend something.

The leadership response is to manage the symptom. Talk about culture. Send a memo. Run a workshop on "billing integrity." The memo is well-meant and the workshop is well-attended and the structure is unchanged the next morning, and the number is still wrong, and the smoothing continues. You can't culture-train your way out of a meter that's measuring the wrong thing.

"Clients won't accept anything else"

The reflexive objection comes up in every conversation about getting off T&M. It is almost always wrong, but not for the reason most people offering the counter-argument think.

Clients already accept everything else. Day rates. Retainers. Fixed-fee. Milestone billing. Capacity blocks. Anyone selling capacity sells these. Large-enterprise procurement, on one extreme, audits every hour. Two-person studios on the other end need hour-by-hour visibility. Almost everything between those two extremes is already negotiable, and most agencies have some contracts that already work this way. The friction isn't client appetite. It's agency muscle memory.

Here is the harder version of the objection, which I heard in a leadership room earlier this spring: if you sell capacity instead of hours, you'd better make sure you're delivering enough.

That isn't a complaint about the model. It's a real constraint on it. The hourly meter, for all its problems, is a continuous accountability mechanism — the client knows what they're paying for in fifteen-minute increments. Switch to allocation billing and you lose that. The client is now buying a sprint, and if the sprint underdelivers, the client has paid for less than they thought they'd get.

The answer to that objection is not to wave it off. It's to acknowledge that switching off T&M raises the bar on what you have to prove every two weeks. The accountability mechanism moves from the timesheet to the increment. You demonstrate value by shipping things, on a cadence, that the client can see and evaluate. That is harder than running a meter. It is also the entire point.

And no, you don't escape by productizing

Y Combinator's Spring 2026 RFS asked for AI-powered agencies with SaaS-like margins.

The framing is half-right. The delivery model is broken. Hours are the wrong unit. AI changes the labor math in ways that haven't fully landed in agency P&Ls. All of that is true and the RFS deserves credit for naming it.

The naive read is wrong. Strip the humans, keep the meter, print SaaS margins. That misidentifies the bug.

In services, value and commoditizability are inversely correlated. The work clients pay agency rates for is judgment under ambiguity and accountability for outcomes — neither of which survives a price card. The moment you can put a clean number next to "build a checkout flow," the platform you'd build it on has already built a better one and given it away. The work that is price-cardable has either already been compressed to commodity prices by the SaaS that owns the category, or will be.

Productized services is the attempt to sell the commoditizable slice at non-commoditized prices. It is a brief arbitrage, not a business model. It misidentifies the bug — the bug isn't "humans are in the loop"; the bug is "the meter measures the wrong thing." SaaS margins come from distribution economics, not from removing humans from delivery. And if you go productized, you've signed up to compete with funded SaaS on their home turf, which is not a fight your services P&L was designed to win.

What replaces it

I am inside the change while making the argument for it. That is the most honest version of this essay I can write, and it is probably the only version worth writing.

The unit of sale becomes capacity against a prioritized backlog over a fixed window. Hours become an internal velocity instrument — useful for steering, never the meter on the wall. Scope flexes inside the window. The window does not. The client is buying a guarantee of delivery on a date, with a known team, and a backlog that they help prioritize. If the work goes faster, they get more of the backlog. If it goes slower, the must-haves still land and the could-haves slide. The meter on the wall stops being hours; it becomes increments shipped.

The operating-model specifics — sprint cadences, capacity caps, riskshare structures, exchange-for-free swaps — are workable and well-documented in the agile-contracts literature. They are not the hard part. The hard part is what comes next.

The hard part isn't the meter, it's the accounting

Utilization. Gross margin. Revenue recognition. Individual contributor compensation. Weekly health metrics. Quarterly forecasting. All of it currently hangs off the billable hour.

Most "we moved off T&M" announcements quietly revert. I've watched a half-dozen of them in the last decade. They revert because the leadership team changed the meter on the wall and didn't change the meters underneath the wall, and the underneath-meters are what determine whether anyone in the company believes the change is real.

Your sales team is still grading itself on bookings denominated in hours. Your delivery leads are still grading themselves on utilization percentages computed from hours. Your CFO is still recognizing revenue against hours billed. Your compensation system still pays people partly on their personal utilization. Three months in, somebody asks why margin looks weird, and somebody else says let's just track hours alongside the new model so we can compare, and four months in the new model is theater and the meter is back.

Each of those underneath-meters is its own load-bearing migration. Utilization has to be redefined against allocations rather than billed hours, which means the denominator changes and so does every dashboard built on top of it — and the people who built those dashboards are not the people changing the pricing model. Revenue recognition rules under ASC 606 are workable for fixed-window allocation billing but the workable answer is not the answer your finance team is using today, and changing it touches audit. Compensation tied to personal utilization has to be unwound on a quarterly cadence that doesn't disrupt people mid-cycle, which means the comp change lags the pricing change by at least two quarters, which means for two quarters your delivery org is being paid on the old meter while being asked to operate on the new one. Sales commissions denominated in bookings-hours have to be reconverted to bookings-capacity, and the salespeople notice immediately. Forecasting models built on a smooth hours-billed signal have to be rebuilt against the lumpier sprint-allocation signal, which the CFO will not love. None of these are pricing decisions. All of them have to land before the pricing decision is real.

The pricing change is the easy part. The accounting change — utilization redefined, comp restructured, revenue recognition rewritten, weekly rhythm rebuilt, forecasting reframed, dashboards re-instrumented — is the part that determines whether anyone in the company actually changes how they work. Most agencies that announce they're getting off T&M are announcing the pricing change without the accounting change. They are announcing pricing theater and they don't know it yet.

The agencies who win the next five years

The meter is the bug. Unbugging it does not save you from the squeeze.

AI is compressing what clients are willing to pay for delivery, and that compression is coming for the entire services category whether or not you change how you bill. Customers know AI makes the work faster. They are not going to keep paying old prices for new speeds. The compression lands on you either way.

The difference — and this is the difference — is who's holding the pen when it lands. The agency that fixes the meter gets to design the compression on its own terms: allocations priced, scope flexed, margin known, AI cost-of-revenue priced into the line, the operating model rebuilt to make all of it legible. The agency that doesn't fix the meter gets the same compression done to it on its customers' terms — sliced off the topline as billed hours quietly fall, with the AI cost line eating margin from the other side, with senior engineers smoothing their timesheets to make the numbers look right, with the structure routing the failure through individual people until the people leave.

The services business that wins the next five years is not the one that bills the most hours. It is the one whose pricing model stops paying people to be slow, and whose accounting model stops grading them on whether they were.

Fix the meter. Then fix the meters underneath the meter. Then prove, two weeks at a time, that the increment is worth what you charged for it.

That is the work. That is the only work.