The Brand Era P&L: What a $15M DTC Looks Like With Three Humans and Seven Agents

Open your last quarterly P&L. Find the line your accountant labels marketing personnel, or growth team, or sales and marketing salaries. For a mid-market DTC brand that line is one of the largest controllable costs in the operating model, and it has historically been the one line every founder was told not to touch, because the team running marketing was the team running growth.

The brand era changes that arithmetic. When seven AI agents handle forecasting, attribution, creative strategy, campaign execution, and post-purchase memory, the marketing and growth headcount that line pays for shrinks to a three-person team. In the model below, that single swap moves about $1.1 million of operating cost on a $15 million brand. All else equal, that is roughly seven points of operating margin.

This post is a composite P&L built from the public 10-K filings of three real DTC brands (Allbirds, Warby Parker, and FIGS), reshaped to model the brand-era org. The filing figures are real and sourced to SEC EDGAR. The org-cost figures (headcount, loaded comp, agent-stack cost) are modeled assumptions, labeled as such wherever they appear. The 3-7 P&L is the operating-model template for a $10M to $25M DTC brand in the brand era: three humans (an operator-CEO, a brand director, a head of memory) plus seven AI agents that run forecasting, attribution, creative, execution, and memory.

The salary line was always the lever. The brand era is the first time anyone could pull it.

Why nobody has shown you this P&L yet

The reason this P&L is hard to find is structural. Private DTC P&Ls are not public, and public DTC P&Ls are not structured for the comparison this post needs. So the honest way to model a brand-era P&L is to build a composite from companies that do file publicly, are genuinely DTC, are marketing-heavy, and have recent filings. Allbirds, Warby Parker, and FIGS fit all four. Their numbers are real, line by line, in 10-K filings anyone can read on SEC EDGAR.

These are not chosen to flatter the model. They span a wide margin range, which is the point: a median across three smooths out any single company's noise. FIGS, the most profitable of the three, reported marketing expense at 14.8% of net revenue in its most recent 10-K; the other two land near that on the marketing line even as their bottom lines diverge sharply. This composite is the baseline. The brand-era restructuring is what we apply on top of it. The funnel-to-loop shift underneath all of this is covered in the funnel autopsy and the Recommendation Loop that replaces it; this post is the P&L that shift produces.

The composite baseline: a real $15M DTC drawn from three filings

Take each brand's most recent reported full year, express every line as a percentage of revenue, take the median across the three, and apply it to a $15 million baseline. The result is the composite below. The gross margin is healthy (a 54% median, anchored by Warby Parker's 54% gross margin in its most recent 10-K). The operating line is honest: the composite runs near breakeven, because two of the three brands are still spending ahead of profitability. That is the real state of public DTC, and it is the right baseline to reshape rather than a flattering invented one.

Introducing the 3-7 P&L: same revenue, restructured org

Now hold revenue and gross margin constant and restructure the org. This is the 3-7 P&L: the same $15 million top line, the same product economics, a different operating team. Three line categories move. The marketing media spend stays roughly constant as a share of revenue but reallocates toward agent-readable surfaces, which is why brands running this model push budget into OpenAI Ads and agent-led optimization rather than legacy display. The catalog work that makes those surfaces pay off is the 12-field Agent SKU that gets the catalog ranked. The two lines that move materially are headcount and the agent stack.

The marketing and growth personnel line, modeled at $1.8 million for a twelve-person team, collapses to about $480,000 for the three humans the brand era org needs. A new line appears: the agent stack, modeled at $180,000 for seven agents (software plus operations). The net operating-cost reduction is about $1.1 million. The reason to restructure now rather than wait is the direction of demand: Gartner projects 90% of B2B purchases will be mediated by AI agents by 2028, representing over $15 trillion in spending, and consumer DTC is moving the same way faster. To be precise about what that figure is and is not: it is a cost-structure delta, not a promise that EBITDA turns positive, because the composite baseline is near breakeven. On a $15 million brand, removing $1.1 million of operating cost is roughly seven points of operating margin, all else equal, whichever side of zero the brand starts on.

Where the cost saving actually goes

The roughly $1.1 million the swap frees up is not free money, and a model that treats it as a clean drop to the bottom line is the model a CFO stops trusting. The honest version reinvests most of it. A modeled allocation: about $400,000 into agent-stack maturation (memory infrastructure, integrations, the unglamorous plumbing that keeps the seven agents reliable); about $300,000 into brand, because creative quality goes up rather than down when humans stop doing execution and focus on craft; about $200,000 into supply chain and margin work; and about $200,000 to the bottom line. The proportions are a starting point, not a prescription.

The reason to reinvest rather than pocket the saving is that the brand-era advantage compounds, and the compounding is the asset. The demand moving to agent surfaces is not a niche to harvest once: Bain projects agentic AI will account for 25% of U.S. ecommerce sales by 2030. A brand that reinvests the saving into being the agent's trusted recommendation widens a lead that a brand pocketing the cash does not. The mechanics of how an agent reaches the buy step in the first place are in the frame-by-frame walkthrough of how an agent actually buys.

Three humans: the JD for each role

The three humans are not a shrunken version of the old team. They are three distinct mandates, two of which barely existed in 2022. The operator-CEO owns the P&L and the agent oversight at the meta level. The brand director owns everything that becomes the agent's training data: voice, visual system, customer experience. The head of memory, the genuinely new role, owns the post-purchase loop that decides what the agent remembers about the brand. That role exists because agent-routed demand is now material enough to staff against: ChatGPT alone drives 20% of Walmart's referral traffic, and the memory written after each of those purchases shapes the next recommendation. What the agent reads to form that memory is the 7 things that make an AI agent recommend your brand.

What changes in the next board pack

The 3-7 P&L changes the board deck as much as the org chart. Half the slides a 2024 DTC board reviews measure behaviors the agent buyer does not perform. The volume forcing the change is not speculative: Adobe reported AI-driven traffic to U.S. retail sites grew 393% year over year in Q1 2026. A board reviewing only click-funnel metrics is reviewing a shrinking slice of how revenue actually arrives.

The brands that reach the 3-7 P&L by 2028 will out-earn the brands still carrying a twelve-person marketing org by enough operating margin to fund the next product line, on the same revenue. That is the whole case in one sentence: the cost structure that funds growth in the brand era is lighter on people and heavier on the agent layer that the people now direct. The composite here is real, the restructuring is modeled and labeled, and the direction is not in doubt. Cresva runs the seven-agent layer that makes the 3-7 P&L possible: forecasting, attribution, creative, execution, memory. Request early access to see your own brand's 3-7 P&L modeled from your actuals.