M&A in the Brand Era: How to Value and Sell a Three-Person, Seven-Agent Company

The DTC exit playbook your founder peers are still running is the 2018 playbook. Build to $50M or more in revenue, hire a 40-person team, sell the team and the brand together to a strategic acquirer at three to five times revenue. That playbook still works. It just is not the highest-multiple playbook anymore.

The 2028 DTC exit looks different. A $15M-revenue brand with three humans and seven agents, run on the 3-7 cost structure to a modeled 25% EBITDA margin, a Recommendation Loop instrumented and producing measurable memory polarity, a Machine-Readable Voice fingerprint that survives an ownership transfer, and a clean agent-commerce surface (the 12-Field Agent SKU, the 8-Signal Trust Stack, the Agent Return Triangle) that an acquirer can scale on day one.

The Agent-Adjusted Multiple is the valuation framework for that company. It lowers the discount the market applies to small headcount, raises the premium for measurable memory polarity, and treats agent-stack continuity as the central deal-side risk. This post is the framework, the comparables, the term-sheet implications, and the explicit 2028 prediction. It is the tenth and closing post of a ten-post cluster on Agent Commerce and the Brand Era; the framework lands here because everything else in the series feeds it.

The buyer is no longer paying for the team. They are paying for the loop.

Why the 2018 exit playbook stops working

The 2018 exit priced a team. Headcount was a proxy for stability and capacity, so a 40-person marketing org read as a durable, transferable asset and a tiny team read as key-person risk. When a brand runs on three humans and seven agents, that logic inverts. The acquirer is not buying a team that could leave; it is buying a loop that keeps running. Lower headcount stops being a discount and becomes a positive signal, because it means the brand's growth is embedded in instrumented systems rather than in people who walk out the door at close.

The reason this is a now-problem rather than a someday-problem is the scale of the shift underneath it. Gartner projects agent-mediated purchasing will represent over $15 trillion in B2B spending by 2028, and consumer DTC is on the same curve. A brand that built itself as a loop rather than a team is the asset class that curve produces, and it is valued against the 3-7 P&L the multiple is calculated against. The cost structure there is real and sourced; the 25% EBITDA margin in this post is a modeled target for a well-run version of it, not a claim about the sourced composite, which runs nearer breakeven.

Introducing the Agent-Adjusted Multiple

Start from the baseline and adjust. A profitable, growing mid-market DTC brand trades in a range of roughly three to five times revenue (industry composite, mid-market DTC, 2024 to 2026; not a quoted deal). The Agent-Adjusted Multiple applies four adjustments to that baseline. The headcount discount that used to subtract from a small team is removed, and at the clean end turns slightly positive. A memory-polarity premium is added, the single biggest lever, based on the measured quality of the brand's agent recall. An agent-stack continuity risk is subtracted, based on how locked-in the brand's agent vendors are. And a brand-voice transferability premium is added, based on how consistent and portable the Machine-Readable Voice fingerprint is.

Net it out and a clean 3-7 P&L brand with strong loop signals can reach five to seven times revenue, against three to four for a comparable headcount-heavy brand with weak loop instrumentation (all ranges industry composite, not quoted deals). Real DTC M&A history is the reference class here: the public trajectories of Allbirds, Warby Parker, FIGS, Yeti, Honest Company, and Solo Brands, and the PE activity of firms like Apollo, Cerberus, and KKR, are where these multiples get set, and Pitchbook and CB Insights are where the composites get summarized. The market backdrop that makes the premium real is the adoption curve: ChatGPT alone reached 900 million weekly active users by February 2026, which is the demand the loop monetizes and the acquirer is buying access to.

What acquirers actually buy: the four assets that transfer

A brand-era acquisition transfers four assets, and they do not transfer equally. Brand IP (the trademark, the voice, the content corpus) transfers cleanly, the way it always has. The agent stack configuration and training data transfers only partially, and only cleanly if the agent vendors are non-exclusive and the data is portable. The Recommendation Loop instrumentation transfers conditionally: it keeps working only if the buyer maintains the brand-voice signal that feeds it, which is why Machine-Readable Voice and its transferability through ownership change is a deal variable, not a branding footnote.

The fourth asset, customer memory continuity, is the one acquirers underbid because they do not yet know how to value it. Some of the agent's memory of the brand is brand-side and transfers with the company; some is consumer-side, attached to the buyer's own agent, and does not transfer at all. Pricing that split is the frontier of brand-era diligence. The reason it matters more every quarter is the volume riding on it: Adobe reported AI-driven traffic to U.S. retail sites grew 393% year over year in Q1 2026, so the memory layer the acquirer is buying governs a fast-growing share of the brand's future demand.

Deal-side risk: agent vendor lock-in

The biggest new diligence item in a brand-era deal is agent vendor lock-in. The acquirer has to answer three questions: which agent vendors does the company run on, are those vendors substitutable, and what happens to the memory layer if a vendor is switched. A brand running all seven agents on a single closed vendor carries a continuity discount, because the buyer inherits a dependency it cannot easily unwind and a memory layer that may not survive a migration. A brand running its agents on open protocols carries a vendor-neutrality premium, because the buyer can swap infrastructure without losing the loop.

This is where the protocol map that defines vendor-neutrality becomes a balance-sheet question: a stack built on ACP, AP2, and the open commerce protocols is portable in a way a single-vendor stack is not, and portability is what the continuity-risk adjustment prices. The stakes are not small, because the demand flowing through these agents is large and growing: Adobe's 393% year-over-year jump in AI-driven retail traffic is the traffic a locked-in memory layer puts at risk in a migration. Paid amplification compounds on a portable, trusted stack, which is part of why acquirers value a brand that can keep running OpenAI Ads through an ownership change.

Term sheet implications: what changes vs a 2018 DTC deal

A brand-era term sheet adds five clauses a 2018 DTC deal never had. Agent-vendor continuity clauses lock in the agent infrastructure for a 12-to-24-month minimum so the loop does not break at close. Brand-voice transition periods of 90 to 180 days commit the seller to maintaining Machine-Readable Voice consistency while the buyer learns it. Memory-polarity earnouts vest on agent-recall metrics measured after close, aligning the seller to the asset the buyer is actually paying for. Trust-layer maintenance commitments keep the Schema.org Organization markup, sameAs links, and registry profiles current through the transition, which is the trust-layer maintenance commitment written into the deal. And Recommendation Loop access gives the incoming acquirer day-one access to the instrumented dashboards.

Two of those five (the memory-polarity earnout and the day-one loop access) are the ones most acquirers have not yet put in a term sheet, and they are the ones founders should negotiate hardest, because they protect and price the loop the buyer is paying for. The reason these clauses have teeth is that agent-routed demand is already material: ChatGPT alone drives a fifth of some major retailers' referral traffic, so the loop these clauses protect is governing real revenue at close, not a future option.

The 2028 prediction: when brand-era multiples become the median

Here is the dated bet, stated as a prediction. Brand-era exits are a small minority of mid-market DTC M&A today, a modeled mid-teens to twenty percent share in 2026. The prediction is that by Q2 2028, half of all mid-market DTC exits in the $10M to $50M revenue range will be 3-7 P&L brand-era companies, and the median deal will be priced with the Agent-Adjusted Multiple rather than the headcount-era one. The exact quarter will be wrong; the direction is the durable part. What is not a prediction is the destination: Bain projects agentic AI will account for 25% of U.S. ecommerce sales by 2030, and an acquirer cannot ignore a quarter of the market's demand mechanism when it prices an asset.

The operator consequence is simple and it is the same logic as the rest of this series: a brand built to the framework in 2026, while brand-era exits are still a minority, is inside the median when the median moves in 2028; a brand that was not is below it. The framework is not a forecast you wait on. It is a structure you build toward now, so that when the valuation method changes, your company is already the kind the new method rewards.

The buyer is no longer paying for the team; they are paying for the loop, and the Agent-Adjusted Multiple is how that loop gets priced. This is the tenth and closing post of the cluster, and the framework lands here because every earlier post is an input to it. An acquirer valuing a brand-era company is really valuing the whole stack the series described: the 7-Stage Agent Checkout the brand transacts on, the Recommendation Loop it runs, the 12-Field Agent SKU and the 8-Signal Trust Stack that get it surfaced and verified, the Agent Return Triangle that protects its memory, the Machine-Readable Voice that survives the transfer, the 3-7 P&L that produces the margin, the Operator-CMO Profile that directs it, the B2B Mirror that shows where it is all heading, and the Agent-Adjusted Multiple that finally puts a number on it. Ten frameworks, one operating model. By 2028 that model is the default the median DTC exit is priced against, predicted here and certain well before the decade is out. Cresva is the operating layer that produces the loop the acquirer pays for: forecasting, attribution, creative, execution, memory. Explore OpenAI Ads for the channel side, and request early access to see your own brand modeled against the Agent-Adjusted Multiple.