Six Layers of Sediment: The GTM Stack Nobody Designed
Go-to-market tooling has turned over five times since 2010, and five of the six eras added a layer while retiring nothing, because each was an answer to…
Go-to-market tooling has turned over five times since 2010, at intervals close enough to three years that the industry calls each turn a generation. Generations replace each other. These did not. Sales engagement did not retire the CRM. Revenue intelligence did not retire sales engagement. Intent data did not retire enrichment, and workflow automation did not retire any of it. Six eras produced six strata, all still running, all still invoiced, all still sitting on somebody's renewal calendar. What a revenue organization operates in 2026 is not a stack. It is sediment. The claim is testable in one sentence, so test it. Name a go-to-market tool category that a later era actually retired. Two candidates survive the first pass. The standalone dialer stopped being a purchase once sales engagement platforms shipped calling inside the sequence. Salesforce retired Data.com, the contact database it built out of its Jigsaw acquisition. Neither was a stratum. A dialer was a feature that had been priced as a product, and Data.com was a table that a better-funded table replaced. Neither is an era retiring the layer beneath it, which is what the generational story promises. Marketing automation is still bought. CRM is still bought. Sequencers are still bought, including by the companies that later bought the intelligence layer sitting on top of them. Features get absorbed and vendors get outbid. Strata do not get removed. The timeline itself is borrowed. A version of it circulates in go-to-market writing and did not originate here. The boundaries were blurry and several of these vendors moved between eras. The cadence and the additive habit are the evidence. The vendor names are not. Compressed, it runs like this. From 2010 to 2014, CRM and marketing automation: Salesforce and Marketo. From 2015 to 2016, sales engagement: Outreach and Salesloft. From 2017 to 2019, the layer that read what the sequencer produced: Gong on the conversation, Clari on the forecast. From 2020 to 2022, intent and account-based targeting went mainstream: Bombora and 6sense. From 2023 to 2024, workflow automation: Clay and n8n. From 2025 into this year, agentic go-to-market, with agents running the workflows and callable interfaces displacing the dashboards built to operate them. Six eras, and all six are still being paid for. A CRM instance bought early in that first era still sits underneath everything, carrying custom fields nobody can name. A sequencer is wired to it and writes activity back, feeding an intelligence tool that reads that activity, cross-referenced against an intent feed. An orchestration layer stitches the whole thing together because the other four do not talk to each other. Nothing was deprecated. Contracts renewed on their own schedules, each renewal individually defensible, and the layers compacted. Stack is the wrong word for the result. A stack implies somebody designed it. The additive habit is not an accident of procurement discipline. Each era was a response to the waste product of the one before it. Sales engagement existed because the CRM recorded what happened and did nothing about it. Revenue intelligence existed because sequences produced volume without judgment, and somebody had to say which of ten thousand touches was working. Intent data existed because enrichment told you who to contact and never told you when. Workflow automation existed because signals began arriving faster than a team could act on them. Agentic go-to-market exists because orchestration produced more workflows than any team can operate. Read that chain a second time and the additive habit explains itself. A layer that solves the exhaust of the layer beneath it cannot replace that layer, because it consumes the exhaust. The intelligence tool has nothing to score without the activity the sequencer writes. The orchestration layer has nothing to orchestrate without the signals the intent feed emits. No era arrived as a successor. Every one arrived as a dependent. The job titles deposited on the same cadence and for the same reason, which is the habit showing up on the org chart instead of the invoice. The current era looks like the seventh entry on the list and is not. Every prior era shipped a category: a thing to buy, a line item, an admin to hire, a seat count to negotiate. This one ships a removal. The dashboard, the queue view, the sequence builder, the pipeline board, the report a director opens on Monday morning: those are the parts an agent has no use for, because work now reaches the system through a callable surface instead of through a screen somebody looks at. For the first time the thing being retired is not a vendor category. It is the layer everyone had assumed was the product. Additions and subtractions do different things to an organization, and the difference is not one of size. An addition changes what a team can do. The team is the same team, with one more surface to operate and one more line on the requisition. A subtraction changes what a team is. Remove the interface and the person who spent four years becoming fluent in a specific screen holds fluency in something that is no longer how the work happens. The operator stops being a person clicking. It becomes a system calling. The org chart changes. The tooling budget may not move at all. The mistake is easy to spot and it is already running. A company signs an agent platform, points it at the six strata underneath, and within two quarters the agent is opening sequencer records, writing activity back into the CRM, and reading the intent feed on a schedule. That is a description of the job a person held. The stack did not shrink. It acquired a faster operator. The signature was easy to get. One more platform on a stack that already has six, approved the way every prior addition was approved. The workflows run faster. The renewals all clear. What was bought is speed applied to a shape that was deposited, never designed. The decision worth making this year is not which agent platform to add. It is which of the six strata is now only a database charging a fee for a screen nobody opens. That question does not require an outside opinion to answer. Take each vendor in the stack, name the workflow a human still performs inside it by hand, and ask whether that workflow has become a call an agent makes. The interface has been declared dead before. Conversational selling was going to move the pipeline review into a chat window. Voice was going to turn the CRM into something a person asked instead of opened. Both waves ended by shipping more screens, because somebody still had to see what the system had done. The mechanism in this note predicts the same outcome here. If every era answers the exhaust of the one before it, the exhaust of agentic go-to-market is actions nobody watched, and the layer that answers it is a review queue and a run log. Those are screens. What leaves is the screen a person operates the work through. What comes back is a screen a person supervises the work from. The question is not whether a dashboard exists. It is how many people need one. We take execution authority inside a client's organization for 8 to 16 weeks, and reading the stack the way this note describes is part of that work, not a separate deliverable: what each layer is actually doing, which workflows a person still runs by hand, and what is left of a vendor once the work stops going through it. The deliverable is a system operating inside the business rather than a recommendation about which platform to buy.