Block Is Rebuilding Around an Intelligence Layer. What That Means for a Sales Floor

Block published From Hierarchy to Intelligence. Its argument: layers exist to route information, and AI now does the routing. Block is a payments company with a transaction graph. Revenue teams have messier data, so here is what that argument looks like on a sales floor.

GTM is the most hierarchy-bound function in the modern company. SDR to AE to manager to director to VP to CRO. Marketer to manager to director to VP to CMO. Two pyramids that meet at the executive table and argue about pipeline.

That structure was built to move information through humans, because humans were the only routing layer available. There is now another way. This is my attempt to write the GTM version of that essay.

The revenue pyramid was a workaround, and the workaround is over

Walk into any B2B revenue org and you will find the same shape. A wide base of SDRs and BDRs. A layer of AEs above them. Managers above that. Directors. VPs. A CRO at the top. Marketing mirrors it exactly.

Every box on that chart exists for one of three reasons: to do work the system above it could not do, to route information between the box below and the box above, or to make a decision the box below was not trusted to make.

In 2005 those were real reasons. The SDR existed because nobody else had time to dial. The manager existed because the VP could not coach 40 reps. The director existed because the VP needed someone to translate strategy into territory plans. The campaign manager existed because the CMO could not personally check every nurture flow. Every layer was a workaround for the same constraint: humans cannot hold enough context, route enough information, or make enough decisions on their own.

The buyer no longer needs the SDR to start a conversation. They start it in ChatGPT. The AE no longer holds context the buyer cannot get elsewhere. The buyer has read the G2 reviews, the Reddit threads, the pricing page, the earnings call, and three competitor comparisons before they take a meeting. The manager no longer needs to relay forecast information up the chain. The system already knows. The director no longer needs to compile the QBR deck. An agent builds it from source data in 90 seconds.

Gartner estimates that 75 percent of B2B sales organizations will augment traditional playbooks with AI-guided selling by 2026, and that by 2028 a third of enterprise software will include agentic AI capable of autonomously making at least 15 percent of day-to-day work decisions. Forrester's buyer research shows generative AI has passed vendor websites and sales reps as the most influential information source in many segments. McKinsey estimated in June 2023 that gen AI could raise marketing productivity by 5 to 15 percent of total marketing spend and sales productivity by about 3 to 5 percent of sales spend. The figure is a projection.

Translate that to the org chart. The roles whose primary job was to move information through the system are the ones being absorbed by the system. The roles whose primary job is judgment, trust, and edge cases are gaining reach.

The pyramid was never the goal. It was the price of not having a better option. We finally have a better option.

The operating intelligence model for revenue

Block's insight was that you can replace the coordination job of hierarchy with a system that holds a continuously updated model of the business. The same shape works for revenue, and arguably works better, because every buyer step leaves a signal. Every visit, every call, every email open, every product touch, every renewal, every churn, every expansion. That is the revenue equivalent of a transaction graph. Most companies already have the signal. Almost none of them use it as a coordination layer.

Layer 1: Revenue capabilities

The atomic primitives of how revenue gets created. Capabilities: lead qualification, account research, buyer education, demo delivery, pricing exploration, objection handling, procurement support, onboarding, adoption, renewal, expansion, reference creation. Each one gets an owner, a quality bar, a latency target, and a definition of done. Some run on humans. Some on agents. Most on both.

Layer 2: The revenue world model

Two sides. A company-side model that knows what every account is doing, what every rep is working on, what every campaign is producing, where every forecast is soft, and where every renewal is at risk. A buyer-side model that knows who the account is, what they have read, what they have asked, who else is involved, what they have objected to, and where they are in their actual decision process. It is the system of record for what is true right now, and it updates continuously from every interaction.

Layer 3: The intelligence layer

This is the part most companies are missing entirely. The intelligence layer composes capabilities into the right action for this buyer at this moment based on what the world model knows. A buyer watches the pricing video for the third time this week, and the system composes a personalized ROI summary, a calendar slot with the right AE, and an alert to the deal team with context attached. Three users go inactive on a renewal account, and the system composes a usage analysis, a CSM outreach, and a budget signal for the AE working the expansion. Nobody assigned it. Nobody routed it. The system recognized the moment and composed the response.

Layer 4: Revenue interfaces

The website. The product. The email. The deal room. The QBR. The renewal conversation. These are delivery surfaces. The value lives in the model and the intelligence layer. The interfaces deliver it to the buyer, or the seller, at the right moment.

Now invert the chart: intelligence in the system, people on the edge

The edge is where revenue actually happens. The first call with a strategic prospect. The pricing negotiation that needs a human read on the room. The QBR with a customer whose CFO just changed. The board-level reference call. The renewal conversation with a champion who is leaving. The competitive bake-off where trust decides the deal. Those are the moments where humans are irreplaceable, and they are the moments your best people should get all of their week.

The middle layer goes. The people move to new roles. The information-routing manager. The status-meeting coordinator. The deal-desk-as-bottleneck. The campaign approver. The pipeline recompiler. Those jobs were workarounds.

In place of the middle, three roles:

• Revenue ICs. Sellers, marketers, CSMs, RevOps engineers. Deep specialists in one capability or one segment. The world model gives them the context a manager used to provide. They act on the signal when it arrives.
• Directly Responsible Individuals. Own a cross-cutting outcome: reduce mid-market churn in 90 days, lift enterprise win rate in financial services, compress cycle time in manufacturing. They can pull resources across capabilities. They may persist or rotate by quarter.
• Player-coaches. Senior operators who still do the work and develop the people. They replace the managers whose primary job was to compile and forward information. The system does the compiling. The coach does the coaching.

BCG's research on agentic AI suggests this kind of redesign can accelerate core business processes by 30 to 50 percent in the workflows where it lands cleanly. The productivity story is real, but it is the second-order story. The first-order story is that this technology rewards a different org chart from the one you inherited.

The playbook: five questions for your next exec staff meeting

You cannot ship this in a quarter. You can start it in a week. These five questions are designed to surface the gap between the org you have and the org your buyer is already operating against.

1. Which roles exist primarily to route information? Walk every layer and mark the roles whose main output is a status update, a rolled-up report, a forwarded request, or a meeting summary. That is your compression candidate list. The work stays. The role goes.
2. What capabilities does our revenue motion actually need? Write the capability list, then map every current role to the capabilities they deliver. You will find capabilities owned by no one and roles delivering capabilities the buyer no longer values.
3. Where does our world model live, and who owns it? Data scattered across CRM, marketing automation, product analytics, support, billing, and 14 spreadsheets is a pile. If no one owns it, no AI investment downstream will compound.
4. Which moments should never require a human, and which should never not? Education, qualification, scheduling, basic objection handling, pricing exploration, and content delivery belong to the intelligence layer. Trust, negotiation, executive alignment, complex procurement, escalation, and references belong to the human edge. Draw the line on purpose, before someone else draws it for you.
5. If we were starting this revenue org today, what would we not rebuild? This is the most uncomfortable question in the room, and the honest answer almost always includes layers and roles that are sacred today. That discomfort is the signal.

So here is the exercise for your next QBR. Pull up the org chart. Cover the names. Ask the room: if the system held the intelligence, which boxes would still need to exist? If the room goes quiet, you have found the most useful conversation of the year.

Redesign how revenue gets coordinated

Treat this as a redesign of how revenue gets coordinated. The pyramid is done. Companies that build the operating intelligence first compound on every cycle: better signal feeds a better world model, a better world model composes better action, better action creates better outcomes, better outcomes create more signal. Hierarchy cannot keep up with that loop. Nothing built around human routing can.

The CRO and CMO who get this right will be operating an intelligence in 2027. The work will look like tuning a system. Fewer layers. Cleaner signal. More time on the edge where revenue actually gets made.