Golden Data

Issue #9 · July 5, 2026

The SDR job got sold in pieces.

Paul Brown Writes Golden Data from Ogden, Utah 6 min read

Every Sunday I look at the job function a vendor used to sell a productivity tool for. This Sunday it is the sales rep — specifically the SDR, the first rung on the enterprise-sales ladder. For fifteen years the CRM sold a seat license against the rep’s calendar. Faster research, faster outreach, faster follow-up. The rep’s day was the unit. The seat was the invoice.

This fortnight four vendors quietly moved. They bought or shipped the parts of the SDR’s day — the research, the outreach, the inbound triage, the named-contact list — and reset the pricing surface. The rep’s calendar stopped being the constraint. The rep’s tasks got a price sheet.

The buyer noticed. The buyer that used to add three SDR seats to hit next quarter’s coverage now buys a Prospecting Agent bill and a per-lead unit price. The rep’s day got a bill of materials.

What’s actually shipping this week

1. HubSpot agreed to acquire Warmly on June 30, 2026. Terms undisclosed. Warmly, founded six years ago, is best known for person-level website intent that identifies more than half of the visitors who never fill out a form, plus two agents that turn that intent into outreach — the Inbound Agent and the TAM Agent. HubSpot’s own framing: this closes the CRM’s biggest blind spot, the 95%-plus of visitors who leave without identifying themselves. Existing Warmly contracts and pricing stay in place. The CRM vendor didn’t build the SDR-substitute. The CRM vendor bought it.

2. Salesforce signed a definitive agreement to acquire Fin (formerly Intercom) for approximately $3.6 billion on June 15, 2026. Closing expected Q4 FY27. Fin’s AI Agent resolves customer queries end-to-end across live chat, email, WhatsApp, SMS, phone, and Slack — powered by Apex, Fin’s proprietary customer-support model, which Salesforce’s own release says outperforms top commercially available frontier models on resolution rate. Salesforce’s Q1 FY27 print already had Agentforce at $1.2 billion ARR, up 205% year-over-year. Salesforce paid $3.6 billion for the agent that answers the customer before the SDR sees the ticket.

3. Salesforce Summer ‘26 shipped Momentum and a Customer Engagement Agent on June 15, 2026. Momentum — acquired February 18 — captures every call, email, and meeting from Zoom and Google Meet and writes structured data back to Salesforce in real time, with no change to how reps work. The Customer Engagement Agent qualifies inbound leads autonomously, twenty-four hours a day. The Summer ‘26 State of Sales Report: 55% of sales professionals use AI for prospecting today, another 38% plan to; top-performing sellers are 1.7 times more likely to use prospecting AI than underperformers. The rep does not have to change how they work. The rep also does not get to charge for the work the agent now does.

4. 6sense shipped AI-Recommended Leads on June 22, 2026. The GTM-intelligence vendor turned the engaged-account list — the SDR’s morning research queue for a decade — into a downloadable set of named, CRM-ready contacts per qualified account. The campaign that used to end with an account name and “check LinkedIn” now ends with the specific person, whether or not that person clicked the ad. Follow-up rates go up. Sales-marketing friction goes down. The prospect list stopped being research the SDR did. The prospect list became an output the vendor sells.

What I’d ship in your app this week

The rep’s day became a checkout page. Two small tools your team can ship in two weeks so the AI feature in your product can be sold against the outcome, not against the seat, before the CRM vendor prices the outcome for you.

Feature one: the outcome-priced tier next to your seat-license tier. Pick one AI feature you already ship. Publish an alternative pricing tier billed per successful outcome — per resolved ticket, per generated lead, per completed workflow. Existing customers stay on their seat license; five pilot customers switch to the outcome tier. Read the results at ninety days. The tier customers migrate to is the one the market is telling you your feature is actually worth.

  • Shape. One new SKU on the billing page. One meter that increments on the outcome event. Existing seat-license unchanged.
  • Data shape. Per (tenant_id, feature_id, outcome_event_id, tier, unit_price, invoice_line). Joins to the existing invoice generator.
  • System shape. Metering job reads the outcome-event topic; writes one line per invoice cycle. No new LLM call.
  • Latency budget. None on the production path. Meter updates within one hour of the outcome event.
  • Cost ceiling. Under $50 per month per tenant on metering compute plus storage.
  • Eval. Every pilot customer's finance lead reconciles the meter against their own log monthly. Meter accuracy above 99% before the tier opens to general availability.
  • Instrumentation. Track migration rate, net-revenue impact per customer, and retention on the outcome tier. The customer who cuts their bill by switching is a churn risk in six months; the customer who doubles their bill validates the feature.
  • Two weeks in. Three pilot customers on the outcome tier with clean invoices. If they haven't, the meter is triggering on the wrong event — walk the trace with a pilot customer's ops lead to find the right one.

Feature two: the agent-vs-human scorecard on the customer’s admin surface. For the same task category your AI feature handles, tag every completed instance with who did it — agent or human. Publish the two-column scorecard on the customer’s admin surface: outcome per instance, dollars per instance, minutes to complete. The ops lead decides which slice to expand next quarter. Trust is built from the split shown, not the pitch made.

  • Shape. One new column on the existing task-log table (handler = agent | human). One page on the customer admin surface showing the two-column readout.
  • Data shape. Per (tenant_id, task_id, task_category, handler, outcome_tag, dollars, minutes, timestamp).
  • System shape. Aggregation query on the task log. No new LLM call. One optional summary line per column per week — "agent slice closed 62% at $4.10; human slice closed 71% at $18.40."
  • Latency budget. Zero on the production path. Under 2 seconds to render the last 12 weeks.
  • Cost ceiling. Under $30 per month per 5,000-seat tenant.
  • Eval. 40 customer-facing readouts per quarter reviewed by a senior CSM on "would the ops lead expand or shrink the agent slice from what this page shows?"
  • Instrumentation. Track which task categories the ops lead expands the agent slice on. Those categories are the ones ripe for outcome-based pricing next quarter.
  • Two weeks in. Three customers cite a specific task category from the scorecard on the next renewal call. If they don't, the categories are too broad — split them by workflow, not by feature name.

Both ship in two weeks with the team you already have. Both put a price on the work that used to sit on the seat license. The vendors who moved this fortnight bought the pieces of the SDR’s day and are now selling them back by the outcome. The vendors who didn’t will be the seat license the buyer negotiates down at the next renewal — after the outcome-priced competitor showed the exact dollar delta on the customer’s admin surface.

Sources

Send me an email and we will talk. If something here landed close to what you're working on, the door is open. No calendar funnel, no pitch deck — I read every note that comes in.

Doing the work rather than deciding what to build? Crafting is the column for that chair.

Email me Read more Signal

← All Signal issues · Drafted with Claude · Edited by Paul Brown