A rep finishes a 45-minute discovery call. On that call, the prospect mentioned their budget got cut by a third, their champion is leaving in August, a competitor came up twice, and the timeline slipped to Q1. Real signals. The kind of information a forecast should be built on.

Here is what lands in Salesforce: “Good call. Sending proposal. Follow up next week.”

Everything that mattered stayed inside a recording nobody will ever rewatch. Multiply that across every rep, every call, every quarter, and you start to understand why pipeline reviews so often feel like archaeology. The CRM is not lying, exactly. It just only knows what someone bothered to type.

Salesforce put a name on this in the Summer ’26 release announcement: most conversation data never makes it into Salesforce, which leaves deal signals incomplete and agents working blind. That second part is the one architects should sit with. Agentforce Sales can only reason over data it can see. If your richest deal context lives in Zoom recordings and email threads, your prospecting and engagement agents are running on the thin typed residue of the real relationship.

Momentum is a conversational insights and revenue orchestration platform. Its core asset is a universal ingestion engine: it captures customer interactions wherever they happen, including third-party video platforms like Zoom and Google Meet, plus calls and email, and converts that unstructured audio and text into structured intelligence. Salesforce signed the definitive agreement on February 18, 2026 and closed the deal on March 2. By June it was already listed on the Agentforce Sales release page, available standalone, with native integration targeting GA this quarter.

That pace tells you something. Salesforce President and Chief Product Officer Steve Fisher framed the acquisition around one idea: to deliver on the promise of agents, the platform needs visibility and context from every meaningful interaction, so Agentforce 360 can incorporate the true voice of the customer into complex, multi-step workflows.

Zoom out and Momentum is one move in a bigger sequence. Over roughly six months, Salesforce completed ten acquisitions, and the pattern is coherent: Informatica for enterprise data quality and governance, Qualified for top-of-funnel demand capture, Cimulate for intent-aware commerce search, and Momentum for conversational context. Salesforce describes its platform as four layers: context, work, agency, and engagement. Almost every one of those deals strengthens the context layer, because context is what separates an agent that acts well from one that merely acts.

I read the acquisition run as Salesforce executing a thesis with real money: agents win or lose on the data underneath them, so own the pipes that produce the data. Momentum is the pipe for the highest-value data source in all of sales, which is what customers actually say.

The mechanics matter here, because this is an architecture story disguised as a product launch. Momentum runs a four-stage pipeline from raw conversation to agent action.

The phrase worth memorizing is the write-back. Plenty of tools transcribe calls. Plenty summarize them. The difference here is that Momentum closes the loop into the CRM data model itself, in real time, so the opportunity record reflects what happened on the call before the rep has even poured their next coffee. The rep stops being the system of entry. The conversation is.

For architects, the write-back is also where your work begins. Structured signals landing on records means field mappings to review, activity timelines to keep clean, and deduplication rules that now have a new, very prolific author writing into your org. An automated writer with perfect attendance changes your data volume assumptions. It also changes your data quality ceiling, dramatically for the better, because the source is the primary conversation rather than a human summary of it.

One more piece deserves mention: the new In-Person Meeting Assistant in Conversation Intelligence captures and transcribes face-to-face meetings on-device through the Salesforce mobile app. Field sales conversations, historically the least documented interactions in any org, now flow into the same structured pipeline. Between Momentum and on-device capture, the coverage map for customer conversations is approaching complete.

Here is why Momentum lands at exactly the right moment. The June 2026 Agentforce Sales release reads like a roster of agents that are hungry for conversation data, and every one of them gets sharper the moment that data flows.

Notice the dependency running underneath all six. Deal Risk Alerts flag activity drops, which only works if activity is captured. Meeting prep briefings are only as good as the record of the last meeting. Prospecting hit-lists ranked on CRM signals need the signals to exist. Momentum is the feedstock. The agents are the machinery. This release is the first time the two halves of agentic selling, the intelligence and the raw material, ship as one connected story.

Momentum being available standalone today, ahead of the native GA, is a gift. It gives you a sanctioned pilot window. Here is how I would sequence it.

Every powerful capability arrives with homework, and pretending otherwise helps nobody. Three things deserve honest planning.

Consent is a design input, not a checkbox. Call and meeting recording sits under a patchwork of regional rules, and your org probably sells across several of them. None of this is a product limitation; it is simply the territory that comes with recording customer conversations. The teams that handle it well design consent into the meeting workflow from day one and configure capture behavior per region.

Trust the signals, verify the stakes. AI-extracted deal data is a massive upgrade over three typed lines, and it will still occasionally mishear a number or attribute a quote to the wrong speaker. Early on, keep a human review step on high-stakes fields like amounts and close dates while confidence builds. In my experience, that review period is short, because reps quickly notice the auto-captured record is more complete than anything they used to write.

Frame it as liberation, not surveillance. Adoption is won or lost in the announcement email. The same capability can be introduced two very different ways, and only one of them works.

Step back and the direction is unmistakable. For twenty years, CRM accuracy depended on the discipline of the busiest people in the company, and it showed. Salesforce just inverted the model: the record keeps itself, and humans spend their attention on the conversation instead of documenting it. Agentforce Sales supplies the workforce. Momentum supplies the memory.

So here is the question I would put to any revenue leader this quarter: if your forecast was rebuilt tomorrow from what customers actually said, rather than from what reps had time to type, how different would it look? If that question makes you slightly nervous, you already know the answer. The good news is that closing that gap just became a product decision instead of a culture war.