Why the AI Battleground in Sales Isn’t the Models

Roughly $2 trillion in software market value has been wiped out over the past year because investors and analysts were asking the wrong question about AI. That’s the framing from a sharp CNBC op-ed published this week, and it lands with more weight given the timing: Salesforce just reported its strongest quarter on record, with accelerating bookings and revenue that made the so-called “SaaSpocalypse” narrative look shaky at best.

The question everyone fixated on was which AI model would win. The question that actually matters for go-to-market teams is simpler and more urgent: who controls the context layer where AI does its work?

The Context Layer Is Where Sales AI Actually Lives

Models are commoditizing fast. That part of the debate is essentially settled. What isn’t settled is who owns the structured, reliable data pipeline that feeds AI agents inside a sales workflow – and that’s where the real competition is playing out. CRM vendors sit at the center of it.

Think about what a sales rep’s AI assistant actually needs to do something useful. It doesn’t need a smarter model. It needs clean, connected data – call transcripts from Gong, activity history from the CRM, deal stage signals from the sales pipeline, and intent data from external sources. Without that context, even the most capable model produces generic output that reps ignore after the first week.

This is why a quiet development on PyPI is worth paying attention to. A new open-source plugin called contextbase-plugin-gong appeared this week, built on PyAirbyte, designed to pull Gong conversation data into a unified context base for AI applications. It’s early-stage – alpha releases, rough edges – but it points directly at the problem serious RevOps teams are trying to solve: getting the right data into the right AI at the right moment.

What Salesforce’s Record Quarter Actually Signals

Marc Benioff publicly called fears of a SaaSpocalypse “nonsense” after the Q2 results dropped. He’s not wrong to push back, but the more interesting signal isn’t the revenue number itself. Accelerating bookings in a quarter where AI skepticism peaked suggests enterprise buyers aren’t abandoning CRM platforms – they’re doubling down on them as the integration layer for AI agents.

That makes sense when you think about where RevOps teams spend their time. The hardest part of deploying AI in a sales org isn’t choosing a model or even writing prompts. It’s ensuring the AI has access to accurate sales forecast data, clean account history, and reliable signals about deal health. CRM platforms that have spent years accumulating that data are now sitting on a structural advantage that’s very difficult for a model provider to replicate from scratch.

The vendors who figured this out early built moats around data quality and integrations, not raw AI capability. That’s a counterintuitive conclusion for a market that spent 18 months obsessing over model benchmarks.

What GTM Teams Should Actually Be Doing Right Now

If the battleground is context, the most valuable investment a sales or RevOps team can make right now is auditing their data infrastructure – not their AI tools. A few things worth examining:

• Are call recordings from tools like Gong actually flowing into your CRM in a structured, queryable format, or are they sitting in a silo?
• Is your ideal customer profile definition codified in your CRM in a way that AI agents can reference and act on?
• Can your AI assistant see deal history, contact engagement, and competitor mentions together, or does it only have access to one source at a time?
• Do you have a feedback loop that lets you measure win rate changes after AI-assisted interactions, so you can tell what’s actually working?

Most teams answer “no” to at least two of those. That gap – not model quality – is what’s limiting AI performance in real sales environments right now.

The open-source community is starting to build toward solutions here. The contextbase-plugin-gong project is an early example of developers trying to standardize how conversation intelligence data gets piped into AI contexts. It won’t be the last. Expect this category of “context infrastructure” tooling to grow quickly over the next 12 months as the model wars quiet down and the data wars heat up.

The CRM Platform as AI Operating System

Here’s the view that’s hardening: the CRM isn’t just a database anymore. It’s becoming the operating system through which AI agents take action inside a sales org – and that reframes how you should evaluate your current tool stack entirely.

A platform with deep integrations, a strong data model, and the ability to connect AI agents to live deal context is worth far more to a GTM team than a standalone AI point solution that produces smart-sounding output disconnected from your actual sales cycle. The net revenue retention case for the former is much stronger, because the AI gets more useful as your data matures – it compounds.

That’s what Salesforce’s quarter is actually telling the market. Enterprises aren’t paying for AI abstractions. They’re paying for AI that knows their accounts, their history, and their customers – and that knowledge lives in the CRM.

If you want to stay current on how this space is developing, the CRM Daily Newsletter covers these shifts weekly – no hype, just what’s actually changing in sales and GTM.

The model race was always a distraction. The data race just started.