Resources in This Newsletter:

Outbound Pipeline Inspection

In March 2026, Salesforce co-founder Parker Harris stood on a stage in San Francisco and said, “Why should you ever log into Salesforce again? Maybe you never will.”

The next month, Salesforce Headless 360 was announced.

This reimagining of the platform no longer ties users to a browser-based UI. Instead, through Model Context Protocols (MCPs), AI agents (like Claude, ChatGPT, or whatever tool you build yourself) can translate a plain-language request into the right SF query without needing to know SF’s API.

With the right workflows, you can now go days, weeks, or even months without needing to open Salesforce.

Truly, a dream come true for sales reps and revenue leaders everywhere!

As the SVP of Delivery for USC, it’s my job to know everything I can about these changes, how to get the most use of them, and the risks involved.

I sat down with our Founder and CEO on the GTM Science podcast to get into it. You can listen to the full episode here, or continue reading for the key points.

What’s Possible Right Now

We’re seeing a wide range of use cases for Headless 360 opening up across the manual-to-automated spectrum.

Here are some examples:

CRM Structure Diagnostics in Minutes

Every Salesforce org accumulates fields and objects over the years that become obsolete. Before, a human had to go field by field figuring out which ones currently matter (which, as I’ve experienced first hand, could take 5 hours). Now, pointing an AI at your object and field structure can surface which fields are actually being used (ie; by last updated, how many records populated, etc.) in minutes.

Data Pulled from Wherever it Already Lives

Now, your data can sit in 20 different systems, as long as the AI has clear instructions on where to go get it. You can get a report that pulls from all of those systems, on demand or on a schedule, no formal integration required. Everything does not need to be updated in Salesforce. As a GTM Ops firm, I never thought I’d see the day we’d be saying that!

Call Notes and CRM Fields Auto-Updated

Call recording tools can populate notes, next steps, contact roles, and decision-criteria fields directly from the call, with the rep reviewing before it pushes to Salesforce.

Fully Automated Call Prep

Instead of a rep spending their precious selling time pulling together account history, you can give an AI instructions once to assemble a prep document (with clear prompts like which fields matter, which sources to check, etc.). Have your AI tool scan a rep’s calendar daily, check the domains and subject lines on each invite, and build the prep sheet automatically.

Autonomous Pipeline Ops

Here, an agent scans open pipeline on its own schedule. It flags deals that have gone quiet or slipped their close date, drafts follow-ups or suggests ownership reassignment, and adjusts forecast recommendations continuously. Although there absolutely should be a human in the loop making the final judgement call on agent recommendations, a manager doesn’t need to pull reports or click through deals to manage pipeline.

Here Are the Risks

I push back on the word “hallucination” here. The bigger risk isn’t Claude or another AI making things up (though that can happen), it’s the tool not having the institutional knowledge a human admin has. Other risks include over-agreeableness, unchecked write-backs, inconsistent reasoning, and tool-specific limitations.

Here are the potential hazards your team should be aware of:

Field Overload

20-30 reasonably labeled fields is manageable. With ~300 fields (many redundant or unused), the AI has no way of knowing which “amount” field or which “open” definition is the right one, unless you tell it. Every field looks equally important to the model until you say otherwise.

Although you can do a massive cleanup project to fix this, you don’t have to. Just document the 2-3 fields that matter for each use case, the way you’d brief a new hire with zero institutional context.

Over-Agreeableness

AI tools default to being agreeable. Left without guardrails, they’ll guess and lean toward telling you what you want to hear. This is also why you probably shouldn’t use AI as a therapist.

Unchecked Write-Backs

Reading data is much lower risk than having AI update (“write-back”) data into your CRM. My advice: take the training wheels off faster on read-only use cases; move slower and add more guardrails anywhere you’re writing back to a database.

Inconsistent Reasoning

For hard numbers, don’t let AI infer the query every time. Document the exact SOQL query (Salesforce’s structured query language) once, and have the tool run that instead of reasoning its way to an answer from scratch.

If you ask Claude “what’s my open pipeline value” without that guardrail, you can get a different number twice in the same conversation. No bueno if you’re building a board deck.

Tool-Specific Limitations

Watch out for point solutions that only write to their own proprietary fields. If a tool can only update its own version of your MEDDIC fields instead of the ones you’ve built and reported on for ten years, you’re looking at a rebuild of your fields and your reports.

What Shouldn’t Be Automated

Some decisions should stay human:

Never let AI move a deal to a loss. That’s a judgement call, for a long list of reasons.

Only sometimes let AI move a deal to a win. Contract signatures triggering a close-won has been an automated thing for years, and that’s fine. Moving a deal on a conversational signal, like something said on a call, is where I draw the line.

Letting AI move a deal between earlier pipeline stages? That depends entirely on your sales motion and deal size. There’s no universal answer here.

The Bottom Line

You no longer need a clean Salesforce instance to do a lot of these tasks, but you do need clear instructions for the things that matter for each specific use case. Start slow with read-only workflows, prove it out, then move carefully into workflows that write back to your database, with humans in the loop checking.

The goal here is to get your reps out of Salesforce and in front of prospects, doing what they do best, while giving leadership better, faster visibility into what’s happening in the pipeline.

Remember that AI won’t help you reach either of those goals if your processes are not dialed in and executed consistently among your teams.

Need help determining if your Outbound motion is ready for AI, and/or determining which foundations need to be laid to get the most use out of Headless 360? Check out our Outbound Pipeline Inspection offer, free for qualifying applicants.

This isn’t a veiled sales pitch, it’s a real consulting session with either myself or our CEO Eddie, where we spend the whole hour finding the most immediate value for your company.

If you’re looking for more content on optimizing your outbound motion, check out our Outbound Efficiency Framework.