First reaction on Claudeforce

What are you all thinking about the Claudeforce announcement this week?

Four thoughts immediately come to mind for me:

1. If your people already like Claude, they’re probably going to like this.

Getting value from AI in a company relies heavily on adoption and confidence.

If your people are already using Claude (or another LLM) in their personal or professional lives, they already know how to start a conversation, how to prompt, how to explore, and roughly what these tools are good at. You already eliminate that first adoption barrier we’ve seen with Agentforce/Einstein Copilot.

If you’re thinking about preparing your team for the next steps with AI in Salesforce, maybe giving them an afternoon to go somewhere and experiment with Claude could be hours well spent. Or a team event sharing personal and professional best practices and use cases in action.

2. What happens when the CRM data isn’t perfect?

Looking at the use case in the demo — “show me high-potential customers with no recent touchpoint” — I immediately heard that nagging voice in the back of my head:

“But what if the recent touchpoint wasn’t recorded?”

Because we all know that happens.

You can design processes to make sure data is updated, but there will always be edge cases. And if we’re asking AI to help with critical thinking like which customers should I prioritise?, we’re asking it to make those recommendations based on data we know isn’t always perfect.

But I’m increasingly thinking:

So what?

Let Claudeforce help your sales team prioritise customers using the best information you have.

If someone says, “Hang on, I spoke to them yesterday” and Claude doesn’t know about it because they didn’t update Salesforce, they’ve just been given a pretty compelling demonstration of why keeping Salesforce updated matters.

AI doesn’t just depend on good CRM adoption. AI gives people a reason to adopt the CRM as intended.

What might you lose along the way? And what might you gain?

And if KPI tracking remains in reports and dashboards, someone is still making sure there aren’t abandoned nooks and crannies of potential.

3. And then there’s pricing.

This is probably the biggest question for me.

At the same time as some companies are exploring vibe-coding CRMs partly to reduce software costs, which companies will choose to pay extra if a credit model is introduced for Claudeforce?

And how should we calculate the ROI?

Maybe some of the value isn’t directly attributable revenue at all, but instead becomes partly an employee experience benefit:

“Come work here. You don’t have to manually update records in the CRM and click around looking for information.”

Could access even become partly based on user preference — not too dissimilar to height-adjustable desks? Some people aren’t bothered. Others wouldn’t want to work without one.

I’m really looking forward to hearing about real Claudeforce use cases where companies have calculated the ROI.

4. Where does this go next?

One thing immediately comes to mind for me: meeting transcripts and call transcripts.

I’ve personally used AI to help me read between the lines of conversations and understand the obstacles that weren’t actually said out loud. Slightly unnervingly, AI can be better at this than me.

I’d love to get to the point where Claude can tell me:

“Your customer went quiet when you raised this point. That’s interesting because, in a meeting three months ago, they said they were against a similar idea.”

Imagine combining the structured history in Salesforce with the unstructured context buried in months of meetings, emails and conversations.

That’s a very different kind of CRM.