Why are you still logging into Salesforce?

Death by a thousand clicks

Think about what it takes to prepare for a single customer call in Salesforce. Open the account. Scroll to the related lists. Click into the open opportunity. Back. Click into the last case. Back. Open a report, change the filters, wait, export it because the view you need doesn't exist. Then do it again for the next call.

None of those clicks is hard. Together they're a tax on every person who uses the system, every day. It's why reps put off updating records, why managers build side spreadsheets, and why so many companies pay for Salesforce and still don't trust what's in it. It isn't that people don't care. The interface makes the simple things slow.

That's what's about to change.

What we're seeing with our clients

We're building Claude artifacts for clients right now: custom, well-designed screens that pull their Salesforce data into Claude, where people can see it, ask about it, and update it. Once a team has one, something happens quickly. They stop opening Salesforce.

Not because Salesforce stopped mattering. Because the screen in Claude is simply a better place to do the work. It shows what that person needs and nothing else, it answers questions in plain language, and it doesn't make them hunt through tabs to find it.

Our view is simple: soon, nobody at a company will need to log into Salesforce, not the reps, not the managers, and not even the admins and developers. Claude will be the interface where they interact with and manage their Salesforce data. Salesforce will be the engine behind it.

The infrastructure is already moving that way. Salesforce now offers a hosted connector service for AI tools called Headless 360, and the Claudeforce partnership puts Claude directly on top of Salesforce data and permissions. The pieces are in place. What's left is designing the experience people actually use, and that's the work we're doing.

What replaces it: a Claude artifact built for each role

Instead of a generic page layout that tries to serve everyone, each role gets a screen designed for its actual job. Built as a Claude artifact, it pulls live Salesforce data through the connector, using that person's own permissions, and shows exactly what they need:

• An account executive's morning: today's meetings, each with the account's open deals, last activity, open cases, and what changed since the last call, on one page.
• A sales leader's pipeline: the deals that will decide the quarter, what's slipping, and why, without building a report.
• A customer success manager's book: renewals coming up, health signals, and the accounts that need a call this week.
• A finance view: closed deals waiting to be invoiced, anything that doesn't match, and, for accrual-based accounting, the status of active projects, so finance knows what revenue to expect to recognize and when.

And because it's Claude, people don't have to know where the data lives. They ask: "Which of my deals haven't had activity in two weeks?" "What did this customer buy last year?" "Draft the follow-up from yesterday's call." When something needs to change in Salesforce, it goes back through approved actions, with the same rules and permissions as before.

Salesforce stays the system of record. The clicking goes away.

What doesn't go away

Going headless doesn't make Salesforce less important. It makes the foundation more important, because now everything people see comes straight from it.

• The data has to be true: duplicates, stale records, and empty fields show up directly in every artifact and every answer.
• Permissions have to be right: the artifact sees what the user can see, so overprovisioned access becomes visible fast.
• Definitions have to be written down: a data dictionary and query guide so every artifact and every AI tool agrees on what "pipeline" or "active customer" means.
• Admins and developers stop clicking too: Claude can build the configuration, write the automation, and make the changes through the connector. The people with years of Salesforce experience direct the work and review every change, instead of doing it screen by screen. Their judgment matters more than ever.

How to start

Don't try to replace the whole interface at once. Pick one role and one daily routine, and measure it first: how many clicks, how many minutes, how many side spreadsheets. Clean the data and permissions that routine depends on. Build the artifact for that one screen, train the people who'll use it, and measure the after.

When the numbers move, do the next role. That's how the UI goes away: one routine at a time, each with a measured result.

Spend your time wisely — let agents do the rest.