By Remco Livain in IMHO — 21 Oct 2026

Salesforce isn't the system companies run on. That's the problem.

SAP runs the back office. Salesforce runs the front of the funnel. In an AI-rebuilt company, that difference decides who gets replaced and who just gets slower.

I spent part of this week thinking about a company I don't work with and have no stake in: Salesforce. Not because of the stock, but because it's the cleanest test I know for one of my own convictions. If organisations really get rebuilt around AI, what happens to the software they were built around?

My short answer: Salesforce will still be here in 2030. It just won't be the company it is today. And the reason has less to do with AI than with where Salesforce sits inside a business.

Front of the funnel, not the back office

The comparison people reach for is SAP. Both are giant enterprise software vendors with huge consulting ecosystems around them. Both get called expensive, slow and hard to leave. But they are not in the same position.

SAP runs finance, payroll, supply chain and production. Those are the processes a company literally cannot operate without. Ripping SAP out is a multi-year, board-level risk, which is why it survives even while its own customers complain about it.

Salesforce sits at the front of the funnel: leads, pipeline, service tickets. Important, yes. But it is not the operating system the company runs on. It's a database people log into when they have to.

SAP runs the processes a company can't operate without. Salesforce runs the ones it would rather not lose.

SAP has a migration problem. Salesforce has a substitution problem.

What AI actually threatens

Salesforce has spent two decades selling seats. One user, one licence. That model works beautifully as long as humans do the work. It stops working when a sales team of forty becomes a team of ten with a layer of agents doing the rest.

Illustrative. Per-seat SaaS pricing breaks when agents take over work that used to need a licence per person.

Salesforce knows this. It has been moving toward outcome-based pricing, and investors have been nervous enough that the stock was down roughly a third this year by late May on fears of seat compression. Meanwhile its agent business is growing fast but is still small next to the core: about $800 million in Agentforce ARR against a $41.5 billion company.

Source: Salesforce Q4 FY26 results, via Constellation Research.

The deeper problem is the surface. Agents start where employees already are: their email, their documents, their calendar, their chat. That's Microsoft and Google territory. And the intelligence layer on top, the part that reads the customer data and decides what to do next, is increasingly something OpenAI, Anthropic and Google can offer directly. I've written before about how agents often don't need the interface we assume they need. A CRM screen is one of those interfaces.

The squeeze: agents start on the surface people already use, and the intelligence on top can be bought directly. The CRM screen sits at the bottom.

The ecosystem is there, but its work is exposed

Salesforce does have what SAP has: thousands of partners, certified admins and developers. The difference is what that network does all day.

SAP's experts customise core processes that AI won't replace any time soon. They'll simply do it faster with AI, and keep building for customers in a fairly traditional way. Much of the Salesforce ecosystem configures workflows, dashboards and automations, which is exactly the kind of work agents and AI-native tools are getting good at. Same size of network, very different exposure.

This is the same point I keep making about teams. The leverage is the domain judgement, not the tool. Configuration without judgement is precisely what gets automated first.

Why it won't happen overnight

Here's where I disagree with the "SaaS is dead" crowd. Those of us working at the forefront of this shift consistently overestimate how fast the rest of the world moves. Multi-year contracts, compliance, procurement and plain inertia are real. They buy Salesforce time to reprice and repackage.

So I don't expect a collapse. I expect a split. The large, regulated enterprise base holds through 2030. The mid-market leaks, because that's where switching to something cheaper and AI-native is easiest. Growth slows, margins come under pressure, and more of the value flows to whoever owns the daily surface and whoever owns the intelligence.

My expectation for 2030: not a collapse, but a split. The risk isn't dying; it's becoming the IBM of CRM.

The real risk for Salesforce isn't dying. It's becoming the IBM of CRM: profitable, entrenched, and no longer where the action is.

What this means if you're the customer

If you run marketing or sales, the question isn't whether to leave Salesforce tomorrow. It's whether your next system design still assumes one licence per person doing a task a person no longer needs to do. The same goes for your storefront: AI agents are already shopping on behalf of your customers, and they don't care which CRM you bought. Design for the team you'll have in three years, not the org chart you have today.

Being too big to fail isn't the same as being too important to replace.

Sources & further reading

External

Constellation Research — Salesforce Q4 FY26 results and fiscal 2030 outlook
CX Today — Salesforce's outcome-based pricing and AI seat compression
Gotrade — Salesforce Q1 FY27 and the SaaS seat debate