SaaSpocalypse or Interface Revolution? What Dreamforce '26 Really Means for Your App Rationalisation Roadmap

When Jensen Huang and Marc Benioff share a Dreamforce stage and both call the same idea 'crazy nonsense', you should probably pay attention — not because they're right, but because two of the most powerful people in tech agreeing on something usually means someone stands to make a lot of money from you believing it.

The idea they were dismissing is the so-called SaaSpocalypse: the theory that AI agents will hollow out traditional software-as-a-service by making the underlying apps interchangeable, or worse, unnecessary. Why pay for a subscription seat when an AI can do the work the software used to gate behind a login? Benioff, unsurprisingly, thinks that's nonsense. Salesforce sells seats. Huang sells the chips that power the AI everyone's excited about, so he's happy to talk up AI while reassuring the software world it isn't going anywhere.

Meanwhile, out in the actual market, SaaS stocks spent much of 2026 getting sold off. Investors clearly haven't fully bought the 'nothing to see here' message. So which is it — apocalypse or overreaction? And more to the point, what should you do about your own stack of applications when the people telling you the answer all have a financial stake in it?

The keynote you were meant to remember

Dreamforce 2026 leaned hard into the Claudeforce narrative — Salesforce's deepening partnership with Anthropic, positioning Claude-powered agents as a native layer across the platform. The pitch is elegant: you don't rip out your software, you put an intelligent layer on top of it. Your CRM stays, your service desk stays, your marketing tools stay, and AI agents stitch them together and do more of the busywork for you.

It's a comforting story if you're a vendor. 'Keep everything you're paying for, and pay us a bit more for the clever layer' is a much easier sell than 'half your apps are about to become redundant'. The 'AI on top of SaaS' thesis is self-serving precisely because it protects existing revenue while opening a new line of it.

That doesn't make it wrong. But it does mean you shouldn't take it at face value when you're deciding where your budget goes for the next three years.

What the stock sell-off is actually saying

The 2026 SaaS sell-off isn't a vote for the full apocalypse. It's a reasonable market reflecting an uncomfortable truth: a lot of software has been priced as if per-seat growth would continue forever. AI genuinely does threaten that model in specific places.

Think about the tools you pay for per user, where the value is really about a human doing repetitive input or lookup work. Basic ticket triage, first-draft copy, data entry, report assembly, simple scheduling. If an agent handles a chunk of that, do you need as many seats? Probably not. That's the part of the SaaS world that's genuinely under pressure, and the market is right to reprice it.

But plenty of software isn't just a labour interface. Your finance system holds the record of truth. Your CRM stores relationships and history no agent can regenerate. Your compliance and payroll tools carry legal weight. Those aren't going anywhere, and their value doesn't evaporate because an agent got clever.

The honest read is neither apocalypse nor revolution. It's a sorting. Some apps become more valuable, some become features inside something bigger, and some quietly lose their reason to exist.

A framework that ignores the keynote

App rationalisation is unglamorous work: going through everything you pay for and deciding what stays, goes, or merges. Most SMEs we work with in Nottingham and across the East Midlands are carrying more subscriptions than anyone can name off the top of their head. AI is a reason to do this exercise now, not a reason to panic-buy the layer being sold to you at a keynote.

Here's how to think about each application, stripped of vendor spin.

Ask what the app actually holds versus what it does. If an app mainly does something — processes, formats, routes, drafts — its job is exposed to AI and agents. If an app mainly holds something — the customer record, the ledger, the contract archive — it's far more defensible. Rank your stack on that axis first. The 'holders' are keepers. The 'doers' are candidates for consolidation.

Follow the per-seat money. List every tool you pay for by user, then ask honestly how much of each seat's value is a person doing work an agent could reasonably take over in the next 18 months. That's your exposure. It's also your savings opportunity — not by cancelling on day one, but by watching whether you can reduce seat counts as workflows change.

Look for overlap before you look for AI. Most SMEs could cut cost today without touching AI at all, simply by killing duplicate tools. Two project trackers. Three places to store files. A CRM and a separate contact database that half the team ignores. Rationalisation pays for itself here long before any agent gets involved. Do this first — it's real money and it has nothing to do with Dreamforce.

Be sceptical of platform lock-in dressed as convenience. The Claudeforce pitch — one intelligent layer across everything — sounds like simplification. It can also be a route to depending entirely on one vendor's ecosystem, at a price they set. There's a real trade-off between an integrated platform and keeping your options open. Neither is automatically right, but you should choose it deliberately rather than drift into it because the demo was slick.

Judge AI features on your data, not their stage. Keynote demos run on clean data and rehearsed prompts. Your business runs on messy, inconsistent, half-migrated data. An AI layer is only as good as what it's built on. Before you pay for the clever bit, ask whether your underlying records are tidy enough for it to work. Often the honest answer is that you'd get more value spending on data hygiene than on another AI subscription.

What to actually do this quarter

Don't restructure your entire stack around a partnership announcement. Do use the noise as a prompt to do the boring, valuable work.

Start with a full inventory of what you pay for, who uses it, and what it costs per year. Nearly everyone finds surprises. Cut obvious duplicates and dead subscriptions — that's guaranteed savings with no downside. Then mark each remaining tool as a holder or a doer, and flag the per-seat doers as the ones to watch. For those, run one small, real pilot of AI or automation against your own data before committing to anything platform-wide.

And treat every vendor's AI story — including Salesforce's, including Anthropic's, including whoever your account manager is — as a claim to be tested, not a truth to be adopted.

The uncomfortable middle ground

The SaaSpocalypse won't happen the way the doomers say, and it won't be the non-event Benioff and Huang would like you to believe either. What's coming is a slow reshaping where the software that stores your most important information gets more valuable, and the software that just moves work around gets squeezed.

Your job isn't to pick a side in a keynote argument. It's to know your own stack well enough that when the vendors change their pricing and their pitch — and they will — you already know what you'd keep and what you'd let go.

If you'd like a straight, vendor-neutral look at what you're paying for and where AI genuinely helps versus where it's just a sales line, that's exactly the kind of review we do. No stage lighting, no crazy nonsense — just your actual costs and a clear plan.