For the previous yr, the software program business has been wrestling with an uncomfortable query: Will the frontier labs’ AI brokers eat SaaS?
At Dreamforce yesterday, Salesforce CEO Marc Benioff supplied a provocative reply. The SaaS apocalypse is not essentially in regards to the finish of software program, he famous, however it might be in regards to the finish of software program that makes people do all of the work. That is an attention-grabbing repositioning of the talk.
In my earlier evaluation of the SaaS apocalypse, I argued that the menace extends past brokers changing purposes. It contains the erosion of seat-based subscriptions and the margin strain created by variable AI prices.
These pricing variables are tougher to foretell, defend, and finances for, and findings from G2’s 2026 Purchaser Habits Report reveal patrons are already demanding extra flexibility in response. Particularly, 70% of patrons say the tempo of know-how innovation is pushing them towards shorter contracts, and 80% say their group offers builders or technical groups with a devoted token or LLM utilization finances, whereas choice for outcome-based pricing greater than doubled in a single yr (from 11% to 23%).
Benioff’s keynote might not have squarely addressed these financial questions, but it surely supplied a compelling imaginative and prescient for the way established software program corporations can evolve and stay aggressive.
Listed below are my greatest takeaways.
1. AIforce may flip trapped enterprise information into discovered cash
Benioff launched AIforce as Salesforce’s new AI interface high layer, bringing collectively its information, purposes, brokers, and enterprise intelligence. His central worth proposition? Unlocking worth that is already sitting inside enterprise techniques of file.
Consider it as discovered cash. Corporations have spent many years accumulating buyer information, constructing purposes, and defining enterprise processes. A lot of that intelligence stays trapped in techniques that require people to navigate screens, generate experiences, and join the dots.
Benioff believes AIforce can change that equation, making enterprise intelligence accessible by a dynamic interface that finds patterns and delivers solutions.
He captured the imaginative and prescient in a single assertion:
“And now for the primary time we’re speaking about this, a dwell interface on Salesforce, unlocking this entice worth, discovering the patterns throughout all of your information, getting solutions in seconds, not days.”
Marc Benioff
CEO, Salesforce at Dreamforce 2026
That is the actual alternative for enterprise AI. Not merely doing present work quicker, however discovering enterprise worth that was beforehand too troublesome or costly to extract.
2. The enterprise AI breakthrough comes from the platform, not the mannequin
One among Benioff’s most essential observations was that fashions alone can’t run the enterprise.
LLMs are probabilistic, extremely able to reasoning, producing content material, and navigating advanced workflows — however they don’t seem to be inherently grounded in an organization’s supply of fact.
Enterprise techniques, in the meantime, include deterministic information, enterprise guidelines, permissions, and workflows. Benioff outlined 4 layers that should work collectively: information, apps & semantics, brokers, and interface.
His thesis is that combining these worlds creates the inspiration for dependable, enterprise-grade AI. I agree with the path. The mannequin offers intelligence and adaptability, whereas the enterprise platform provides the context, governance, and controls essential to show that intelligence into reliable enterprise outcomes.
The aggressive benefit might more and more belong to corporations that may join these capabilities, not merely these providing entry to the neatest mannequin.
3. Headless purposes will make the normal software program interface out of date
Salesforce’s transfer towards headless purposes was one other fascinating a part of the keynote.
As an alternative of requiring customers to log into conventional purposes, Salesforce can expose its underlying performance and intelligence by AI interfaces. Some would possibly see this as destroying the worth of the normal software program interface, however Benioff sees it as creating a completely new one.
He described an interface that is dwell, dynamic, and clever, somewhat than static. It will possibly adapt to customers and assist them construct, administer, and function purposes.
Take into account the evolution from DOS to graphical interfaces, then browsers and cellular apps. Every transition essentially modified how individuals interacted with software program. AIforce suggests we’re getting into one other such transition. And Salesforce says its merchandise are constructed with zero information retention, which means buyer information is not used to coach the underlying fashions. Benioff emphasised this as a important a part of sustaining buyer belief.
The broader lesson for software program distributors is that they now have to rethink the interface with out sacrificing the enterprise capabilities beneath it.
4. Claudeforce is Salesforce’s reply to the SaaS apocalypse
In an act of business-judo, Benioff additionally launched Claudeforce, integrating Salesforce’s enterprise intelligence straight into Anthropic’s Claude Cowork surroundings. He defined that Salesforce is already utilizing it internally, together with on the COO stage.
That is notably attention-grabbing given considerations that prospects will use Claude Code and comparable instruments to construct their very own CRMs. As an alternative of competing with Claude, Salesforce is becoming a member of forces with it. Salesforce retains its system-of-record worth whereas permitting prospects to work together with it by an AI surroundings they already use.
Benioff additionally launched Slackforce, which affords the flexibility to jot down, assessment, and ship code collaboratively with brokers inside Slack. He shared his statement that coding is turning into a multiplayer sport. Pair programming is not new, however bringing collaborative growth and AI brokers straight into the move of labor may change how software program will get constructed.
(Clearly, Benioff has learn The Artwork of Struggle.)
5. The agentic enterprise is beginning to produce critical numbers at Salesforce
Maybe probably the most eye-opening second got here when Benioff mentioned Salesforce’s personal AI brokers. He claimed that its Hunter agent generated $500 million in pipeline final quarter. If sustained, that is a $2 billion annualized pipeline run fee attributed to an AI agent.
I have not encountered a declare of this magnitude over the previous few years. In fact, pipeline is not income, and the keynote did not set up how a lot of that pipeline was incremental or how a lot will convert. However the scale of the declare is outstanding.
Benioff additionally highlighted uneven AI adoption, contrasting autonomous autos in San Francisco along with his expertise in Europe, the place he hadn’t encountered them. He sees comparable disparities amongst enterprise prospects.
The chance is big, however adoption stays uneven throughout the globe.
My takeaway: The following enterprise AI battle is about unlocking worth
Dreamforce bolstered the purpose that the way forward for enterprise software program is not about changing people with brokers. It is about combining trusted enterprise intelligence with probabilistic AI and making that intelligence accessible by interfaces that may really get work carried out.
Benioff’s imaginative and prescient is bold, and the execution, economics, and measurable buyer outcomes will finally decide how a lot of it turns into actuality.
However one factor is turning into clear: The businesses that work out the way to unlock the worth already sitting inside their prospects’ techniques might have a completely new development alternative forward of them.
If you happen to missed it, watch Marc Benioff’s full Dreamforce keynote.
