AI agents are coming for the SaaS stack: five signals from September
Mega-rounds, acqui-hires, prompt-to-app builders and MCP all point the same way — software that does the work, not software you use to do it.
For a decade, the SaaS playbook was simple: find a workflow, build an app for it, charge per seat. September 2026's news suggests that playbook is being rewritten. Funding, acquisitions and product launches are all pointing to the same idea: software that does the work, not software people use to do the work.
Here are five signals from the past few weeks, and what they mean if you buy or build SaaS.
1. Agent platforms are getting growth-stage money
- Ema raised $77 million to run teams of agents across HR, IT and finance. It says it has $150 million+ in bookings and ~180% net dollar retention.
- Wonderful raised $550 million at a $5 billion valuation, more than double its value six months earlier.
Both pitch the same thing: a layer that sits above the apps a company already pays for and gets them to work together.
2. Incumbents are buying agent teams
- Asana paid $75 million for StackAI, a no-code agent builder.
- Adobe bought Rilo, a marketing workflow automation startup, and is shutting down Rilo's standalone product.
Established vendors are deciding it's faster to buy agent expertise than build it, and to fold it into suites customers already have.
3. Low-code platforms want to generate the whole app
Microsoft's September Power Platform update made its prompt-to-app builder generally available. It generates tables, forms, views and security roles, not just screens. Softr's AI Co-Builder, launched in March, takes the same approach for portals and internal tools.
4. AI assistants are becoming the interface
Softr's MCP launch lets Claude, ChatGPT and other assistants build and edit apps directly. The chat window becomes the place work starts, and the no-code tool becomes the system that runs it. More on that trend here.
5. The infrastructure underneath is scaling fast
Gimlet Labs' $300 million round is a reminder that all of this depends on cheaper, faster inference. As running agents gets cheaper, companies can afford to run more of them.
What it means for you
If you buy SaaS:
- Before renewing a point tool, ask whether an agent layer could run that workflow across the systems you already have.
- Push vendors on usage-based or outcome-based pricing. Seat counts make less sense when agents do the work.
- Treat data export and API access as must-haves. Acquisitions like Rilo's show that young tools can disappear.
If you build SaaS:
- Being the system of record matters more than ever. Agents need somewhere trustworthy to read and write.
- Ship an MCP server or a solid API so assistants can operate your product.
- Rethink pricing before your customers do it for you.
The bottom line
The SaaS stack isn't going away. The apps are still where data lives and rules are enforced. But more and more, agents will be the ones using them. The winners of the next few years are likely to be the products that are easiest for an agent to operate safely.
The Appboxs newsroom covers launches, funding, acquisitions, pricing changes and AI across the SaaS and no-code world. Every story links to its primary sources. Have a tip, a correction or a story we should cover? Send it through our contact page.
Related stories
Ema raises $77M to replace SaaS busywork with teams of AI agents
Newsroom · 23 Sept 2026 · 2 min readWonderful raises $550M at a $5B valuation for its enterprise AI operating system
Newsroom · 2 Sept 2026 · 1 min readAsana buys no-code agent builder StackAI for $75M
Newsroom · 28 May 2026 · 1 min read