Future of Work7w ago

SaaS moats are dissolving as AI agents replace human workflows

Reddit Community

Community Problem

Elevator Pitch

As AI agents become capable of executing tasks directly via APIs, traditional SaaS 'workflow stickiness' is becoming obsolete. Companies must pivot to become the essential 'agent-callable' services, rather than relying on human user engagement.

Full Description

VCs are apparently done funding workflow stickiness. That phrase used to be the magic word in every SaaS pitch deck. Get humans to do their work inside your product, make it hard to leave. That was the playbook for a decade.

But something Jake Saper from Emergence Capital said this week stuck with me. "Pre-Claude, getting humans to do their jobs inside your software was a powerful moat. But if agents are doing the work, who cares about human workflow?"

He pointed to Cursor vs Claude Code as the canary. One owns the developer's workflow, the other just executes the task. Developers are increasingly choosing execution over process.

If an agent can just call your API and get the job done, nobody needs your dashboard. Nobody needs your onboarding flow. Nobody needs your notification system designed to pull them back in. The moat shifts from "keep humans inside your product" to "be the thing the agent calls."

I work at a SaaS company and I keep asking myself whether we're building for humans who use our product or for agents that use our capabilities. Probably needs to be both.

Anyone else thinking about this shift? Curious how other SaaS teams are approaching the "build for agents vs build for humans" question.

Get involved

Discussion

No comments yet. Be the first to share your thoughts.

From the Reddit thread(4 top comments)

  • 35·Reddit commenter·1mo ago

    If they still need your API, then it's a moat. Some SaaS will move more pricing towards API usage if this picks up, but imo the VCs who say that the SaaS don't have a moat anymore are probably heavily invested in AI and they have to push that narrative. AI is not reliable enough to work unsupervised. And people still need to be in the loop and have interfaces to work with. And be building SaaS equivalents in house is not feasible in most cases. The complexity of software, while.some of it lies in the code, most of it lies on infrastructure, processes, cost management, support.

    permalink ↗
  • 28·Reddit commenter·1mo ago

    Rule 1: The opinions of VCs don’t matter Rule 2: Follow rule 1 Be happy

    permalink ↗
  • 16·Reddit commenter·1mo ago

    I think the real shift isn’t “SaaS is dying”, it’s that switching costs are collapsing. A few years ago, vendors won by locking you into their workflow. Now with APIs, LLMs, and better data export tools, migrations are way less scary. That weakens the old “stickiness” moat. A couple practical takeaways: 1. Own a capability, not just a UI. If an agent can call your API to get a result faster than a human clicking through a dashboard, that’s where the value sits. 2. Make your product usable both ways. Humans still need visibility, debugging, and control, agents just need clean, reliable endp…

    permalink ↗
  • 8·Reddit commenter·1mo ago

    In another life I used to sell SaaS. The big thing that people in SaaS managements aren't talking about is that the sales process has changed due to AI. The earlier process for SaaS used to be to land at a company having a specific problem and have discussions with a function leader. Then expand the discussion with other related functions to finally create a combined value proposition. This has changed now. In the new discussions I find that many of my ex-customers dont want SaaS software discussions or meet sales guys because they are solving problems at the department level with workflows…

    permalink ↗

More in Future of Work