We’ve spent 20 years building incredible SaaS applications. Salesforce. HubSpot. Workday. ServiceNow. Thousands of others.
They’ve transformed how businesses operate. They’ve created hundreds of billions in value. They’ve made things possible that were unimaginable in the on-prem era.
And yet.
They’re still “terrible” to actually use.
Not terrible like “doesn’t work.” Terrible like: you need a 6-month implementation. A dedicated admin. A consultant. A certification. A YouTube tutorial just to find the setting you’re looking for.
The average enterprise runs 130+ SaaS applications. Each one has its own:
- Mental model you have to learn
- Navigation paradigm that’s slightly different from the others
- Settings buried 4 clicks deep
- Workflow builder that requires a PhD
- Documentation you need documentation to understand
We’ve normalized this. We shouldn’t have to in the Age of AI.
The Tax on Every Business, On Everyone’s Time
What SaaS complexity actually costs:
The Ops Tax. You don’t just buy Salesforce or Workday or ServiceNow or even HubSpot. You hire a Salesforce admin or agency to deploy HubSpot or even Shopify. Then a Salesforce consultant. Then a RevOps team to connect it to your other 47 tools. A $50K software contract becomes a $300K+ line item when you add the humans required to operate it.
The Time Tax. Your AEs spend 20% of their time fighting with tools instead of selling. Your marketers spend more time in workflow builders than on strategy. Your CS team toggles between 8 tabs to answer one customer question.
The Knowledge Tax. When your ops person leaves, half your institutional knowledge walks out the door. Because it’s not documented—it’s trapped in automations and configurations that nobody else understands.
The Opportunity Tax. The thing you should build never gets built. Because the backlog of “configure the tool we already bought” is infinite.
We’ve just accepted this as the cost of doing business.
What We Actually Wanted
Think back to why we bought these tools in the first place.
We didn’t want “a workflow builder with 47 configuration options.”
We wanted:
- Leads that get followed up with automatically
- Customers who get help instantly
- Data that tells us what’s working
- Processes that run without babysitting
We wanted outcomes. We got interfaces.
For Now—It’s Actually Getting Worse
Here’s the irony: AI was supposed to fix this. Instead, it’s making it worse.
Every SaaS vendor is now shipping AI agents. Salesforce has Agentforce. HubSpot has Agent.AI. ServiceNow has AI Agents. Zendesk, Intercom, Outreach, Gong—everyone’s got one.
And they’re genuinely powerful. These aren’t chatbots. They can reason, take action, handle complex workflows.
But here’s the problem:
If you were already managing 130 SaaS tools, you’re now managing 130 SaaS tools plus 130 AI agents.
Each agent has its own:
- Personality and tone settings
- Knowledge base to configure
- Guardrails to set up
- Escalation rules to define
- Permissions to manage
- Monitoring dashboard to watch
You didn’t eliminate complexity. You doubled it.
And worse—these agents don’t talk to each other.
Your sales AI agent doesn’t know what your support AI agent knows. Your marketing AI agent has no idea what your CS AI agent just learned. Each one is siloed inside its parent application, optimizing locally, blind to the full picture.
So now you need… what? An AI agent orchestration layer? A meta-agent to coordinate your agents? A RevOps person who specializes in agent configuration?
We’ve taken the fragmentation problem of SaaS and added a fragmentation problem of AI on top of it.
This is not the future. This is a transition state. An awkward, expensive, confusing transition state.
The vendors are building what they can ship, not what you actually need.
The AI Opportunity Everyone’s Missing
Here’s what’s wild: we finally have the technology to get what we actually wanted.
Large language models can understand intent. They can translate “I want to follow up with demo no-shows” into actual system actions. They can iterate based on feedback. They can remember context.
And yet.
What is the SaaS industry doing with AI?
Adding copilots to the existing UI.
Salesforce built Agentforce—AI that helps you navigate… Salesforce. HubSpot added AI assistants that help you… use HubSpot faster. Every SaaS company is bolting an AI helper onto their existing complexity.
This is like adding a GPS to a horse and buggy. Helpful, sure. But missing the point entirely.
The Interface Should Disappear
Here’s the radical idea: What if you could just talk to your software and have it do what you want?
Not “talk to an AI that helps you click the right buttons.”
Actually talk. In plain language. And have it happen.
“Set up a sequence for prospects who no-show their demo. Three touches over a week. Friendly tone. Stop if they reply.”
Done.
“Show me which deals are at risk of slipping this quarter.”
Here’s the list, with reasons why.
“When a customer’s usage drops below 50% of their plan, alert their CSM and draft a check-in email.”
Running.
No workflow builder. No settings screen. No certification required.
The configuration layer becomes a conversation.
Old SaaS Needs to Die. It Needs to Be Work Like Claude or ChatGPT on Steroids
1. Ops becomes accessible to everyone.
Today, building a lead scoring model requires a RevOps specialist. Tomorrow, a founder can say “flag leads that look like our best customers” and it happens. The bottleneck shifts from “who knows the tool” to “who knows what outcomes matter.”
2. Implementation timelines collapse.
Six-month Salesforce implementation? What if you could describe your sales process in conversation and have Claude configure it in an afternoon? The time-to-value equation completely changes.
3. The ops hire gets 10x more leveraged.
Your RevOps person stops spending 80% of their time in workflow builders and starts spending it on strategy. They describe what they want, AI builds it, they refine through conversation.
4. Institutional knowledge becomes conversational.
Instead of knowledge trapped in automations, it’s encoded in natural language. “Why do we do it this way?” becomes a question you can actually ask—and get an answer to.
What’s Actually Required
This isn’t science fiction. The intelligence layer exists today. Claude, GPT-4, and others can absolutely understand intent and translate it to action.
What’s missing:
Integrations with write access. Most AI today can read your systems. Few can actually take action in them. The companies that build robust, secure, bi-directional integrations win.
Trust and guardrails. You need AI that can act autonomously but within boundaries. “Send up to 100 emails without approval. Flag anything above that.” This is a product design problem, not a technology problem.
Memory and context. The AI needs to know your business—your ICP, your processes, your preferences. Not starting from scratch every conversation. Persistent, compounding understanding.
Someone willing to rethink the UI paradigm. This is the hardest part. Existing SaaS vendors are trapped by their current interfaces. They can’t cannibalize the product their customers have already learned.
The Uncomfortable Math for SaaS Vendors
Here’s what keeps SaaS execs up at night—or should:
Today you’re charging $50K/year for software + the customer spends $200K/year on humans to operate it.
Tomorrow, Claude with integrations costs $3K/month and eliminates 80% of the operational burden.
Your customer isn’t paying for your software. They’re paying for outcomes. If they can get the outcomes through conversation with an AI, your workflow builder isn’t a feature—it’s a liability.
The new entrant doesn’t need to rebuild Salesforce. They need to build a Claude integration layer that makes Salesforce’s UI irrelevant.
Who Wins in This World?
Horizontal AI platforms that nail integrations. If Anthropic or OpenAI build robust connectors to the major business systems, they become the operating layer. The SaaS apps become dumb databases underneath.
Vertical AI-native startups. New entrants who build for conversation-first from day one. No legacy UI to protect. No existing mental model to preserve.
The SaaS incumbents who cannibalize themselves. Rare, but possible. The vendor who says “forget our UI, just talk to us” and actually means it.
Losers: Everyone defending their existing interface. Adding AI to help people use your complex UI is a transition strategy, not an end state.
What Should You Do About It?
If you’re a B2B founder:
Start asking what your product looks like if the UI disappeared entirely. If customers could only interact through conversation, what would need to be true? That’s your roadmap.
If you’re a buyer:
Stop evaluating tools on features. Start evaluating on “how easy is it to get outcomes?” The tool with fewer features but natural language interaction might be worth 10x more than the tool with a 47-page feature comparison.
If you’re in ops:
Your job is about to get much more interesting. Less time in configuration screens, more time on strategy. Start thinking about what you’d build if you had no technical constraints.
If you’re an investor:
The current wave of “AI copilots for existing SaaS” is a bridge. The real opportunity is AI-native that makes the old interfaces obsolete. Different bet, much bigger outcome.
SaaS Is Now Painful
We’ve spent two decades training ourselves to use software on the software’s terms.
Click here. Navigate there. Learn this workflow builder. Get certified in that platform.
It was always a compromise. We just didn’t have a better option.
Now we do.
The question isn’t whether conversational interfaces replace configuration UIs. It’s when, and who builds it.
SaaS isn’t going away. But the SaaS experience—the clicks, the complexity, the consultants—that’s on borrowed time.
It’s time for software that just does what you ask.
It’s time to be able to talk to it.
