AI Agents Are Rejecting Marketo, Outreach, and Salesloft

AI models choose their own tools—and reject human productivity software

When three leading AI models—Claude, OpenAI, and Gemini—were asked to evaluate 120 APIs and select the best tools for agent workflows, they reached a surprising consensus: marketing automation platforms like Marketo, sales engagement tools like Outreach and Salesloft, and conversation intelligence software have no place in agent-driven work.

The reasoning was direct. An agent doesn't need Outreach to send an email sequence because it can craft and send better, personalized emails itself. It doesn't need Marketo's templating system because it can generate 50,000 unique messages in real time. The productivity layer these tools provide to humans is exactly what agents replace.

Stripe earned the only A+ grade in the evaluation, while human productivity tools were dismissed outright.

Why it matters

This isn't about AI features being added to existing software—it's about entire product categories becoming obsolete as agents handle workflows directly. Even if agents manage just 30% of these tasks by late 2026, that represents 30% of the customer base with no native need for the product. The shift threatens billions in enterprise software revenue built on human constraints that no longer apply to AI workflows.

The categories agents bypass

Marketing automation platforms exist because humans can't manually write thousands of personalized emails. Agents can generate each message fresh, pulling context from CRM data, recent prospect activity, and product updates. The templating and scheduling layer that defines Marketo, HubSpot's enterprise tier, and Eloqua becomes redundant.

Sales engagement tools like Outreach and Salesloft help SDRs manage follow-up sequences across hundreds of leads. Agents run these cadences natively and generate each touchpoint based on real-time prospect behavior, eliminating the need for sequence builders.

Conversation intelligence platforms record and analyze sales calls to surface insights for human action. Agents ingest transcripts directly and act on them without needing a dashboard translation layer.

Project management tools—Jira, Confluence, Asana, Monday—coordinate work across human teams. Agents don't need Kanban boards or wikis; they have memory and context windows that eliminate the coordination overhead these tools were built to solve.

What survives the agent transition

Not all B2B software faces obsolescence. Systems of record that hold customer, financial, or operational data remain essential—agents need to read and write to Salesforce, Workday, NetSuite, and Stripe. These become infrastructure for agents rather than tools for humans.

Communication infrastructure like Twilio and SendGrid survives because agents need delivery mechanisms, even as they eliminate the templating layers above them. Compute infrastructure, AI platforms, code repositories, and vertical software with deep domain integration also remain relevant.

The pattern is clear: infrastructure survives, human productivity layers get cut.

Real-world evidence

At SaaStr, Salesforce costs increased 80%—from $12,000 to $22,000 annually—as agent usage drove higher API consumption, even as human seats dropped from 10 to just over 2. Meanwhile, the company eliminated its marketing automation equivalent entirely. The AI VP of Marketing writes campaigns, segments audiences, sends emails, and measures results without needing a Marketo-style platform.

This vendor consolidation is capturing 30-50% of new AI spending, according to Gartner. The first casualties are tools whose entire value proposition is "let a small team do what a larger team would do." Agents represent an even smaller team doing even more.

The strategic question

For companies in affected categories, adding AI features to products agents don't need misses the point. The real question is what the business becomes as agent-driven work grows quarterly and the addressable human workflow shrinks 20% annually. Some companies may pivot to becoming infrastructure for agents—requiring complete product, pricing, and go-to-market rewrites. Others face managing a durable but declining human user base while searching for a second act.

Public markets are already pricing in this shift. Atlassian and Monday have declined as investors recognize project management lacks native agent demand, while Twilio and Cloudflare have outperformed as agent infrastructure plays.

These details were first reported by Jason Lemkin at SaaStr, based on direct API evaluation experiments with multiple foundation models.