The Agent Verdict
Jason Lemkin ran an experiment: he had Claude, OpenAI, and Gemini grade the top 120 B2B APIs for agent compatibility. Stripe got the only A+. That part was predictable.
What surprised him: when asked what they would use for marketing automation, sales engagement, and conversation intelligence, all three models mocked Marketo, Outreach, and Salesloft. Their reasoning was blunt: an agent will never send an email through these platforms because it can just craft and send a better one itself.
No use to agents. That was the consensus.
Why These Categories Exist
Marketing automation platforms like Marketo exist because a human cannot manually send 50,000 personalised emails. Sales engagement tools like Outreach and Salesloft exist because an SDR cannot track 400 cadences manually. Conversation intelligence platforms like Gong exist because humans need dashboards to extract insights from transcripts.
Agents have none of these constraints. They generate emails in real time, pull context from your CRM, and run cadences natively. The templating layer, the sequence builder, the insight dashboard: these are workarounds for human limitations. Agents do not need workarounds.
What This Looks Like in Practice
At SaaStr, the Salesforce bill went from $12,000 to $22,000 annually. Seats dropped from 10 to 2 plus one agent. Token consumption is up because agents run constantly. Salesforce captures more value per agent than per human.
Their Marketo equivalent? Gone. Their AI VP of Marketing writes campaigns, picks segments, sends emails, and measures results. No Marketo seat to upgrade. No Outreach contract to renew. The category got cut.
Gartner says vendor consolidation is taking 30 to 50% of new AI spending. The first tools to go are pure productivity layers for humans: the ones built around the constraint that a small team cannot do what a large team does. Agents are an even smaller team that does more.
The Shifting Ratio
Even if agents only handle 30% of these workflows by end of 2026, that is 30% of the customer base with no native need for the product category. The real number will be higher.
This is not about agents using the same tools faster. Entire categories of software built to compensate for human constraints have no reason to exist in agent workflows. Project management tools like Jira exist so humans can coordinate work. Agents have memory and context windows: they do not need a Kanban board.
Design tools are next. Why would an agent open Canva to drag assets around when it can generate them directly? The agent does not want a canvas. It wants an output.
What to Watch
The question for every B2B vendor: are you agent-ready? That means clear data models, reliable action models, and task execution paths that work without human-oriented interfaces. Stripe got the A+ because its API is clean and agents can act on it.
Legacy platforms with large installed bases are now scrambling to prove agent compatibility. Newer AI-native competitors are positioning as alternatives in marketing automation, sales engagement, and conversation intelligence.
For sales teams: audit your stack. Which tools exist purely as productivity layers? Which ones compensate for human constraints that agents do not have? Those are the categories where budget will shift first.