The AI vision: From Insight, to Execution, to Skills

I started my career building enterprise software. For most of that time, the software’s job was the same: capture what happened, make it searchable, make it reportable. Systems of record. Then came systems of engagement. Useful, but fundamentally passive. The work itself still sat with the person in front of the screen.

When I joined Outreach, I wasn’t interested in building another system that watched revenue teams work. I wanted to build the system that did the work with them. That’s the journey I want to share in this post, because the Spring release we’ve shipped is the clearest expression yet of where this has been heading. At its core, this shift is about moving toward an agentic AI platform for revenue teams where AI operates directly within workflows to help complete tasks, support execution, and drive measurable outcomes, not just provide insights.

From tool to platform to agentic platform

A few years ago, we made a deliberate bet: Outreach would become an agentic AI platform for revenue teams, driving execution from directly within workflows. Not a sales engagement tool with AI features bolted on. Not a CRM replacement. A platform where AI operates inside the revenue workflow, on behalf of the seller, and drives measurable outcomes.

That's also why, alongside this release, we've become outreach.ai. It's a statement of intent, about what this company builds, who it serves, and the AI-native platform we've built from the ground up to support agentic AI. Every architectural decision has been in service of that bet.  

The early work looked a lot like what you see across the industry now. We started with AI that could answer questions. Now, we’ve evolved and introduced agents that take pre-built actions, draft emails, summarize calls, and flag deals at risk. The goal was always a platform where agents run directly inside workflows.

Getting there required rebuilding the foundation: unified data, shared context, and enterprise-ready governance and observability. Our agents are what makes this kind of foundational AI possible.

At the core is the Outreach agentic platform, built on three layers:

• Revenue data that unifies every signal across every revenue motion: customer engagements and conversations, deal and pipeline movement, first and third-party data, and company knowledge
• Revenue context that makes that data actionable, so Outreach agents and third-party agents can retrieve, reason, and act on it in real time
• Agent harness that blends revenue domain expertise with your data and context, with enterprise-grade governance and guardrails. It acts as the control layer, with shared memory and skills, evals and observability to orchestrate multi-agent workflows.

It is this architecture that makes agentic revenue execution possible at an enterprise scale.

Why insight alone is not enough in revenue AI

I’ve said this in conversations with CROs and CIOs for the last year: most AI in revenue today is still a passenger. It summarizes the call. It scores the deal. It tells the seller what might be happening. Then it hands the real work back.

Revenue teams win because they execute their pipeline consistently, at scale, and across hundreds of deals in motion at any given moment. Insight without execution just produces more and more colorful dashboards, but not necessarily better outcomes.

This is the line the market is crossing right now.

Agentic AI refers to AI systems that operate within workflows to support execution, not just provide insights. Truly agentic AI, prioritizes tasks, drafts outreach, conducts research, and moves deals forward.

The next shift: Hiring agents, not licensing seats

For thirty years, the SaaS model has been the same, licensing software, assigning seats, and eventually deploying teams to do the work inside those seats. The unit of value was a user.

Agentic AI melts that model away. The unit of value is shifting from seats to skills.  

In this model, AI agents act as extensions of your team, helping execute specific tasks within revenue workflows.

When you bring on Outreach, you’re no longer just licensing a tool for your reps. You’re hiring agents. Agents that come with specific skills like prospecting, research, deal progression, follow-up, meeting prep, and the capacity to do that work on your team’s behalf. You distribute those agents the way a sales leader would distribute headcount. Your commercial team gets a different mix than your enterprise team. Your top AEs get more capacity where their books demand it.

Every seller, in this model, gets their own agent that knows their accounts, their workflow, their priorities. Agents that learn what’s working across the organization and raise the performance of the reps who haven’t figured it out yet. The bottom of the distribution catches up to the top. That’s the promise that has kept me up at night for the last two years, and it’s the one we deliver on.

Skills have become the currency. As you expand your investment, you unlock more skills and more capacity. The agents get better as your team gets better, because they’re learning from your team.

The real enterprise requirements

Enterprise AI must be governed, explainable, and secure from day one. Outreach agents operate with the same permissions as the user inside the same guardrails the organization has already defined. Every action is logged. Every output is traceable. Nonnegotiable. The Agent harness is the very reason this architecture can be trusted with the autonomous and active work in the first place.

And because no enterprise is going to build its entire stack on a single vendor, interoperability matters. Our MCP server and open APIs are there so customers can extend what we do, integrate their own data, their own agents, their own workflows, all without leaving the platform.

Enterprise AI must be built for real-world deployment from day one. At a minimum, enterprise AI must be:  

• Governed: operating within defined permissions and organizational guardrails ‍
• Explainable: so teams can understand how outputs are generated  ‍
• Secure: ensuring sensitive data remains protected ‍
• Traceable: with every action logged and auditable

These are the requirements that make autonomous execution possible.

What this Spring release means

Our latest release is the clearest manifestation of this shift toward agentic AI in revenue workflows. Agents act as teammates, supporting research, prioritization, and follow-up within the flow of work, helping scale what top performers do across the entire team.

Sellers spend more time on what matters. Managers scale coaching. Leaders gain a clearer view of what’s working and how to make it repeatable.

The first wave of enterprise AI was experimentation. The next wave is operationalization, moving from insight to execution within workflows. What comes next belongs to the leaders who stop asking what AI can do and start leveraging it to augment their teams, increase critical skillsets, and win more market share.

Frequently asked questions about agentic AI

What is agentic AI in sales?

Agentic AI refers to AI systems that operate within workflows to support execution, not just provide insights.

How is agentic AI different from copilots?

Copilots provide recommendations. Agentic AI helps users take action within the same workflow.

What is an agentic AI platform for revenue teams?

An agentic AI platform is a system where AI is embedded directly into revenue workflows, helping teams move from insight to action more efficiently.