Oscar Vail has spent years at the intersection of emerging technology and business efficiency, navigating the shifts that define how companies grow. As the Go-To-Market landscape undergoes a fundamental transformation, Oscar is currently focused on how the Model Context Protocol is moving AI from a novelty into a functional operational layer. In this conversation, we explore the transition from simple content generation to complex, agentic workflows that leverage deep buyer intelligence. We delve into the specific tools leading this charge—from technical-buyer specialists like Onfire to CRM giants like HubSpot—and identify the strategic hurdles revenue teams must clear to capture real business value in an era where data quality and workflow control are the new competitive advantages.
When shifting from simple content generation to active workflow layers, what operational changes are necessary for a modern revenue team?
The shift requires moving past the “speed for speed’s sake” mindset that dominated the first stage of AI adoption. Early on, teams were primarily using AI to draft emails or summarize calls, which certainly improved individual velocity but didn’t actually solve the core revenue problem of disconnected systems. To move into active workflow layers, you have to address the reality that 51% of sales leaders feel disconnected systems are actively slowing down their initiatives, even though 87% of sales organizations have already adopted some form of AI. Operationally, this means you can no longer treat AI as a standalone drafting tool; it must become a context-aware layer that understands your CRM, your specific ICP configuration, and real-time market signals. You are moving from a world of generic prompts to a world of governed actions where the AI doesn’t just suggest a message but knows exactly which account to prioritize based on a live technical signal. This transition demands a much tighter partnership between Sales and RevOps to ensure that the “active” part of the workflow is built on a foundation of high-quality data and approved logic, rather than letting an agent hallucinate its way through a prospect’s technical stack.
How does the Model Context Protocol specifically bridge the gap between static AI prompts and the dynamic, often messy reality of a company’s CRM?
The beauty of the Model Context Protocol, or MCP, is that it provides a standardized bridge for AI agents to interact with business data without needing to rebuild the underlying model every time your data changes. Traditionally, a CRM integration was a rigid, predefined connection between two systems, but MCP allows an agent to request context or take actions in real-time based on the specific needs of a conversation. For instance, when you look at how HubSpot utilizes a remote MCP server, it allows an AI tool to have permission-aware access to contacts, deals, and engagement history. This means the agent isn’t just working from a static memory; it is reaching into the system of record to see the latest meeting notes or lifecycle stage before it suggests a next step. It solves the “context gap” by giving the agent a way to “see” the truth of the customer relationship as it exists right now. It turns the AI from a writer into an operator that can read from and, with the right permissions, write back to the CRM, logging calls and updating tasks as the workflow progresses.
For companies selling to technical audiences—like DevOps or cybersecurity—why is general firmographic data no longer enough to drive a successful GTM strategy?
In technical sales, the evaluation process often begins in the “public footprint” of engineering and technology decision-makers long before a lead ever shows up in a CRM. If you are relying solely on company size or industry tags, you are missing the nuanced signals that actually indicate a technical buyer’s intent. This is where a tool like Onfire becomes essential because its Account Intelligence Graph connects customer data with the public activities of more than 50 million engineering and technology decision-makers. You need to know which people are showing relevant technical interest in specific infrastructure or cloud solutions. A general prospecting database might tell you someone is a “VP of Engineering,” but it won’t tell you they are currently grappling with a specific DevOps bottleneck. By using AI to refine this technical intent into actionable context, teams can prioritize accounts based on actual technical fit and signal discovery, moving from broad-stroke marketing to precise, ICP-aware execution that resonates with a buyer who has likely already done a significant amount of research.
We see a lot of “experimentation fatigue” in marketing today; what distinguishes the tools that actually scale from those that remain stuck in the pilot phase?
The distinction usually comes down to the ability to connect insight directly to execution. McKinsey found that while 90% of CMOs are experimenting with AI, fewer than 10% have actually scaled it to capture real value across their workflows. The tools that make the leap are the ones that don’t just provide a research report but allow a rep to stay within their workflow to finish the job. Take Apollo as an example; it doesn’t just help you find a company; it allows you to enroll a prospect into an approved outreach sequence directly from an AI assistant like Claude or ChatGPT. That compression of steps—from discovery to enrichment to enrollment—is what allows a process to scale. If a rep has to leave their AI tool, log into a separate database, export a CSV, and then manually upload it to a sequence tool, the “intelligence” of the AI is lost in the friction of the manual labor. Scaling happens when the “action depth” is high, meaning the tool can actually complete the revenue-generating task rather than just talking about it.
Revenue leaders often struggle with “pipeline truth” versus “pipeline volume.” How can activity intelligence change the way a CRO evaluates deal health?
This is a critical pain point because a CRM can often be a “lagging indicator” or, worse, a reflection of a sales rep’s optimism rather than reality. A deal might look healthy on paper with a high dollar value and a close date in the current quarter, but the activity history might tell a different story. Platforms like Backstory are designed to solve this by capturing the “truth” found in emails, call transcripts, and meeting cadences. If an AI agent can access this activity intelligence through MCP, it can flag when a champion has gone quiet or when recent meeting notes show a lack of urgency that contradicts the forecast. Instead of a CRO asking “How is the pipeline looking?” they can ask an agent “Which of our top five deals has had a drop-off in stakeholder engagement over the last 14 days?” This shifts the focus from pipeline volume to pipeline reality, allowing managers to intervene in at-risk deals with evidence-based coaching rather than just guessing.
What role does conversation intelligence play when it is no longer locked in a separate dashboard but available to external AI agents?
When conversation data is siloed, it’s only useful to the people who have the time to go back and listen to the recordings or read the summaries. But when a platform like Gong supports MCP as both a client and a server, that intelligence becomes a fluid asset for the entire GTM team. As a server, Gong allows external agents in tools like Microsoft Copilot to pull in account summaries, objections raised in the last demo, or specific competitor mentions. This means that if a marketing manager is building a campaign, they can ask their AI tool to summarize the top three objections heard by the sales team this month, and the AI can reach into Gong to get that answer. It turns every discovery call and renewal conversation into a live data feed that informs every other part of the GTM motion, ensuring that the “voice of the customer” is actually used to drive next best actions across the organization.
As teams move toward using AI agents for prospecting, what are the most common governance mistakes that lead to “messy” data or damaged brand reputation?
The most frequent mistake is treating MCP as a feature you just “turn on” rather than an architecture you have to govern. We see teams giving agents access to messy CRM data and then being surprised when the agent generates outreach based on outdated information or duplicate records. Another major pitfall is automating an unclear ICP logic; if you haven’t defined exactly what a “good” lead looks like, the AI will just find you more of the “bad” leads faster. There is also the danger of letting agents update CRM records without any human-in-the-loop oversight. Without proper permission scopes and audit trails, you can end up with a database full of “AI-enhanced” garbage. Security and RevOps ownership are non-negotiable here. You need to define what an agent can read, what it can write, and which actions require a human to hit “approve” before an email goes out or a deal stage changes.
How should a RevOps leader evaluate which of these seven tools—from Clay to ZoomInfo—actually fits their specific sales motion versus just adding more “tool sprawl”?
The evaluation shouldn’t start with the tool’s feature list; it should start with the specific revenue workflow you are trying to fix. If your problem is that your reps spend four hours a day researching technical buyers, then Onfire is likely your lead. If your problem is that you have great data but no one is enrolling prospects into sequences fast enough, Apollo’s execution focus might be the answer. For teams with complex, custom data requirements and creative outbound needs, Clay’s enrichment and scoring logic is incredibly powerful. My advice is always to choose one measurable workflow—like lead qualification or account research—and test two or three tools against the exact same data and success criteria. The platform that provides the most reliable context and controlled action within your actual sales motion is the winner. More features do not equal more value; reliability and “CRM fit” do.
What is your forecast for GTM intelligence?
I believe that by late 2026, the concept of a “standalone” GTM tool will feel as dated as a paper rolodex. We are moving toward a fully “agentic” GTM stack where the CRM, the enrichment layer, and the conversation intelligence platform are all essentially “servers” providing context to a centralized AI orchestrator. The winners in this space won’t just be the companies with the biggest databases, but the ones with the most accessible and high-quality MCP integrations. We will see a massive shift where the quality of your “context layer” becomes your primary competitive advantage. If your AI agent knows more about the technical challenges of your buyer than your competitor’s agent does, you will win the deal before the competitor even knows there is an opportunity. The barrier to entry for GTM success is rising, and it’s being built on a foundation of real-time, permission-aware intelligence.
