The Agentic AI Shift: What I Learned Building Revenue at an AI Platform

When I Joined SteppingStones.ai, Nobody Had a Budget Line for This

When I stepped into the Chief Revenue Officer role at SteppingStones.ai, agentic AI was still conceptually foreign to most enterprise procurement teams. It wasn't that buyers were skeptical of AI broadly — by that point, everyone had a ChatGPT story and half of them had a pilot underway somewhere. The problem was more fundamental: agentic AI didn't fit cleanly into any category they already understood. It wasn't automation in the legacy RPA sense. It wasn't a chatbot. It wasn't a copilot bolted onto an existing workflow. It was something genuinely new — and in enterprise sales, genuinely new is simultaneously the most exciting and most dangerous position to occupy.

What I learned over those months — earning through deals won, deals lost, and deals that stalled for reasons that had nothing to do with the technology — reshaped how I think about selling at the frontier. It also directly informed how I'm building the commercial architecture at HedgeNova, where we're navigating a structurally similar challenge in an even more regulated context. So I want to share that thinking in some depth, because I think the lessons generalize well beyond any single company or vertical.

Selling the Unknown: Why the Traditional SaaS Playbook Breaks Down

The conventional SaaS revenue motion is built on a set of assumptions that, taken together, form a fairly reliable foundation: the buyer has a named problem, they're actively evaluating vendors, they have a budget category that maps roughly to your solution, and your job is to win the evaluation on product-market fit, price, and trust. That's a gross simplification, but the underlying logic holds — you are competing within a recognized market.

Agentic AI, at least in the 2023–2024 window when I was running revenue at SteppingStones, didn't fit that model. We were often the first conversation a buyer had ever had about autonomous agents operating inside their enterprise environment. There was no RFP waiting for us. There was no budget that said "agentic workflow automation." There was curiosity, sometimes excitement, and a healthy dose of institutional caution — which is entirely rational when you're asking an organization to let software agents take actions, make decisions, and interact with live systems on their behalf.

This meant the sales motion had to accomplish something that traditional SaaS sales doesn't: we had to create the category in the buyer's mind before we could win within it. That's a fundamentally different job. It requires a different kind of discovery, a different narrative structure, and a very different relationship between the sales team and the technical team.

What Actually Worked: Anchoring on Outcomes, Not Architecture

The breakthrough in our messaging came when we stopped leading with what the technology was and started leading with what it eliminated. Enterprise buyers — whether they're in financial services, healthcare, or enterprise SaaS operations — are not buying technology for its own sake. They're buying relief from a specific, often painful operational reality.

For us, that meant walking into discovery with a pointed question: What does your team do today that requires human judgment but is largely repetitive, high-volume, and expensive to staff for? That question unlocked more productive conversations than any product demo. It surfaced the real pain — the analyst team manually reconciling data across systems, the ops team hand-routing exceptions, the compliance function spending thirty hours a week on tasks that followed a clear decision tree but still required a warm body to execute.

Once you name the operational pain precisely, you can position the agent as the solution to that specific pain — not as a general-purpose AI capability in search of a use case.

This outcome-first framing did something else important: it changed the internal champion dynamic. When we anchored in technology, we were talking to IT or innovation teams who found it interesting but had limited budget authority. When we anchored in hours eliminated, error rates reduced, and cost per transaction lowered, we were suddenly in conversations with CFOs, COOs, and heads of operations — people who control budget and have direct accountability for the numbers we were discussing.

The Demo as Proof of Outcome, Not Feature Tour

One of the most consequential tactical shifts we made was in how we structured product demonstrations. Early in my tenure, demos followed a familiar pattern: here's the platform, here's the interface, here's the configuration layer, here are the integrations. Feature-by-feature. Buyers would nod politely and ask when they could see pricing.

We rebuilt the demo from the ground up around a single principle: show the agent completing a real workflow that the specific buyer cares about, in real time, with real stakes. We did the work upfront in discovery to identify a workflow that was both meaningful to that buyer and well-suited to demonstration. Then we built a working proof of concept — not a sandbox with fake data, but as close to their actual environment as we could responsibly construct — and let the agent run it live.

The effect on deal velocity was significant. When a VP of Operations watches an agent intake a document, extract the relevant fields, cross-reference three systems, flag an anomaly, route for human review, and log the action — all in under two minutes, without a human touching it — the conversation shifts immediately. You stop explaining what's possible and start negotiating what deployment looks like. The demo stops being a feature tour and becomes a proof of outcome. That's when sales cycles compress.

The Trust Infrastructure Problem Nobody Talks About

Here's what I didn't fully anticipate coming in: the biggest obstacle to closing agentic AI deals in enterprise is rarely the technology and rarely the price. It's trust infrastructure — specifically, the buyer's ability to answer the question, how do I know what the agent did, why it did it, and whether I can rely on it?

This is where many agentic AI vendors, in my observation, underinvest. They build impressive agent capabilities and then treat auditability, explainability, and human-in-the-loop controls as secondary features. In enterprise sales, especially in regulated industries, that ordering is backwards. The audit trail is the product for the buyer's legal, compliance, and risk teams — who will inevitably have a seat at the table before any contract closes.

• Explainability: Can you show the buyer exactly what decision logic the agent followed, and why?
• Auditability: Is every agent action logged in a format that satisfies compliance and legal review?
• Override and escalation: Is it unambiguous how and when a human takes back control?
• Failure behavior: When the agent encounters something outside its operating parameters, does it fail gracefully or create a downstream mess?

Getting fluent on these questions — and being able to answer them credibly — became as important as the product demo itself. The deals we lost almost always involved a security or compliance review that exposed gaps in how we talked about these dimensions, not gaps in what the product could actually do.

What I'm Carrying Forward Into HedgeNova

Building HedgeNova as CEO — where we're deploying agentic AI in the context of alternative investment research and risk intelligence — I'm applying every one of these lessons from day one. The buyer profile is different (hedge funds, family offices, institutional allocators), but the structural challenge is identical: we are operating at the edge of what buyers know to expect, in a domain where trust and auditability are non-negotiable, and where the fastest path to revenue is a live demonstration of a specific outcome the buyer already knows they need.

The agentic AI shift is real, and it's accelerating. But the companies that will win the enterprise aren't necessarily the ones with the most capable agents. They're the ones that understand how to sell a category that doesn't yet have a procurement code — and who do the hard work of making trust legible to the buyer before they're ever asked for it.

That's the playbook I'll keep refining. And I'll keep writing about it here as it evolves.