With most AI projects failing to deliver measurable value, Dom shares a detailed six-part orchestration framework built from lessons learned across dozens of deployments. From prompt management to hallucination mitigation, this conversation is a reality check — and a roadmap — for RevOps professionals trying to move beyond flashy demos and into scalable, integrated, AI-powered execution.
If you're building workflows that rely on AI — or planning to — this is your blueprint for getting it right the first time.
The session kicks off with a poll: What percentage of AI projects fail to deliver value?
Spoiler alert — the majority of attendees believe it’s well over 50%. And Dom agrees.
“AI isn’t plug-and-play. It’s not magic. What we’re dealing with is still, fundamentally, a fragile system that breaks without the right scaffolding.” – Dom Freschi, Jr., Director of Operations at Openprise
Why is this happening?
Dom’s message is clear: Treat AI like you’d treat any enterprise system. It needs structure, not just strategy.
For more on this topic, check out AI can’t fix what your data is breaking.
To move AI from sandbox experiments to revenue-impacting workflows, Openprise developed a six-part orchestration framework. Internally they call it the “beach ball,” but functionally it’s a practical operating model for RevOps teams.
Check out the full white paper on how AI orchestration makes AI ready for reliable RevOps deployment.
AI is the ultimate “garbage in, garbage out” system. Without clean, enriched, and structured data, even the best prompts and models will return low-quality results.
Examples of failure include:
Your orchestration layer should support:
“We’ve seen AI generate entire emails using completely incorrect titles — promoting someone to CEO. That’s not just embarrassing. It kills trust.” – Dom Freschi, Jr.
For more on this topic, also check out 8 Steps to Creating a Data Utopia.
Prompts are the instruction set that guides the model’s behavior. But unlike code, they’re often untracked, inconsistent and injected haphazardly into tools.
Dom’s best practices:
Dom notes that operators need to understand that every prompt is made up of two components:
Managing both is essential to reduce hallucinations and ensure repeatability.
No single LLM will outperform in every scenario.
OpenAI’s GPT may be best for creativity, Claude excels at long-context summarization, and Google Gemini might win on multilingual tasks.
Key orchestration needs include:
LLMs are designed to sound human — not to be accurate.
The result? High-confidence wrong answers. And in enterprise systems, that’s dangerous.
Dom shares tactics to mitigate risk:
“We’ve seen AI cite research papers that don’t exist — and then other AI tools cite those fake papers. Hallucinations spread unless you put controls in place.” – Dom Freschi, Jr.
Once your AI delivers a clean output — now what?
AI doesn’t magically write to Salesforce, Outreach, or Marketo. You need a translation layer to route outputs into structured fields or trigger downstream workflows.
This is where orchestration platforms shine:
Think of this as the plumbing layer that turns insights into action.
AI costs are notoriously hard to predict. Most operators don’t know what a “token” is — let alone how many tokens their prompts will consume.
Dom recommends:
“It’s not just about if the AI works. It’s about whether it’s worth it — and that means measuring cost and output with the same rigor as any other GTM investment.” – Dom Freschi, Jr.
Throughout the session, Dom returns to a central theme: thoughtful acceleration.
That means:
And perhaps most importantly: Know when AI isn’t the answer.
“Not everything needs AI. Sometimes what you need is a deterministic workflow with no hallucination risk. Know the difference.” – Dom Freschi, Jr.
For more on this topic, also check out a prior digital event Taking the Sexy out of AI: AI for Ops.
Dom and Matthew walk through a common GTM use case: automating outreach. What sounds simple — generate emails with AI — is actually a multi-system problem:
AI without orchestration would fail at nearly every step. But with the framework in place, operators can generate, validate, and route messages with confidence.
Too many companies treat AI as an add-on.
But what Dom makes clear is that AI is a system — and like any system, it requires architecture, governance, and operational oversight.
If you’re serious about using AI to power GTM, don’t start with tools. Start with a framework.
✅ Learn more about Openprise’s orchestration platform
✅ Explore more frameworks on the RevOps Co-op blog
✅ Join the RevOps Co-op community and connect with 18,000+ operators deploying enterprise AI the right way
Don’t let hallucinations, hidden costs, or broken data pipelines derail your AI dreams. With the right orchestration, you can finally deliver the value your executives are asking for — and your GTM systems are begging for.
.jpg)
.jpg)
.jpg)
.jpg)
