Building AI Agents for RevOps: From Data Hygiene to Deal Insights

AI is changing the way revenue operations teams think, act, and scale—but without clean data and cross-system orchestration, even the smartest models fall flat. In this session, Matthew Volm, CEO and Founder of RevOps Co-op and Eventful, sits down with Chris Ferraro, Brian Knaupe, Sandeep Kumar, and Josh Azorsky from Celigo to unpack one of the most practical questions facing modern RevOps teams:

‍

How do you build AI agents that actually work—delivering cleaner data, faster insights, and real operational leverage across GTM systems?

‍

What follows is a step-by-step breakdown of how the Celigo team is operationalizing AI across their revenue engine, from data hygiene and automation to human-in-the-loop orchestration—and what every RevOps leader can learn from their journey.

‍

For more on this topic, check out the Celigo blog post Beyond AI hype: Building AI agents that actually deliver for RevOps.

‍

AI in RevOps: The Hype vs. the Hard Reality

AI has become a boardroom buzzword, but implementation remains elusive. While 74% of CEOs say AI will most impact their industry, a stunning 88% of AI pilots never reach production. The reason? Bad data and fragmented systems.

‍

As Brian Knaupe, Celigo’s Head of Go-to-Market Operations, explained:

‍

“Turning on an AI tool isn’t a strategy. The easiest thing to do is flip a switch—but in RevOps, that never solves the root problem. We’re only limited by our creativity, but our creativity can’t overcome bad data.”

‍

Most revenue organizations suffer from tool sprawl and inconsistent data schemas across CRM, ERP, and CS platforms. Even the best large language models can’t reason effectively without context and structure. That’s why Knaupe’s team shifted focus from “building AI use cases” to “engineering an AI-ready data foundation.”

‍

“Every new tool promises automation. But if your data is inconsistent or siloed, you’re just automating noise.” – Brian Knaupe, Head of GTM Operations at Celigo

‍

From AI Pilots to Production: Building on Solid Ground

Josh Azorsky, Solutions Consultant at Celigo, drew a sharp line between experimentation and operationalization.

‍

While pilots prove potential, production AI requires three pillars:

1. Connectivity – AI must reach every relevant system—CRM, ERP, analytics, support—without manual stitching or one-off scripts.
2. AI-Ready Data – The data must be clean, current, and contextualized, structured for accuracy and interpretability.
3. Context – Models need business logic: how deals progress, what triggers matter, what “good” looks like.

‍

When these layers align, RevOps teams can embed intelligence directly into daily workflows rather than running disjointed “AI experiments.”

‍

“AI without integration is like running a race blindfolded—you’re moving fast, but not in the right direction.” – Josh Azorsky, Solutions Consultant at Celigo

‍

The goal isn’t to bolt on AI; it’s to operationalize it into the core revenue infrastructure, turning insights into automatic, actionable outcomes.

‍

Celigo’s Framework: Signal → Context → Action

At the heart of Celigo’s approach is a deceptively simple framework: Signal → Context → Action.

‍

This three-step model transforms abstract AI ambitions into executable automations:

1. Define the Signal Identify the triggering event or insight. Examples:
   • A deal that slipped past close date
   • A customer using specific competitive keywords on a call
   • A website visitor matching ICP criteria
2. Add Context
3. Pair the signal with additional data—account owner, deal stage, ARR band, or region—to ensure precision.
4. Define the Action Determine what happens next:
   • Post an alert in Slack
   • Create a Salesforce task
   • Update an account field
   • Route an escalation to CS

‍

This model gives RevOps a common language for AI orchestration, ensuring that automation happens intelligently—not recklessly.

‍

“Write it down. Every great automation starts as a documented process. Your documentation becomes your AI’s playbook.” – Brian Knaupe

‍

Case Study: Automating Deal Insights from Gong

Before AI augmentation, Celigo’s reps were manually updating Salesforce after every Gong-recorded call—painstakingly entering applications mentioned, intent signals, and next steps. The result:

• Incomplete records
• Inconsistent data
• Slower follow-up

‍

Celigo replaced that manual process with an AI data-augmentation agent:

1. AI reads call transcripts from Gong every five minutes.
2. It extracts structured data—products mentioned, deal type, buyer intent.
3. It sends the results to Slack for human validation (a “human-in-the-loop” step).
4. Once approved, Salesforce updates automatically—within minutes, not hours.

‍

This saved hours of manual entry while improving CRM accuracy and forecast precision. Over time, as confidence in AI accuracy increased, the team removed the human review step—transitioning from assisted automation to full autonomy.

‍

“Every automation begins with humans in the loop. Once we trust the output, we move to autonomy—but guardrails always stay in place.” – Sandeep Kumar, Head of AI Strategy at Celigo

‍

The Role of Human-in-the-Loop Governance

AI success isn’t about eliminating people—it’s about augmenting human judgment. The Celigo team begins every AI workflow with validation checkpoints, gathering user feedback via Slack workflows and Asana integrations before granting full autonomy.

‍

For example:

• Early-stage agents: Require sales reps to confirm AI-generated insights before syncing with CRM.
• Mature agents: Run autonomously but still flag anomalies or high-value changes for manual review.
• Exception thresholds: Expense approvals over certain amounts still trigger human oversight.

‍

This creates an adaptive governance model—balancing automation with accountability.

‍

Integration: The Unsung Hero of AI Orchestration

A key discussion point was whether to use native app integrations (e.g., Gong → Salesforce) or orchestrate through an integration platform (iPaaS) like Celigo.

‍

Their conclusion? Native integrations solve small problems; iPaaS solves systemic ones.

‍

Native integrations are simple but rigid—offering limited error handling and zero visibility into failures.

‍

You might also want to check out OpenAI App SDK and AgentKit: Why integration still matters.

‍

An iPaaS layer, on the other hand:

• Centralizes data pipelines across CRM, ERP, CS, and analytics.
• Offers full control and monitoring over API calls, retries, and data lineage.
• Allows embedding AI workflows directly into the data orchestration layer.
• Feeds a unified warehouse (e.g., Snowflake) for downstream analysis.

‍

“When you use an iPaaS, you’re not just integrating systems—you’re future-proofing your data strategy.” – Sandeep Kumar, Head of AI Strategy at Celigo

‍

Knaupe added:

‍

“If you’re in RevOps and not using an iPaaS yet, it’s not a question of if you will—it’s when. It’s the foundation every AI workflow needs.”

‍

For more on this topic, check out Scaling smart: A practical guide to integration technologies.

‍

Tangible Results: Real Impact Across the Revenue Engine

Celigo’s internal deployment of AI-driven automation has generated measurable business outcomes:

• 30% faster onboarding – Using an AI-powered Knowledge Bot that references internal documentation and FAQs.
• 10% faster support ticket resolution – Through automated triage and Slack notifications.
• 27% increase in upsell opportunities – Enabled by unifying customer data for more contextual account reviews.

‍

Beyond metrics, the qualitative impact was equally significant: smoother collaboration, faster decision-making, and cleaner system-of-record data that fuels both human and machine intelligence.

‍

“When data is unified, the entire GTM motion moves faster—from deal creation to renewal forecasting.” – Chris Ferraro, Product Marketing at Celigo

‍

For more on this topic, check out AI Agents: Powerful, Promising… but Not for Every Use Case.

‍

Lessons Learned: The RevOps AI Playbook

After four years of embedding AI across their own go-to-market motion, Celigo’s team offered five core principles every RevOps org should apply:

1. Start Small, Scale Fast: Begin with one workflow—like deal hygiene or data enrichment—and iterate once accuracy is proven.‍
2. Document Relentlessly: Every process map, playbook, or SOP is not just operational overhead—it’s the raw material for AI agents.‍
3. Think Integration-First: The foundation of AI success isn’t model quality—it’s system connectivity and clean data movement.‍
4. Govern with Feedback Loops: Keep humans in the loop early; use that feedback to refine prompts, confidence thresholds, and triggers.‍
5. Measure Business Impact: Track metrics like response time, data completeness, and cycle acceleration to quantify ROI.

‍

“Knowledge is power—but integrated knowledge is leverage.” – Brian Knaupe, Celigo

‍

For a specific example of a large enterprise using automation to succeed, check out this case study Celigo + Walmart: Automating Multichannel Fulfillment Ahead of Peak Season.

‍

The Future of AI in RevOps

The Celigo team anticipates a shift from “AI assistants” to embedded AI agents that live inside workflows. These agents won’t replace RevOps—they’ll extend operational capacity, helping teams diagnose friction, model outcomes, and act in real time.

‍

But this vision hinges on one universal truth: you can’t automate chaos. Clean, connected, and contextualized data is the currency of tomorrow’s revenue engine.

‍

As AI becomes the nervous system of go-to-market strategy, integration becomes its circulatory system—and RevOps becomes the architect of both.

‍

Want More Like This?

Check out the RevOps Co-op Blog and Celigo Blog for more deep dives on AI orchestration, data governance, and operational frameworks, and join our community of 18,000+ operators building the next generation of RevOps systems—together.

‍