Everyday AI Podcast – From Automation to Agents: Why Weak Data Makes AI Guess

Date: December 11, 2025
Host: Jordan Wilson
Guest: Edmoski, Chief Product & Technology Officer at Boomi

Episode Overview

In this episode, Jordan Wilson interviews Edmoski (“Ed”), Chief Product and Technology Officer at Boomi, to explore the transition from traditional automation to AI agents in business workflows. The discussion centers on why high-quality data is more critical than ever, how agentic (AI-powered) systems differ fundamentally from deterministic scripts, and the practical steps business leaders should take to harness these new capabilities while maintaining trust and ROI.

Key Discussion Points & Insights

1. The Trend: From Automation to AI Agents

• Agentification is everywhere:
  As we move into 2026, everything that could be automated is becoming “agentified”—standard processes, features, and even classic automations are being reimagined as autonomous agents (00:16, Jordan).

• Old automations vs. new agents:
  Traditional automations fail when data or configuration is off—no output is produced. In contrast, AI agents will still produce an output, even if it's based on flawed data—which can lead to unexpected or even made-up results (00:16, Jordan).

2. The Current State of Enterprise Automation

• Hybrid approach:
  “Being a little bit of both actually....We will see both deterministic and non-deterministic or agentic workflows in the enterprise forever.” (04:08, Ed)
  Some processes require strict, rule-based automation; others benefit from agentic flexibility. Many organizations are integrating both for effective business outcomes.

• Significant cost savings:
  Real business impact is emerging: team-based agentic workflows save companies millions of dollars, moving beyond pilot projects to real automation at scale (04:08, Ed).

3. Pros and Cons of Moving to AI Agents

Benefits

• Reduced manual intervention:
  “You can take your hand off the wheel a little bit....You can get a lot more out of your workforce....let the humans do more valuable work.” (06:01, Ed)
• Adaptive, less brittle workflows:
  Agents can read and dynamically adapt to business policies instead of relying on rigid, brittle rule sets (07:59, Ed).

Challenges

• Data dependency:
  “You have to build these agents...on a strong foundation of data. Bad data in is worse data out.” (06:01, Ed)
  Data quality is no longer a “nice to have”—agents will act on whatever data they’re given, amplifying errors or inconsistencies.

• Need for data governance and maintenance:
  Internal policies, organizational charts, and resources must be maintained as meticulously as the agentic systems that use them (09:38, Ed).

4. Real-World Example: Expense Reports

• Old way:
  Rigid process, many manual inputs, rule enforcement via hard-coded scripts, prone to errors and workarounds.
• Agentic way:
  Conversational input for users, dynamic rule retrieval/interpretation by agents, reduced friction for managers/approvers, and learning from approval patterns to know when to involve humans (10:10–13:03, Ed).
• Data still matters:
  Even the smartest agent can’t fix or guess its way around missing or incorrect data (e.g., department assignments, policy rules).

Quote:
“Those are all data quality things underneath...in the last number of years in an enterprise, haven’t really paid much attention to because humans have been a crutch...That’s why data quality is becoming more and more critical for these agentic flows.”
— Ed, (12:14)

5. The Vulnerability of AI Agents to Bad Data

• Agents may guess or “hallucinate”:
  LLMs can leap to conclusions or act on faulty assumptions when faced with inadequate, inconsistent, or ambiguous data (13:03–14:35, Jordan).
• Importance of transparency:
  Observing the “chain of thought” is vital for advanced users, but ordinary business leaders must ensure sound data and clear decision records.

6. Safeguarding Agentic Workflows

• Governance “sandwich”:
  Foundational data quality + agent security controls + “agent control towers” monitoring outputs = safer agentic automation (14:35, Ed).
• Multi-agent orchestration raises the stakes:
  As agents collaborate and pass tasks amongst themselves, even small data inaccuracies propagate quickly, making aggregate observability—across platforms—essential (17:42–19:49, Ed).

Quote:
“When you have multi-ecosystem orchestration, there’s the data that resides within the system that the agent is on top of....If that data is not in sync and of quality on both sides, you can have disaster.”
— Ed, (18:39)

7. Advice for Leaders with “Iffy” Data

• Cross-functional collaboration is key:
  “Pull your AI innovators together with your business owners, those that own the data and your AI innovators and get them talking and working together....Focus on the data sets that you need for automating your business or the outcomes that you're looking for and really double down on those.” (20:50, Ed)
• Start small, iterate, don’t try to boil the ocean:
  Instead of big-bang data overhauls, begin with focused, high-impact datasets and expand incrementally.

8. Measuring ROI: Flip the Script

• Don’t start with technology—start with outcomes:
  “Their outcome...is clean data. Well, nobody. Clean data by itself doesn’t mean anything... Flip the script. Start with discrete projects where you want to define your ROI....Then apply the technology and the data projects to those. Then you can’t go wrong.” (23:31, Ed)

Quote:
“Flip the script and start with the outcomes, building back from there. Love to hear it.”
— Jordan, (24:46)

Memorable Quotes & Timestamps

• “Agentic capabilities actually help you augment and expand more and more capabilities...there are certain use cases where you want the performance of a fixed workflow...this is where businesses are today.”
  – Ed, 04:08

• “Bad data in, worse data out.”
  – Ed, 06:18, repeated throughout

• “I love looking to see how models take whatever inputs and what they do and what they think through...sometimes even just one wrong assumption...they might take that and run with it.”
  – Jordan, 13:03

• “Trying to manage these thousands or millions of agents in your organization without using AI to manage them...is going to be nearly impossible.”
  – Ed, 15:42

• “Focus on the data sets that are going to matter for your agentic workflows. Start small and iterate from there.”
  – Ed, 21:46

• “Flip the script. Start with discrete projects where you want to define your ROI and...work backwards.”
  – Ed, 23:31

Timeline of Key Segments

| Segment Topic | Speaker (Min:Sec) | |------------------------------------------------|--------------------| | Automation → Agents trend/risks | Jordan (00:16) | | Differences in enterprise adoption | Ed (04:08) | | Pros & cons of agentification | Ed (06:01) | | Data importance for agentic systems | Ed (06:54, 09:38) | | Expense Report workflow transformation | Ed (10:10–13:03) | | LLM “guessing” & chain of thought | Jordan (13:03) | | Governance & Agent Control Towers | Ed (14:35, 16:41) | | Multi-agent orchestration complexities | Ed (17:42–19:49) | | Advice for leaders with “iffy” data | Ed (20:50) | | Measuring ROI, flipping the script | Ed (23:31) | | Wrap-up / final insights | Jordan (24:46) |

Tone & Style

• Conversational, jargon-free, pragmatic. Both Jordan and Ed aim to demystify the transition to agentic automation for everyday business leaders rather than only tech insiders.
• Frequent analogies (e.g., self-driving cars, sandwiches) help clarify complex ideas.

Takeaways for Listeners

• The shift to agentic workflows is accelerating and offers huge productivity gains – but demands a renewed commitment to data quality and governance.
• Don’t let “AI science projects” distract from measurable business outcomes; start with specific goals and build your agentic infrastructure to deliver them.
• If you don’t trust your data, focus on getting small, critical pieces right and iterate, rather than waiting for a mythical “perfect data” day.

For further reading and more AI news, Jordan suggests subscribing to the show’s free daily newsletter at youreverydayai.com.