Hey Data & AI Teams,

Whether your organization is deep into AI or still building the foundation, one thing is certain. Slow pipelines are the bottleneck that stops everything.

CData Sync was made for this moment, and the results back it up.

1 billion rows. 360 GB. From 11 hours, 6 minutes to 2 hours, 41 minutes.

Same table. Same infrastructure. No destination schema changes. 8 hours and 25 minutes saved on a single run.

CData Sync v26.2 cuts large scale replication time by 76% on that exact workload, through parallel partitioned reads and improvements to the write path. Nothing about the environment changed, which is the point. Most teams assume a result like this needs new hardware and a migration plan. It did not.

The 10-page report goes past the headline number:

- Benchmark methodology and test infrastructure across 3 Sync versions

- Partition tuning results, including why 8 partitions of 500k rows wins on 8 vCPU machines

- Full results for SQL Server into Databricks, Snowflake, and Redshift
- An isolated write improvement benchmark on a 500 column table
- The recommended configuration that got all 3 warehouses under 3 hours

If replication windows are currently deciding how fresh your dashboards, models, and AI agents are, this is worth 10 minutes of your day.

86% of enterprises say they are past AI agent pilots. 34% say they trust what those agents actually do. That 52 point gap was the real subject of Boomi World Tour Sydney last week. Here are my 5 takeaways now that I have had a week to sit with it.

1. The conversation moved from what an agent says to what an agent does.
Michael Barnes at Omdia framed it as a shift from model risk to execution risk. That is the correct framing. Nobody is losing sleep over a badly worded response. They are losing sleep over an agent writing to a general ledger, and no one being able to reconstruct that decision in an audit 6 months later.

2. Finance arrived in the middle of the AI conversation, and it is not leaving.
2 years ago, 31% of FinOps practitioners managed AI spend. This year it is 98%, across 1,192 practitioners representing more than 83 billion dollars in annual cloud spend. Token cost is now a line item somebody has to defend. Agent-level attribution stops being a nice dashboard and starts being the thing that keeps a program funded.

3. In path beats alongside.
Most governance tooling observes. It reads logs, builds dashboards, tells you what happened. Boomi is arguing that enforcement has to sit in the request path, between the agent and the system of record, or it is not enforcement. Rate limits, token budgets and human-in-the-loop approvals only work if the call has to pass through them.

4. Deployment flexibility is the sovereignty answer, not a checkbox.
Running agents, tools, bring-your-own models and small language models inside a customer VPC or on premises within a regional boundary is what makes this usable for regulated APJ organisations. Australian financial services and public sector buyers do not get to move sensitive workloads to a public model endpoint and sort out the paperwork later.

5. Boomi is competing on what it already has, a trusted foundation.
More than 30,000 customers, over 200 billion messages processed a quarter, more than 16 security certifications, 800+ partners. The advantage is that the governed connections to core systems are already there. Agent infrastructure integrates directly with plumbing that has been audited for years rather than stood up fresh.

Where I am still asking questions.
Boomi's Agent Control Plane is available today: an Enterprise Platform solution uniting connectivity, governance, data grounding, and execution under one control point. Newer pieces roll out through 2026; core capabilities are live now.

Gartner's second finding is worth watching here too: applying uniform governance to every agent, regardless of autonomy level, is itself a path to failure. Boomi's human-in-the-loop approval step for high-risk transactional actions is a start toward that kind of tiering. The open question is how far that granularity extends as more agent types and autonomy levels enter the picture.

🔍 Stay Ahead in AI & Data! Join 137K+ Data & AI professionals who stay updated with the latest trends, insights, and innovations.

📢 Want to sponsor or support this newsletter? Reach out and let's collaborate! 🚀

Best,
Ravit Jain
Founder & Host of The Ravit Show