AI agents for business operations are only as useful as the data they can reach. A model can reason through a pipeline review or a month-end variance analysis, but when the numbers it needs are spread across a dozen disconnected systems, the reasoning stalls before it starts.
At CData, we run our own marketing, finance, and business development on the AI gateway we’re building, and on November 5 at Foundations 2026, three people from those teams will show what that looks like day to day. In this post, we preview three functions, three stories, and a look at what each team is building next.
What changes when agents can reach every system
For most operational teams, the slowest part of analysis is assembling the data. Demand generation data lives in Salesforce, Google Analytics, Qualified, and Pardot, sales conversations live in Gong, finance pulls from BigQuery and Stripe, and reporting ends up in Tableau. Connecting those pieces has traditionally meant exporting by hand, reconciling in spreadsheets, and rebuilding the same slides every week, which keeps insight on a delay no matter how capable the people doing the work are.
An AI gateway addresses that constraint at the architectural level. It’s a single, governed layer between AI agents and enterprise systems that controls what agents can access, do, and spend, and it keeps the business context they need to answer correctly. With the gateway in place, agents query the systems of record directly instead of waiting on someone to export the right file, and they do it through the same permissions your teams already use.
Three teams, three teasers
Demand generation. Arielle Daigle, VP, Growth and Demand Generation, used to build her weekly funnel report by exporting from Tableau into PowerPoint and checking pace-to-goal with a calculator. Today, an agent she built herself refreshes that dashboard every morning, and she adds new features through a few rounds of back-and-forth in Claude instead of filing a ticket with the BI team. She’ll also share how Claude now reasons on ideal customer profile (ICP) fit across several systems every day.
Finance. For Kade Tibbetts, Senior FP&A Analyst, month-end close used to mean a sequence of exports, cleanup, and model roll-forwards, each step waiting on the last. Now agents pull from every source at once and run his skills in parallel as soon as the books close. He’ll walk through a client purchasing journey view built from Salesforce, BigQuery, and Stripe data.
Business development. Elayna Leblanc, Manager, Business Development, leads a team that averages around 21 calls a day, far more than anyone could grade against the SPICED methodology by hand. With agents scoring Gong transcripts, her coaching now covers every call instead of a sample. She’ll show how combining Gong, Salesforce activity, and trial telemetry surfaces high-intent leads that no single signal would reveal on its own.
From analysis to action, with an audit trail
Reasoning over live data is valuable on its own, and the larger shift comes when agents act on what they find. On Arielle’s team, an account-based marketing (ABM) program feeds intent signals into Salesforce, and agents reassign account tiers as intent surges, work that previously ran through manual Excel uploads. Elayna’s business development reps work from a single interface where an agent updates records, sends emails, and enrolls leads in sequences, replacing a routine that used to span four or five systems.
Write-back raises a fair question about control, and the answer lies in how requests are authenticated. Every change goes through the requesting user’s own credentials, so the Salesforce admin can see which person made an edit and when it happened. Agents inherit the access each user already has, down to the record, which gives every automated change the same accountability as a manual one. The Inside the Gateway session at Foundations walks through how that governance follows every request.
Why the work survives a model change
Our teams use Claude as their client today, and the gateway is model-neutral by design. Asked during the panel what would happen to everything they’ve built if a new model or host came along, Kade’s answer was simple: the logic, the data structure, and the business definitions all live in the gateway layer, so you can put a different model or host on top and keep the same results. That design choice is what turns individual agents into durable infrastructure for your organization.
The Context Engine session goes deeper on that idea, showing how definitions and approved corrections become knowledge every agent reuses, so your context stays yours whatever model you run. The panel closes on the question each team is still answering: where its function goes next. Expect to hear about dashboards that explain the why behind the numbers, forward-looking finance analysis, and reps who spend their time on conversations.
See how AI agents run real business operations at Foundations 2026
Hear the full “CData on CData” conversation, including what each team is building next, on November 5. Register for Foundations 2026 to see how CData Connect AI Gateway gives agents one governed path into every system you run.
Frequently asked questions
When is CData Foundations 2026?
Foundations 2026 takes place on November 5, 2026, at 11:00 AM ET. Registration is open on the event page.
Who should attend Foundations 2026?
The event is built for AI leaders, CIOs and IT leaders, and functional leaders in sales, finance, support, and operations. Anyone deciding how agents will work with enterprise systems will find a session that fits.
What sessions are on the Foundations 2026 agenda?
Six sessions cover the AI gateway opportunity, a guided walk inside the gateway, the context engine, a CIO’s rollout at Anaqua, CData Labs research, and the “CData on CData” panel with teams from marketing, finance, and business development.
