CData's Connect AI Gateway is a control layer that sits between AI agents (or the people using them) and the enterprise systems where work actually happens. Launched on September 29, 2026, it gives organizations a single control point for the models, tools, data and actions used by AI agents and the individuals who work with them, connecting them to enterprise systems through governed tools, applying company context at every step, enforcing permissions down to the record, and routing each request to the most efficient model for the task. The pitch to founders and operators is simple: an agent is only as useful as the company knowledge and access controls behind it, and stitching those together per system does not scale.
The Gateway is built on CData's existing connector portfolio, which means it arrives already fluent in the systems most companies run on. It understands the schemas, objects, relationships and operations of connected systems from day one, and imports the knowledge a company already holds, including metric definitions, business terminology and existing semantic models, so context accumulates inside the customer's environment and under its control. Schema-aware tools also send models only the information a request needs, cutting AI cost while keeping AI accurate across systems. Hundreds of CData-built connectors are exposed as governed tools over live systems the moment an agent connects, covering SAP, NetSuite, Salesforce, warehouses and databases, plus on-premises and legacy environments, and MCP servers a customer already runs are governed under the same policy.
The launch matters right now because enterprise AI is moving from chat to action, and buyers are asking harder questions about accuracy, permissions, and cost per request. Amit Sharma, CEO and co-founder of CData, frames the shift as agents taking action inside systems where the questions become whether the answer is right, whose permissions applied, and what each request cost, with the Gateway meant to answer all three from one control point. The product is led by Raviv Levi, CData's Chief Product and Technology Officer, who posted the launch. Early access is open through cdata.com/ai.
Comments (12)
Sign in to join the discussion.
P
Priya Vaswani14d ago
System of record for how your company thinks is a spicy phrase. What happens when the record disagrees with the CEO?
T
Tomasz Kwiatkowski14d ago
That B-roll of the dashboard zoom at 0:12 was clean, but the voiceover sounds like it's reading a whitepaper at a funeral. Punch up the hook.
A
Adeola F.14d ago
Before I let this route anything, where does the learned context live and can I pin it to a region? Asking for every GC watching this thread.
M
Mireille Okafor14d ago
Gateways are a crowded rack. Curious what the net retention story looks like once teams realize governance isn't a feature they emotionally connect with. Interesting.
B
Bora Şentürk14d ago
Every AI gateway demo looks incredible until someone asks it about the return policy for a 2019 SKU. I'll wait for the ugly screenshots.
K
Kenta Aramaki14d ago
We're shipping something adjacent in about three weeks, and I have to admit the framing here is sharper than what we had on the whiteboard yesterday.
R
Ravi Menon14d ago
Third AI gateway launch this month by my count, but the routing-plus-learning combo is a cleaner pitch than the last few I saw.
Z
Zoya Balogun14d ago
Naive question but if the gateway learns from requests, and someone asks it a wrong thing confidently, does that become part of the record too?
L
Lin Marchetti14d ago
Already forwarding this to two portfolio CTOs. The 'teach it like an employee' framing does a lot of heavy lifting here.
H
Hakim Ostrowski14d ago
Working on a piece about the gateway layer consolidating. Happy to chat off the record about who the early access design partners are, whenever you have a minute.
M
monkmode14d ago
The company that cannot explain itself cannot delegate itself.
D
Dr. Anouk Vermeulen14d ago
Curious whether the 'learning' is retrieval over prior traces or actual preference modeling. Very different failure modes, and the landing page is coy about it.
Work With Us
Book an intro call to explore our suite of services. We might have a waitlist.
CData
The gateway that makes AI more accurate, for less
The gateway that makes AI more accurate, for less
Tags
TLVC Rating
Community Rating
No ratings yet
to rate this launch.
CData's Connect AI Gateway is a control layer that sits between AI agents (or the people using them) and the enterprise systems where work actually happens. Launched on September 29, 2026, it gives organizations a single control point for the models, tools, data and actions used by AI agents and the individuals who work with them, connecting them to enterprise systems through governed tools, applying company context at every step, enforcing permissions down to the record, and routing each request to the most efficient model for the task. The pitch to founders and operators is simple: an agent is only as useful as the company knowledge and access controls behind it, and stitching those together per system does not scale. The Gateway is built on CData's existing connector portfolio, which means it arrives already fluent in the systems most companies run on. It understands the schemas, objects, relationships and operations of connected systems from day one, and imports the knowledge a company already holds, including metric definitions, business terminology and existing semantic models, so context accumulates inside the customer's environment and under its control. Schema-aware tools also send models only the information a request needs, cutting AI cost while keeping AI accurate across systems. Hundreds of CData-built connectors are exposed as governed tools over live systems the moment an agent connects, covering SAP, NetSuite, Salesforce, warehouses and databases, plus on-premises and legacy environments, and MCP servers a customer already runs are governed under the same policy. The launch matters right now because enterprise AI is moving from chat to action, and buyers are asking harder questions about accuracy, permissions, and cost per request. Amit Sharma, CEO and co-founder of CData, frames the shift as agents taking action inside systems where the questions become whether the answer is right, whose permissions applied, and what each request cost, with the Gateway meant to answer all three from one control point. The product is led by Raviv Levi, CData's Chief Product and Technology Officer, who posted the launch. Early access is open through cdata.com/ai.
Comments (12)
System of record for how your company thinks is a spicy phrase. What happens when the record disagrees with the CEO?
That B-roll of the dashboard zoom at 0:12 was clean, but the voiceover sounds like it's reading a whitepaper at a funeral. Punch up the hook.
Before I let this route anything, where does the learned context live and can I pin it to a region? Asking for every GC watching this thread.
Gateways are a crowded rack. Curious what the net retention story looks like once teams realize governance isn't a feature they emotionally connect with. Interesting.
Every AI gateway demo looks incredible until someone asks it about the return policy for a 2019 SKU. I'll wait for the ugly screenshots.
We're shipping something adjacent in about three weeks, and I have to admit the framing here is sharper than what we had on the whiteboard yesterday.
Third AI gateway launch this month by my count, but the routing-plus-learning combo is a cleaner pitch than the last few I saw.
Naive question but if the gateway learns from requests, and someone asks it a wrong thing confidently, does that become part of the record too?
Already forwarding this to two portfolio CTOs. The 'teach it like an employee' framing does a lot of heavy lifting here.
Working on a piece about the gateway layer consolidating. Happy to chat off the record about who the early access design partners are, whenever you have a minute.
The company that cannot explain itself cannot delegate itself.
Curious whether the 'learning' is retrieval over prior traces or actual preference modeling. Very different failure modes, and the landing page is coy about it.