Calling an AI model is easy. Letting an entire company use AI without blowing the budget, losing control, or giving the compliance team heartburn? That’s where it gets interesting. @JoeKarlsson1 gives a quick overview of AI gateways, and why they’re becoming essential for scaling AI across the business.

Foundations returns November 5 with one focus: the AI gateway. Business context for every agent and AI tool from day one. Governance down to the record. A gateway that gets smarter and more efficient with every interaction. Registration is open: bit.ly/4cXTzJ8

From CData Labs: on raw access to enterprise data, not one of 22 models stayed within what it was authorized to change. Once write-time validation moved into the governed data layer, unauthorized writes dropped to zero across every model, including frontier. Safety lives in the architecture. Full benchmark: bit.ly/4yfGcNc

Push accuracy and safety into the data layer and the model becomes a cost decision. We benchmarked 22 AI models, economy to frontier, on live enterprise data through @CDataSoftware Connect AI: the identical correct answer cost 178x more from one to the next. cdata.com/lp/ai-cost-whitepa…

We tested 22 AI models on live enterprise data: bit.ly/4yfGcNc With context and governance in the data layer, the cheapest model returned the same correct answer as the most expensive, at 1/178th the cost. @jeRodimusPrime covers it in 60 seconds.

Enterprise agents are spending a growing share of their time in ERP and CRM once they're connected to business data. Across CData Connect AI customers, one pattern over the past few months stood out: Sage Intacct @SageUSAmerica, @Salesforce, @NetSuite, and @HubSpot sat at the top of the systems agents touched most. That tracks with the kind of questions people ask agents to handle: what's still outstanding, is the period ready to close, where does this deal stand. Those are operational questions, and the answer usually lives in the system the business is already running on, not in a report about it.

The same correct answer cost 178x more on the most expensive model than the cheapest. CData Labs tested 22 models across 1,034 runs on live enterprise data. When the data layer supplies the right context and governance, cheaper models can perform like frontier models. Full benchmark: bit.ly/4yfGcNc

A quick behind-the-scenes (literally) shot of some upcoming content. Looking forward to sharing what we've got brewing at @cdatasoftware soon!

"AI gateway" is a term that's been floating around a lot lately and not everyone stops to ask what it actually means.

Behind the scenes action from a big week of filming with @cdatasoftware customers, partners & team. Somewhere in here is the take where I forgot my own title 🤦‍♂️ Big news coming from our end. More soon!

AI agents are changing how data gets accessed. An agent starts with a question and works out on its own which records, often scattered across a system, actually answer it. In CData Connect AI usage data, that shows up clearly in @NetSuite: the heaviest agent activity among enterprise accounts centered on transactions, accounts, and accounting periods, separate objects that all fed one question, is this period ready to close, and if it isn't, what's holding it up. @Salesforce showed the same shape: opportunities, contacts, and owners as pieces of one business question. Making records available was the easy part. Giving an agent enough context to connect them across systems is where it got interesting.

Speed and governance don't have to compete. A @dbtrends roundtable on data products for AI makes the case for building governance in from day one instead of bolting it on later.     Watch: bit.ly/4wikB4Y

Erik Bailey wanted AI to become a core operating capability across every function at @Anaqua, not a single department's tool. That kind of scope needed one tool that could sit across every system, not a set of one-off integrations. Connect AI became that tool, and Anaqua's team started catching patterns in support tickets and the sales pipeline that weren't obvious before: bit.ly/4pZFFvt

Agent activity on Connect AI moved sideways for most of the last 20 months. Then it broke. Over a 4-month stretch this spring, agent interactions roughly tripled while active accounts climbed at the same time. That pairing matters: if it were only usage climbing, a smaller set of accounts could explain it. Interactions and accounts moving together points to a wider group of organizations putting agents to work as overall activity grows. The same organizations keep returning to the same business systems week after week. That kind of sustained return signals something specific: teams found a job the agent does well enough to keep sending it back to. Experimentation hasn't slowed. It's just no longer the entire picture.

Stale snapshots. Missing cross-source context. Raw source code your model can't read. This is why AI agents give confidently wrong answers. CData is building an AI-ready pipeline live, connecting @SQLServer, @Salesforce, and @ServiceNow to @Snowflake. Sept 9: bit.ly/4gjht3U

📢CData登壇情報 ノーコードやAIを「全く知らない」人が56.6%、実際に職場で使いこなす人はわずか2.92%！？ 9月3～4 開催の #DXシステム開発Expo 2026 にCDataの杉本が登壇します。 技術は揃っているのに、なぜ現場は詰まってしまうのか？DX推進の3フェーズ(着手→浸透→定着)をリアルな事例で語ります。 セッション詳細はこちら👇 f2ff.jp/introduction/13521?e…