One message coming out of Dreamforce caught my attention: context is moving to the center of the enterprise AI conversation.
Good.
We’ve been arguing for a while that context is what turns AI from an impressive interface into something genuinely useful for the enterprise.
Salesforce’s AIforce approach is about bringing its data, workflows, business logic, semantics, permissions, security, and governance into the places people increasingly want to work, including Claude and Slack. My reaction is simple: yes, and.
Because enterprise context rarely lives inside one application, one data source, or one ecosystem. It lives in Salesforce and SAP, Snowflake and Databricks, AWS and the bespoke system somebody built ten years ago that nobody particularly wants to touch.
Salesforce context should travel. And so should everything else your business depends on.
Context is bigger than the application
Imagine asking an AI assistant why revenue is down in a particular region. Your CRM can tell you what happened to pipeline, and that is useful. But why did it happen?
Maybe product usage changed. Maybe support volumes spiked. Maybe inventory was constrained, or pricing shifted, or the supply chain hiccupped. The real answer is scattered across every system above, not just the one that happens to own the account record. Even Yoav Kolodner, a former Salesforce engineering VP, made a version of this point at Dreamforce: metadata is the layer that actually describes a business, and giving an agent access to Salesforce doesn't erase the technical debt sitting in everything else underneath it.
Here's a real example. Using CRM data alone, the picture for a major region last quarter looked simple: revenue was down, pipeline coverage was thin, case closed.
Add governed account data through MCP and cross-reference renewal risk against product usage, and a very different picture emerges. The biggest driver wasn't competitive pressure or budget. It was a pocket of accounts quietly disengaging, buried inside regional usage numbers that were still climbing overall. The CRM told one story. The usage data told another. Neither was wrong. Neither was complete. Only reading them together found what was actually happening.
That's what "and" should mean in practice. Not another claim. A better answer, built from more than one system's version of the truth.
Trust has to travel too
The same applies to trust. As AI starts reasoning across systems, permissions, lineage, data quality, business definitions, and accountability cannot stop at the boundary of one application.
That is why we have been deliberate at Qlik about connecting AI to governed data and analytical capabilities. Qlik’s MCP Server lets assistants such as Claude securely use Qlik apps, data models, insights, and automation while respecting the permissions of the Qlik user making the request.
You can work in the AI environment you choose while bringing governed Qlik analytics into the conversation.
Freedom is where “and” gets real
This may be the most important part. Nobody knows which model, assistant, or interface will define enterprise computing 18 months from now, and I am not convinced there is supposed to be one winner.
Your architecture should give you room for that. Claude and other models. Salesforce and SAP. Snowflake and Databricks. What you have already invested in and whatever comes next.
That is not philosophy. It is an engineering decision.
Qlik Open Lakehouse, for example, can make one governed dataset available to Redshift, Snowflake, and Databricks without copying the underlying data into each platform. You keep one foundation while different teams continue using the engines that make sense for them.
Earlier this year, Qlik also joined Snowflake’s Open Semantic Interchange initiative, helping build an open, vendor-neutral way to keep business metrics and definitions consistent across analytics and AI tools.
I like that example because openness is easy to talk about when everything stays inside your own portfolio. It means more when you design for interoperability across the technologies your customers actually use.
Stop picking sides
At Qlik Connect, we framed the enterprise AI challenge around three things: Context, Trust, and Freedom.
For years, enterprise technology has been sold through false choices: cloud or on-premises, suite or best-of-breed, speed or governance, innovation or control. Pick an ecosystem and live in it.
Real businesses do not work that way. They are constantly changing combinations of people, applications, data, infrastructure, and models. AI is not going to make that complexity disappear. If anything, it adds another layer.
So maybe we should stop asking customers to pick sides.
Yes, use the new interface, and connect it to the systems already running your business. Use powerful models, and give them trusted, governed context. Move fast, and keep the freedom to change direction. Choose the best technology for the job, and make those choices work together.
That is what openness should look like in enterprise AI.
Not another ecosystem you have to bet everything on, but the freedom to keep choosing.
In this article:
Executive Insights and Trends