06.13.2025

ForwardThinking 🤔

Salesforce's Slack data ownership grab

I write so much about the importance of data in this publication, but that’s because data is the lifeblood of large language models. If data quality is critical to model performance, then the data within your enterprise applications is even more so. It enables companies to build more relevant applications that take into account your particular business information needs.

That’s why OpenAI’s announcement last week about Connectors was so significant. It allows you to confine your query responses to a particular data repository, like all your data stored in Box. But what if SaaS vendors started limiting access to that application to vendors that customers rely on, effectively putting a gate up in front of that valuable data?

That’s essentially what happened this week when The Information reported that Salesforce was cutting off long-term access to the Slack API, a blow to third-party vendors who rely on that data. The beauty of Slack has always been that it is a central place where work happens and lots of companies have been able to piggyback on the data, creating even more value.

Putting up gates

The move raises serious questions about data ownership and what companies can do with the data they create in SaaS applications. To borrow an example from The Information article: If a company wants to let an AI assistant like Glean access its Slack data, shouldn't it be allowed to? That is an essential question being asked this week in light of this report.

Gate with red sign with white block letters that says 'Do not block the gate.'

Photo by Donald Giannatti on Unsplash

When asked about the accuracy of the report, Salesforce gave kind of a nuanced response. “At Salesforce, trust is our number one value — and that starts with protecting customer data. The new innovations we’ve announced, along with updates to Slack’s API Terms of Service, open the door for more intelligent, context-aware AI experiences and ensure that developers, customers and partners can securely and responsibly interact with Slack data.”

Notice what Salesforce did not say. They did not deny that there was a change in the terms of service, or that it could be having a negative impact on companies like Glean who rely on Slack data. They instead tried to position it as a positive change designed for the AI era. But is it?

Whose data is it anyway?

In my mind, when you buy a product like Slack, you do so because you want a central place to talk about work. The advantage Slack had over previous generations of enterprise communications tools was that it also served as a way to connect to enterprise software and do work in the same place where you’re having the conversations.

Fast forward to today and that activity stream has more value than ever. If I have a vendor helping me put that data to use in other ways, and I want to give it access to my Slack data, shouldn’t I be allowed to do that? It’s a fundamental question.

Slack symbol on a 3D white cube sitting on a green and yellow background

Photo by Rubaitul Azad on Unsplash

Ray Wang, founder and principal analyst at Constellation Research, certainly thinks so. “Customers who want access to their own data from other applications they own should have access. Otherwise why buy packaged software today at huge premiums,” Wang asked. “We are at a point where many customers are weighing ‘build versus buy’ and this is just one more reason not to buy.”

Brent Leary, founder and principal analyst at CRM Essentials, offers a more pragmatic assessment, saying that it appears that Salesforce is looking at Slack data as a valuable asset for enterprise AI, one that it could monetize down the road. Perhaps the change in terms of service is the first step.

“All the data created by those Slack interactions and collaborations typically lead to important workflows, process automation and business model transformations. And the worth of those interactions makes Slack data incredibly valuable, and provides Salesforce with a great opportunity to monetize it if they decide to do so at some point,” Leary said.

“Customers who want access to their own data from other applications they own should have access. Otherwise why buy packaged software today at huge premiums?"
~Ray Wang, Constellation Research

That workflow notion is particularly poignant. If agentic AI is about automating complex workflows, this type of data will become more valuable than ever. Yet paradoxically, by cutting off access, Salesforce could be diminishing the value customers see in using an application like Slack in the first place by making it harder for their partners to put it to use.

Sure, Salesforce can make any rules it wants around data usage, but the company could be treading into dangerous territory with this decision, and calling the Slack value proposition into question. And when you paid $28 billion for a property, you may want to move carefully before making such a move.

~Ron

Note that I’ll be on vacation for the next two weeks. FastForward will return the week of July 4th. Thanks dear readers for your support.

What's new on the blog 📰

For Qualtrics security chief, it’s all about safeguarding the data

I recently spoke to Assaf Keren, chief security officer at Qualtrics. Prior to joining Qualtrics, Keren spent a decade at PayPal, and we talked about the differences between running security at the two companies. He said one of the primary differences was the crown jewels, which involves protecting customer data at Qualtrics.

"One of the baseline things that we talk about when we talk about security is that we don't own that data. We are custodians of our customers’ data," he said.

Read the full article >>

Series of locks in a red chain link fence with a red background.

Photo by Jon Moore on Unsplash

Snowflake snags Crunchy Data to get enterprise-grade Postgres database

Just as I did with the Databricks-Neon deal a couple of weeks ago, I spoke to data analysts Devin Pratt and Sanjeev Mohan to get their viewpoints on the recent Snowflake-Crunchy Data acquistion.

I learned that even though both companies bought Postgres startups, each one had different motivations, and these are very different databases.

Read the full article >>

How the Sam’s Club CTO used computer vision and AI to transform the customer exit experience

Sam’s Club CTO Sanjay Radhakrishnan and his team implemented a solution built with computer vision and AI to transform the shopper exit experience. It had been a manual process that was prone to bottlenecks prior to this. His team took advantage of advanced technology to improve the customer experience.

"What's beautiful about this is that all of this processing happens in real time. So we have computer vision inferencing models that run on edge GPU servers. It's processing multiple 4k video streams and processing all of this in real time."

Read the full article >>

Two blue exit arches at Sam's Club. As shoppers pass under the arches, cameras pick up the contents of their carts and check against the receipt to very quickly make sure everything has been paid for.

Photo courtesy of Sam's Club

Workday wants to be an enterprise agentic management hub

As companies continue to move towards a future agentic vision, every SaaS company is going to want to be the management layer. We've seen moves from every major player. Earlier this month, Workday weighed in with its agentic management vision that combines a communications gateway with an agent management and governance tool. I took a look.

Read the full article >>

News of the Week 📣

Ravenous for compute, OpenAI makes a deal with rival Google

OpenAI is a thirsty company when it comes to money and compute resources. It never seem to get enough of either. Let's not forget that its last raise was $40 billion at a $300 billion valuation in March. If they raised now it would probably be higher.

But the flip side of that need for gobs and gobs of cash is an equally massive thirst for compute resources wherever they can find it, even if that means buying from a competitor.

Axios reported this week that the two companies inked a deal where Google Cloud will provide cloud infrastructure services for OpenAI to help keep feeding that resource beast.

Purple OpenAI logos floating on a purple and black striped background.

Photo by Mariia Shalabaieva on Unsplash

The company has a deal with Oracle to build more data centers as part of the $500 billion Stargate Project announced earlier this year, but chances are it's going to continue to try and grab as much cash and buy up as much compute power as it can.

So far, the plan seems to be working with the company announcing earlier this week that it has reached $10 billion ARR (annual recurring revenue) with its most recent weekly user numbers reaching 500 million, per CNBC. Of course, all that compute power costs a lot too, making profitability a challenge.

Databricks projecting $3.7 billion in annualized revenue by next month

Databricks looks a lot like a public company. It has 11,000 employees, thousands of customers, and this week it projected its annualized revenue will reach $3.7 billion by next month. That's up from $3 billion just last December.

Those revenue numbers are up 50% year over year. That an organization this mature is still growing 50% YoY is pretty remarkable, but perhaps the most impressive part of all this is that the company is still pre-IPO.

Photo courtesy of Databricks

At the end of last year, Databricks raised $10 billion -- its Series J if you're keeping score at home -- on a $62 billion valuation. By any measure, it's a healthy company and would presumably do well whenever it decides to go public. It has to pull the trigger one of these years, but for now it keeps on rolling along with big numbers, strategic acquisitions and growth that would make any company in the public markets jealous.

Cisco updates networking to meet growing AI demands

In case you haven't noticed, AI is making big demands on data centers, and your parent's networking equipment probably isn't going to cut it anymore. Networking giant Cisco certainly understands that, and this week they announced some next-generation hardware designed to deal with AI's ever-growing bandwidth requirements.

Keeping in mind a networking company is behind this research, Cisco recently released a report that found "97% of IT leaders see modernized networks as critical to deploying AI, IoT and cloud, with 91% increasing network investment." But with AI making so many demands on data centers, it makes sense that modernization would follow.

Not surprisingly the new hardware announced this week is designed to not only drive AI workloads, but also use it operationally to improve security and efficiency across the network infrastructure.

"Cisco is delivering the critical infrastructure for the AI era—secure networks and experiences, optimized for AI that connect the world and power the global economy,” Jeetu Patel, president and chief product officer at Cisco said in a statement.

My adventures in AI 🤖

Photo by Andy Kelly on Unsplash

Just this morning, I dropped the stories from News of the week into ChatGPT and asked for a review for clarity, spelling, punctuation and grammar. It gave me this feedback questioning the veracity of the $10 billion raise in the Databricks story:

Yes, thank you, but I've got it and it's right. The thing is I included a link to the company press release announcing the raise, but for now large language models seem incapable of checking links, which feels like a big flaw that needs to be corrected at some point.

I would love to hear how other folks are using AI in your job because I'm sure there is no one way to do this, and we are all figuring it out as we go.

What I'm reading 📚

Person reading book, but all you can see is their hands holding the book. They are sitting on a hardwood floor steeped in shadow, creating an interesting look.

LivPhoto by Blaz Photo on Unsplash

Constellation Research

JPMorgan Chase's Dimon on AI, data, cybersecurity and managing tech shifts

Larry Dignan

TechCrunch

How to delete your 23andMe data

Aisha Malik

Marcus on AI

A knockout blow for LLMs?

Gary Marcus

Axios

Scoop: Mattel is still deciding what its first AI toy will be

Ina Fried

Look who's talking 👄

"Data is the hardest part. It isn't the AI or machine learning. Getting the data in the form that's usable is the hard part."

- JPMorgan Chase CEO Jamie Dimon, speaking at the Databricks Data + AI Summit, as reported by Constellation Research's Larry Dignan.

Please subscribe to the newsletter and feel free to share.

3350 Virginia St, Suite 262, Miami, FL 33133

ForwardThinking 🤔

Salesforce's Slack data ownership grab

Putting up gates

Whose data is it anyway?

What's new on the blog 📰

For Qualtrics security chief, it’s all about safeguarding the data

Snowflake snags Crunchy Data to get enterprise-grade Postgres database

How the Sam’s Club CTO used computer vision and AI to transform the customer exit experience

Workday wants to be an enterprise agentic management hub

News of the Week 📣

Ravenous for compute, OpenAI makes a deal with rival Google

Databricks projecting $3.7 billion in annualized revenue by next month

Cisco updates networking to meet growing AI demands

My adventures in AI 🤖

What I'm reading 📚

Constellation Research

JPMorgan Chase's Dimon on AI, data, cybersecurity and managing tech shifts

TechCrunch

How to delete your 23andMe data

Marcus on AI

A knockout blow for LLMs?

Axios

Scoop: Mattel is still deciding what its first AI toy will be

Look who's talking 👄

Subscribe

FastForward #26: Salesforce’s Slack data ownership grab

You may also like...