In this article, I explain the gap between having advanced data capabilities and leveraging them to drive business outcomes. Google has best-in-class capabilities, but the business lags behind. Its most recent ToS gaff isn’t getting enough attention, but the impacts on earnings will.
BIG ANNOUNCEMENT! I have opened an instructor-led cohort of my AI Strategist Certification course. There are 2 reserved spots for subscribers, but due to pre-registrations and running fewer cohorts this year, space is very limited. Enroll Today.
Having advanced data capabilities doesn’t automatically make a business capable of delivering value with data. Everyone in our field shares this common experience. The data team ramps up, but the business doesn’t adapt. The result is that insights are available, but no one uses them to improve business decisions.
The most recent change to Google Services’ Terms and Conditions highlights the problem. Even a company as large as Google can fail to use its AI muscles to solve basic business problems. Here’s the complete document, but I have highlighted the key areas in the image below.
I’m not a lawyer, and most Google account admins or Workspace users aren’t either. While most consumers won’t read this, companies and organizations will. The new language seems to give Google rights that no business or institution would agree to.
The company explains that users retain rights to their intellectual property, which sounds good. What follows is problematic. By accepting this agreement or continuing to use Google Services after May 24th, organizations and individuals grant Google a broad license to use their intellectual property.
The Terms of Service reads like a content licensing agreement. Google has a license to “modify and create derivative works based on your content.” LLMs do that when our content is included in the training data.
Google defines the purposes it can use the content for, which includes “using automated systems and algorithms to analyze your content … to recognize patterns in data … {and for} … developing new technologies and services for Google.” The examples make these purposes sound benign but don’t limit Google from exercising the full scope of these rights.
My takeaway is that Google plans to mine data from Google Drive, Gmail, its Office apps, and other services to train LLMs. I use Google Drive to share files and course recordings with students, but I will switch those files to AWS this weekend. I can’t risk my IP or student privacy being compromised, and neither can the high schools that use Google Services.
Google Services are prevalent in startups, small businesses, colleges, and nonprofits. After reading this, I don’t know that they can continue using Google Services. Since Google hasn’t clarified the language or done enough to clearly limit the scope, the actual legal definition doesn’t matter. As I said, most of us aren’t lawyers, so our perception is reality.
How Is This A Data Problem?
Most businesses feel like they understand their own operations, but the reality is quite different. Operational workflows are understood in pieces and silos. Only small parts of a complete workflow are visible, and they are only visible to a small part of the business. Data changes that from both sides.
One of my frameworks is called ‘Transparency and Opacity.’ It’s the starting point and justification for a data maturity journey. Any part of the business that doesn’t generate data is mostly opaque to everyone except those who manage it.
The data team supports the business in understanding itself and its customers better by gathering data connected to workflows. Making that data accessible across the enterprise brings transparency to that part of the business. The more transparency that data creates, the better the business can manage itself and how it creates value.
The sales funnel is a long process that is one part internal workflow and another part customer workflow. Transparency helps business leaders understand long and complex workflows like sales funnels. That’s what Google is missing.
They have the capability to map their sales funnel and understand the upstream and downstream impacts of a decision. However, it doesn’t look like the business is doing this. If it were, business leaders would realize that the new ToS will break a critical part of Google’s sales funnel.
What Happens When Decisions Are Made In Isolation
Enterprises avoid Google Workspace because there’s a perception that Google uses customer data for its own purposes. This perception spills over to GCP at the worst possible time. GCP will be a pillar of Google’s AI strategy, so it needs enterprise adoption.
Google’s thinking is that since enterprises don’t use Google Services or Workspace today, any loss in business from Google Services will not matter in the long run. Few large enterprises run on GCP either, so who is GCP’s main customer segment? Startups and tech-first small businesses.
Opacity in the complete sales funnel hides the downstream impacts of the ToS change. Most startups and small businesses that use GCP also use Google Workspace, which is how Google hooked them into GCP in the first place. GCP’s sales funnel starts with Google Workspace.
Another problem with opacity is impacts that play out over long time spans are difficult to trace back to their root causes. Google expects increased churn and decreased adoption of its Services and Workspace products in the near term. Those don’t drive much revenue, so the impact will be minimal.
Reading the ToS will reinforce the perception that Google can’t always be trusted with a business’s data. Some businesses and organizations that migrate away from Google Workspace will eventually migrate away from GCP. Those impacts will not show up immediately, and the connection will not be obvious.
There’s a lag between a startup adopting Google Workspace and moving to GCP. Productivity apps are the first parts to be set up. Most products are developed locally in the early stages and only move to the cloud once that work is done. Customer workflows are just as important as internal workflows.
Fewer startups and small businesses adopting Google Workspace will result in fewer migrations to GCP. Those impacts will not show up for 6-12 months. Since there is opacity in the sales funnel, it’s unlikely that anyone will look at the ToS as a root cause. Opacity makes it difficult to see the impact of decisions in advance or find root causes after a problem becomes apparent.
Lessons Learned For Data And AI Products
When data gathering doesn’t align with user needs, products aren’t adopted. Unclear data usage policies are why so many LLM-supported products are banned by enterprises. The reality doesn’t matter. Business leaders perceive a threat to their intellectual property and have responded.
Trust is critical for AI products and any enterprise customer data platform. Why risk making the perception worse? Google gives enterprises reasons to worry about changes like these while other hyperscalers reassure them.
Adobe clearly stated, ‘If you're on a paid account, we don’t train LLMs with your data. We can use your data if you’re on a free account.’ Businesses aren’t afraid of Firefly.
Microsoft clarified GPT on Azure’s terms early by stating that enterprise data was secure and would not be used to train its LLMs. Copilot’s adoption curve has been strong.
Google Services are already prohibited in many enterprises, and the new ToS will push GCP’s core customer segments in that direction. Avoid making the same mistake with your data and AI products.
Build trust through transparency. Be clear about how customer data is used and how it will not be used. Don’t rely on legal-speak or ToS to get the message across. Discover your customers’ biggest concerns and explain how your data and AI product alleviates them.
Internal data products should make the business more transparent, but they can’t take autonomy away from people. Decision support tools have very high ROI, but only if people use them. Improving transparency enables people to make better decisions, but it won’t make those decisions for them. Explain how products help people maintain autonomy.
Use stories like this one to explain the paradigm and get buy-in. Business leaders connect the dots through stories like these. They are powerful tools.
This story is a good segway into the next article. I will return to abstract systems and how to model them. A sales funnel is an abstract system, and I’ll explain how we deliver actionable data to decision-makers. Engineering a sales funnel and aligning it with marketing, customer segments, and products requires a detailed understanding of how all the pieces interact with each other.
More to come.
Can’t think about anyone else who picked up on this and explained second and third order effects, stellar job.