Dr. Ryan Ries here back to say, "Go Spain!".
In all seriousness, I’ve been thinking a lot about this concept where AI in business takes two different paths: personal productivity and business productivity. This has come up a few times recently in conversations with customers, in our AIM workshops, and in our recent webinar, so I figured we should talk about it here as well.
Before I start though, if you’re in the NYC area, we’re running a half day workshop with AMD on August 13th all about running agentic workloads on CPU, rightsizing instead of just going bigger. We all want to talk about GenAI using big GPUs, but there is a whole ecosystem around every agent that uses all kinds of different compute to get to the final solution. Sign up here!
Two Layers, Not One
AI in business runs on two different layers, and most leaders grade both against the same rubric.
Let’s call the first one the assistance layer. This is where AI acts like a personal assistant baked into your daily tools. Copilot, Claude Cowork, helping with things like drafting emails, creating meeting summaries, etc. This layer optimizes how we work. It cuts friction and busy work, and it runs on a simple loop: you ask, it responds, and you execute. Of course, with modern tools you can create agents that will do a lot of this busy work on a daily basis and have it ready for you when you start the day. The value metric here is time saved and personal efficiency, plain and simple.
Now let’s call the second one the execution layer. This is AI embedded straight into business systems. This looks like: RAG chatbots pulling from your actual data, workflows, and agents connected to enterprise systems that complete tasks on their own.
For example, picture a loan processing agent working through risk assessment, pulling application data, and returning an approved or denied decision, tied into the real workflow from end-to-end. This layer optimizes what work gets done, not how one person gets through their day. The value metric shifts from outcomes and throughput, not actually personal time saved.
All that being said, here is where I see leaders getting tripped up on how to measure productivity and ROI.
A tool built for the assistance layer, but gets measured against execution layer expectations, fails every time.
An executive assistant tool that clears your calendar and drafts your emails was never going to process a loan packet. A loan processing agent connected to enterprise data was never designed to summarize your Tuesday meetings.
To measure how AI is performing, you have to match the tool to the right layer first. Then pick the metric that actually belongs to that layer.
These tools are smart, but you can’t expect them to do double duty.
The 5X Problem
I saw a story recently that I felt perfectly captures this issue.
An engineering leader built out an AI-driven development process for his team. The result landed at 15-20% more efficient. He took that number to his CEO expecting a pat on the back.
The CEO asked why the team wasn't hitting 5X. Apparently that's the figure floating around out there in the market?
Five years ago, a 15% efficiency gain from a new tool would have at least gotten you a pizza party and a gold star. Somewhere between then and now, people have become convinced that anything short of a multiple is a failure.
However, if you ask me, a clean 15-20% efficiency gain, proven and repeatable, beats a fictional 5X that nobody can actually point to on a real project.
Setting Expectations
If you’re having these conversations with leadership at your organization, here’s what I’d tell you.
Have a clear thesis before you start. Write down what you actually expect a tool to change and why. If the honest answer is “we don’t know yet,” that’s ok. But making up potential outcomes and living on a prayer is not the way to go!
Ask your people directly. Dashboards are great but only tell part of the story. A conversation with the team using the tool every day tells you the rest. If they say their week got better, that counts as a result, even if it isn’t a metric to share with your CEO.
Separate the categories before you measure anything. Decide up front whether you're evaluating a personal productivity tool or a business workflow rebuild. Again, grading one against the other's scorecard = bad conversation with leadership down the road.
Push for more outcomes, but in a way that's realistic. Celebrate the 20% efficiency gain! That’s big! But don’t get complacent. Go look for the process that might support three or four times the output and treat that as its own separate bet with its own separate expectations.
My Thoughts
I get why the 5X number spreads so fast. It sells conferences and we all want to see that sexy metric! But a team that gets 15% more done every single week, month after month, will outrun a team chasing fictional outcomes that can’t actually be substantiated.
When you're evaluating your next AI rollout, ask yourself:
- Which lane am I actually measuring, personal or business productivity?
- What number did I write down before the project started, and did it come from my own data projections?
- Have I asked my team how their day actually changed, not just what the dashboard says?
- Am I willing to call 20% a win? Or am I chasing something unrealistic?
Let’s Talk
If the two-layer split makes sense in theory but you're not sure where your own tools and workflows actually land, or you have tons of ideas for AI in your business but you don’t know what to do next to make sure your team actually realizes the value from them, let’s chat! That's exactly what we work through in our Mission Cloud AIM sessions. We help you sort out what's working from what just feels productive, and hand you a plan you can act on. Reach out to our team here.
Until next time,
Ryan
Now, time for this week’s AI-generated image and the prompt I used to create it.
Create an image of me facing off with robots for the world cup. I am on team spain and facing off with argentina. You can see thousands of fans in the stands. use my image as reference for me in the photo.
