Daily BriefAI activity is not AI productivity
More tools, users, and pilots can still leave one important question unanswered: what part of the business is actually working better?
Welcome back.
I've spent the last week thinking about a distinction I suspect a lot of companies are about to run into: AI activity is not AI productivity.
They can look similar from a distance.
Both can produce new tools, busy Slack channels, impressive demos, internal champions, and a growing list of pilots. But activity tells you that AI is happening. Productivity tells you whether the business is actually working better because of it.
That means lower cost, faster execution, improved quality, or growth.
The distinction matters because AI activity is getting easier to create. The models are improving. The tools are easier to access. Employees are experimenting with or without an official company program.
But turning that energy into dependable value is hard.
The gap
Microsoft and LinkedIn's 2024 Work Trend Index captured the tension early.
In its survey of 31,000 people across 31 countries, 75% of knowledge workers said they were already using AI at work. At the same time, 59% of leaders worried about quantifying the productivity gains, and 60% worried that their company lacked a vision and implementation plan. Among people using AI, 78% were bringing their own tools to work.
While these numbers from 2024 and they can’t exactly show where every company stands today, the pattern is a useful one to study: employee activity can spread much faster than the organization around it.
The common response is to buy a company-wide tool and call the rollout an AI program. The employees are then expected to figure out the rest.
They have to learn how to use the tool for real work, identify worthwhile opportunities, safely connect the information the AI needs, evaluate the output, and manage the risks. They are usually expected to do all of this while continuing to perform their normal jobs.
A few motivated people will make it work. They will build strong personal systems and report meaningful time savings. But isolated success is not yet organizational productivity.
If the result depends on one unusually capable employee, nobody owns the wider workflow, and the improvement cannot be measured or repeated, the company has created a useful experiment. It has not yet created a dependable organizational capability.
The workflow is the unit
This is why I keep coming back to workflow redesign.
In McKinsey's 2025 state of AI report, workflow redesign had the strongest relationship with reported EBIT impact among the 25 organizational attributes the researchers tested. Yet only 21% of respondents whose organizations used generative AI said they had fundamentally redesigned at least some workflows.
That finding is correlational and based on self-reported survey data. It does not prove that workflow redesign alone produces financial returns. But it points in the right direction.
AI value materializes inside the actual flow of work.
Imagine a client-services team preparing for a renewal.
Giving every account manager an AI assistant might help some of them write emails or summarize notes faster. That is useful.
Redesigning renewal preparation is different. The team decides which information the AI should assemble, which systems count as authoritative, which commitments it should flag, what a person must review, who owns the final decision, and how the company will measure whether preparation became faster or more reliable.
The tool may be the same in both cases.
The surrounding system is what turns personal assistance into a repeatable business result.
A better test
Instead of asking how many people are using AI, I would start with four questions:
Which workflow changed? Name the actual sequence of work, not the tool or department.
Which business result should improve? Choose cost, speed, quality, or growth, then establish a baseline.
What makes the result dependable? Identify the context AI can trust, the standard for a good output, and where human review belongs.
Who owns the outcome, and can other people repeat it? A personal technique becomes an organizational capability only when it can spread and survive beyond its original creator.
These questions are intentionally simple. They will not produce a complete AI strategy.
They will tell you whether you are looking at activity or the beginning of something the organization can actually rely on.
TEAM READINESS AUDIT
Get your team’s AI readiness score. Leave with an AI readiness brief.
A 15-minute assessment that turns your answers into a real decision artifact: your readiness level, your six-axis shape, where you're strongest, where you're constrained, and what not to build yet.
Start with one workflow
The answer is not to slow down experimentation. Activity is the raw material. The mistake is treating activity as the destination.
Pick one frequent, consequential workflow where the result can be measured, and mistakes can be caught. Establish how it works today. Decide what responsibility should belong to AI and what should remain with a person. Give the system the context it needs. Define the review rule. Run it enough times to learn whether the result is real.
If it works, turn the experiment into a shared practice and expand deliberately. If it does not, you have learned something before creating another tool, pilot, or dashboard.
The companies that make real progress with AI will not simply have more usage. They will become better at turning individual experiments into dependable ways of working.
That is the shift from AI activity to AI productivity.
SPONSORED BY CLUTCH
Hire secure AI teammates that work 24/7.
Hire pre-built AI teammates. Give your engineers and operators a platform to ship their own AI apps. Stop losing sleep about what is running where.
Clutch is the platform behind both: pre-built agents for the workflows your ops team should automate first, plus the integration plane your team's vibe-coded apps and Claude Code projects plug into. One platform. Real production. Visible and safe by default.
Built for ops, engineering, and security teams that are tired of the shadow-AI surface area inside their own company.
Know someone responsible for proving whether AI is actually working? Forward them this issue and ask them to run the four-question test with you.
Haroon
Was this email forwarded to you? Subscribe to AI Ready to get the next issue.
