If AI saved your team 10 hours this month, could you say what those hours produced?
The Spark
Faster is not the same as better. This is a four-stage check you can run on your own team, with your own work, to find out whether AI is making you faster, better, both, or neither. Each stage ends with one small page you can show your boss.
Sponsored resource 100+ ChatGPT Prompts to Revolutionize Your Day Supercharge your productivity with HubSpot's comprehensive guide. This free resource is your fast track to AI mastery:
|
Why this matters
AI makes development feel quick. Code shows up sooner, unfamiliar APIs get explained in seconds, tests appear on demand. It is easy to conclude that productivity is going up. But once money and headcount enter the conversation, "it feels faster" is not an answer. Three traps catch most teams:
Counting activity. Pull requests, lines of code and closed tickets say how busy people are, not how much value they made.
Trusting the feeling. A 2025 METR study of experienced developers found they felt faster with AI while their measured task time said otherwise. Your gut is not a measuring tool.
Measuring only the coding step. Say coding is half of someone's week and AI makes it twice as fast. The week shrinks by a quarter, not by half. And if the saved time goes into waiting for a decision or a review, the bottleneck just moved.
If you are a new tech lead, this check gives you a calm, fair way to answer "is the AI spend worth it?" If you are growing toward director, a defensible answer is worth far more than enthusiasm. The plan below is small enough to run inside normal work: one team, one type of task, about four to six weeks.
Before you start
Pick one team and one kind of task, such as bug fixes or small features.
Agree on what AI-heavy and AI-light mean, in one sentence each.
Tell the team this measures the approach, not the people.
Keep any workflow improvements the same for both modes, so you test AI and not a better process.
Stage 1 · Define
Week 1
Goal: decide what "faster" and "better" mean before anyone collects a number.
Choose one speed measure. Calendar time (start to merged, including waiting) or hands-on effort. Either is fine. Pick one and keep it.
Choose one quality measure. For example changes requested in review, bugs found within a few weeks, or rework.
Choose one cost measure: AI spend per task.
Write down the AI-heavy and AI-light modes so everyone follows the same rule.
ARTIFACT · Measurement sheet
Task type we are testing:
AI-heavy means:
AI-light means:
Speed measure (calendar time or effort):
Quality measure:
Cost measure:
We will stop and review on:Done when: someone outside the team could read the sheet and run the same comparison.
Watch-out: changing the definitions halfway through because the early numbers look odd.
Stage 2 · Compare
Weeks 2–4
Goal: let the same people do some tasks both ways, so each person is compared with themselves.
Write the estimate first. Before the mode is known, each person records how long they think the task will take.
Then assign the mode at random. A coin flip is enough. Do not let people pick the easy tasks for AI.
Do the task, and log the actual time, the mode and the AI cost.
Compare each person with themselves, not with their teammates. Experience, style and estimating habits differ too much between people.
ARTIFACT · Task log row
Task and type:
Estimate (written BEFORE the mode was known):
Mode (AI-heavy / AI-light):
Actual time:
AI cost:
Review changes requested:
Notes:Done when: every person has at least a few tasks in each mode.
Watch-out: estimating after you know AI will be used. You end up measuring how much people believe in AI.
Stage 3 · Follow the work
Weeks 4–5
Goal: see the whole path of a task, not only the part AI speeds up.
Split each task's time into coding, waiting for decisions, waiting for review, and rework.
Watch the review queue. If code is written faster but reviewed at the same speed, work just piles up.
Check back two or three weeks later for bugs and follow-up fixes on AI-heavy work.
Ask the team what they did with the time they saved.
ARTIFACT · Where the time went
Coding:
Waiting for a decision:
Waiting for review:
Rework after merge:
What the saved time was used for:Done when: you can point to where the saved time went, or show that it did not turn into anything.
Watch-out: declaring victory because coding got faster while the queue got longer.
Stage 4 · Decide
Week 6
Goal: turn the numbers into a clear spending decision, even if the answer is "it depends".
Compare each person's AI-heavy tasks with their AI-light tasks. Look at speed, quality and cost together.
Split by task type. AI may help a lot with one kind of work and barely at all with another.
Price the extra speed. What did the more expensive setup buy you that the cheaper one did not?
Write a one-page recommendation: expand, keep, narrow or stop. If the result will drive a big budget, ask an analyst to check your sample size.
ARTIFACT · Decision note
We tested AI on [task type] with [team] for [weeks].
Speed: [result, per person]
Quality: [result]
Cost: [result]
What surprised us:
Decision: expand / keep / narrow / stop
We will check again on [date].Done when: your boss can read the note in two minutes and know what you recommend and why.
Watch-out: going in to prove AI works (or to prove it does not). A good check can come out either way.
Questions to ask your team
Where did AI save you real time this month, and where did it cost you time?
What do you do now with the time you save?
Which tasks would you never give to AI, and why?
Has anything slowed down since we started, such as reviews or debugging?
Are newer engineers still learning the why, or only getting the answer?
Mistakes to avoid
Comparing one team against another instead of people against themselves.
Comparing "before AI" with "after AI" when the process also changed.
Treating the most expensive tool as the best by default.
Using the numbers to rank individuals. That ends honest logging fast.
Reporting one company-wide number when results differ by task.
Your 5-minute action
☐ Open a new doc called AI check v0. Write two lines: the one kind of task you will test first, and the one number that would make you change your mind.
That is enough. A rough plan you can start beats a perfect one you never run. That is your first spark.
Hit reply and tell me: how does your team measure AI's impact today, and what is the part you trust least? I read every answer.
Keep going, one small spark at a time.
Erwin
Sponsored resource Unlock Your Next Income Stream: 100 Side Hustles for Ambitious Founders In a world where financial freedom feels like a distant dream, smart women are building wealth on their own terms.
|
PROMO CONTENT
Can email newsletters make money?
As the world becomes increasingly digital, this question will be on the minds of millions of people seeking new income streams in 2026.
The answer is—Absolutely!
That’s it for this episode!
Thank you for taking the time to read today’s email! Your support allows me to send out this newsletter for free every day.
What do you think for today’s episode? Please provide your feedback in the poll below.
How would you rate today's newsletter?
Share the newsletter with your friends and colleagues if you find it valuable.
Disclaimer: The "Tiny Big Spark" newsletter is for informational and educational purposes only, not a substitute for professional advice, including financial, legal, medical, or technical. We strive for accuracy but make no guarantees about the completeness or reliability of the information provided. Any reliance on this information is at your own risk. The views expressed are those of the authors and do not reflect any organization's official position. This newsletter may link to external sites we don't control; we do not endorse their content. We are not liable for any losses or damages from using this information.