Making Sense of Sticky Notes — AI 4 Agile

Let's assume that your team is using

Scrum to create value for customers.

However, they have that trouble with

sprint goals. Like so often, um they

basically first pick the work and then

somehow attach a sprint goal on top of

it because it's part of the Scrum guide

and they're doing scrum, right? Um of

course, this is not how it is supposed

to be. You know uh just as a brief

reminder um sprint planning works in the

following way. Product owner brings the

business objective to the sprint. What's

most valuable the team could accomplish

over uh the course of the sprint? Then

collectively

the scrum team agrees on a sprint goal.

You know taking everything into

consideration. the capacity, who is

there, are people joining, people

leaving, do they have all the tools, are

there any dependencies,

are they skilled in the area they're

supposed to work, etc.

The point is that the sprint goal is

actually the beginning for the rest

because now that they have a sprint goal

that they can accomplish within the

sprint that um basically supports the uh

notion of the product owner what would

be uh most valuable from a business

perspective.

Um and the developers commit to the

sprint goal. they now are in a position

that they can pick the work that they

deem fit and necessary to accomplish the

sprint goal. That's it. Uh the the whole

balance in the in the scrum team, you

know, checks and balances and no one has

any authority to tell anyone else what

to do, when to do it, and how to do

this. This is really the the core of

self-management. Okay. So, product owner

brings a business objective, scrum team

collectively creates a sprint goal,

developers commit and pick the work. So

of course completely useless that you

first pick the work and then for

whatever reason because you probably

feel better or you believe you have to

do this because you're making scrum work

here um that you somehow bold a

arbitrary sprint goal on top of a

complete use of time. So um you try to

help your team and the team uh agrees on

running a retrospective and you trigger

a few prompts here to figure out okay

what is the background of of this this

issue here what are the root causes and

uh now we are actually closing in on the

use case I like to um share with you

when a large language model can actually

make use of unstructured data and help

us interpret this. So let us change

here. Um I created um the prompt in

collaboration with chat jpg as you might

have guessed by now. Um the link uh is

in the description below where you can

see how I did this and uh let's have a

look at this sprint uh goal prompt that

we have here. So um we have a role right

and uh we have a task analyze the

stickies here and uh we provide the

context you know team is struggling with

meaningful sprint goals um goals are

often created after selecting work items

and um now we get more into the details.

Step one extract the data. Extract all

sticky note text from the whiteboard.

preserve the position so that it can

group by color or placement

categorize by retrospective prompts.

There are seven prompts here in the in

the um in the graphic and we just list

them here. So um the previous job on

creating this prompt identified those uh

prompts from the retrospective

and um then we have you know uh the

output you know each sticky notice is

part of a corresponding group you know

flag them as ambiguous if you can't

place them correctly analyze the

response you know summarize key themes

for each of the seven groups I patterns

and misconceptions actions. Highlight

signals of progress or insight. Suggest

one or two coaching opportunities or

follow-up actions per category. Output

format group sticky notes. Summary of

themes. Inside coaching opportunities,

use bullet points. Be concise but

insightful, you know. Um, important

notes. Okay. First of all, I'm attaching

a graphic here. And on the other side,

it aims to shift the team's mindset

towards sprint goals driven planning,

not task first. Right? Emphasize scrum

guide align thinking and detect anti

patterns. Be constructive and solution

oriented.

Now, this is the second time I'm

recording this video and the first time

it was completely rubbish. The whole

outcome, it was disastrous.

And you will find a link to that one um

in the in the in the notes below.

And I wondered, okay, what was going on

here? And uh before I recorded this one

here, I ran an experiment because I

turned on the temporary chat, you know,

and by comparison to the uh junk chat I

had, I also changed the model from chat

uh 40 to 04 mini. And um 10 minutes ago

was working just fine. Let's hope for

the best when I click the go button.

So, thinking, thinking, thinking. This

is good. Transcribing sticky note text.

Yes. So, that would be awesome.

Okay.

With complex work.

So eight

four. Four seems to be the few

three. So there there's missing one.

This one seems correct.

How mess the sprinkle help team

understand the valuables.

Um if you don't ah again number six

there's there's there's one that is very

often missing. Just in the last one I

did not record

it was correctly placed and okay. So um

what you would do now before you

actually continue working on this you

would compare um the um the

transcriptions

a uh and more importantly typically it

gets the text right um but more

importantly you want to also understand

okay where are they placed in what

vicinity right um so think of it this

thing as as color blind um so it's not

that easy um and even In this point we

have um so a vicinity mapping here it's

also not that clear. I mean OCR is not

the the the top job of any large

language model. So bear with it a bit

and uh you would now go through step by

steps and point to the errors. So for

example what we are missing in number

six is lack of focus and also where is

lack of focus?

typically it's somewhere else

probably

okay

lack of

focus. Oh, it didn't catch that one in

the first place. Well, that's

interesting. I never had that before.

Anyway, you know where this all all is

going, right? Um you would now reach out

to the to the model and say, "Hey, you

are missing something here." And then

you would ask it to actually um re

recreate um the um allocation of sticky

notes to the individual points and then

you take it from there.

Generally this works well. I mean in

this situation okay insights coaching

yada yada yada yada. Okay. Now you could

ask okay

could you please

draft a retrospective?

Another retrospective um based on the

analysis

question mark.

What should the team

focus on to improve?

Okay. And then you're basically off to

the races. So, um,

yeah. Let's go first edition duration 90

minutes. Check in. Okay. So, um, check

this out when you uh when you go through

uh when you open this chat yourself and

uh think about this and oh, by the way,

I can't share that with you because it

is a temporary one and temporaries I

can't share with you. Okay. Okay.

Anyway, so um nevertheless, I wanted to

to um mention this that you can use

large language model also to analyze um

graphical information. I mean we used it

before to uh populate our alignment

canvases and um create u the content for

test on learning cards for example. But

you can also use it to actually um

document and analyze sticky notes from

exercise that you have. Of course, it's

a bit more tricky if you have

handwritten um sticky notes. Um if

people are using the wrong pens or

pencils or whatever utensil they use to

write, you know, you need to have clear

black lines. There need to be a decent

uh um a decent level of contrast between

the the the sticky note itself and the

writing on top of it. But um if you

figure this out correctly, uh it's

actually a good way to um uh engage a

large language model to help you analyze

these things and then take it a step

further.

I hope you give it a try and practice.

Um, of course, the online version is a

bit simpler. Um, also because typically

large language the uh the online

whiteboards allow you to export the text

of this. This is just um an exercise in

showing what you actually could do here.

And by the way, the exercise here is

part of the classic advanced scrum

master class when we talk about the

importance of sprint goals. Just it's

real data, so to speak.

Have fun.