One of the things I've considered is that maybe we should have more steps, and fewer outcomes. Or possibly do a simple survey that compares different steps against other steps to see where in the longer job map end users are more frustrated, or where they would prefer to accomplish more.
But, what do you think? There is no single right way
I’m starting to experiment with ChatGPT and I must say I’m impressed. While it’s takes quite bit of experimentation with the prompts, the time it saves is worth it!
Idea: do a workshop over Zoom with the goal to produce a job map and refine it live together? :-)
Here is an example of the output for the step "Schedule Future Communication"
1. Have a clear and specific date and time for the next communication
2. Ensure that the scheduled time is convenient for both parties
3. Minimize the chances of the communication being rescheduled or canceled
4. Ensure that there is adequate time for the communication to be meaningful and productive
5. Enable each party to prepare appropriately for the communication in advance
6. Show respect and consideration for each party's schedule and availability
7. Provide flexibility to adjust the scheduled time as needed to accommodate unexpected circumstances
8. Avoid scheduling conflicts with other important obligations or commitments
9. Maintain consistent and regular communication over time to build a strong and lasting relationship
These do not follow an ODI structure and they break a lot of rules (so do the humans, trust me). But, I think they do get to the heart of the matter. I'm sure there is more tweaking necessary but right now I'm only concerned about "good enough." In fact, these look more human than a desired outcome statement. I wonder how a survey respondent would mess up answering two questions:
1. How important is it to you that you "Avoid scheduling conflicts with other important obligations or commitments"?
2. How difficult is it for you to "Avoid scheduling conflicts with other important obligations or commitments"?
100% I'm very familiar with how difficult it is to craft JTBD survey questions out of strict desired outcome statements, and how difficult it is for respondents to answer them.
Hi Mike, agreed that we may not always be at our best whilst interviewing. While the output looks good, how do we know it is if we haven’t done the work and interviewed real humans. Are these actually the steps people go through? How would Chat GPT know?
Edwards Deming said "A bad system will beat a good human every time."
This is scalable. It's consistent. How do we know the human is good? How do we replicate the human? How would we know these are steps people go through if a human did the map? How many people did they interview? Did they really capture every situation, context? Let me share a secret, these interviews begin with strawman maps created by desk research. At a minimum, we're speeding our way into validation, and at some point we will likely learn that these outputs are pretty darn good.
How do we innovate innovation methods? Is anyone willing to?
There are a lot of professionals invested in interviewing real humans just like there were a lot of laborers invested in digging ditches. We always find better use for those resources. Things we don't have time for now.
I can make it generate desired outcomes statements. It's simply about the proper instructions. If that's the format you really want, fine. The important thing is finding the "object of control", or the success measure. The rest of it not that important and forces an uncommon structure into common brains. The pushback is common, and often intense. Instead, use the language common people use.
One of the things I've considered is that maybe we should have more steps, and fewer outcomes. Or possibly do a simple survey that compares different steps against other steps to see where in the longer job map end users are more frustrated, or where they would prefer to accomplish more.
But, what do you think? There is no single right way
Looking for comments, likes, shares, etc. 🤣
I’m starting to experiment with ChatGPT and I must say I’m impressed. While it’s takes quite bit of experimentation with the prompts, the time it saves is worth it!
Idea: do a workshop over Zoom with the goal to produce a job map and refine it live together? :-)
I’d love if you shared the prompts you used. I found the job steps it produced pretty good but desired outcome statements were a bit “circular”
My prompt for metrics (without all of the formatting mumbo jumbo) is 380 words long
Here is an example of the output for the step "Schedule Future Communication"
1. Have a clear and specific date and time for the next communication
2. Ensure that the scheduled time is convenient for both parties
3. Minimize the chances of the communication being rescheduled or canceled
4. Ensure that there is adequate time for the communication to be meaningful and productive
5. Enable each party to prepare appropriately for the communication in advance
6. Show respect and consideration for each party's schedule and availability
7. Provide flexibility to adjust the scheduled time as needed to accommodate unexpected circumstances
8. Avoid scheduling conflicts with other important obligations or commitments
9. Maintain consistent and regular communication over time to build a strong and lasting relationship
These do not follow an ODI structure and they break a lot of rules (so do the humans, trust me). But, I think they do get to the heart of the matter. I'm sure there is more tweaking necessary but right now I'm only concerned about "good enough." In fact, these look more human than a desired outcome statement. I wonder how a survey respondent would mess up answering two questions:
1. How important is it to you that you "Avoid scheduling conflicts with other important obligations or commitments"?
2. How difficult is it for you to "Avoid scheduling conflicts with other important obligations or commitments"?
See how this works?
100% I'm very familiar with how difficult it is to craft JTBD survey questions out of strict desired outcome statements, and how difficult it is for respondents to answer them.
I'm sure everyone would love if I shared the prompts :)
Hi Mike, agreed that we may not always be at our best whilst interviewing. While the output looks good, how do we know it is if we haven’t done the work and interviewed real humans. Are these actually the steps people go through? How would Chat GPT know?
Edwards Deming said "A bad system will beat a good human every time."
This is scalable. It's consistent. How do we know the human is good? How do we replicate the human? How would we know these are steps people go through if a human did the map? How many people did they interview? Did they really capture every situation, context? Let me share a secret, these interviews begin with strawman maps created by desk research. At a minimum, we're speeding our way into validation, and at some point we will likely learn that these outputs are pretty darn good.
How do we innovate innovation methods? Is anyone willing to?
There are a lot of professionals invested in interviewing real humans just like there were a lot of laborers invested in digging ditches. We always find better use for those resources. Things we don't have time for now.
And I’d since tried it out for myself. Pretty good, though desired outcome statements were a bit lacking. Could have been my prompts
I can make it generate desired outcomes statements. It's simply about the proper instructions. If that's the format you really want, fine. The important thing is finding the "object of control", or the success measure. The rest of it not that important and forces an uncommon structure into common brains. The pushback is common, and often intense. Instead, use the language common people use.