What can AI do to make Jobs-to-be-Done Research Better
We might be getting close to a complete disruption of the consulting industry
In the Jobs-to-be-Done (JTBD) world we like to say that customers want to get more jobs done on a single platform, or more of the job done on a single platform. They are really almost the same thing because job steps are also jobs themselves.
When applying JTBD methods to research, there are a number of competencies required that are difficult for a single person to develop. To keep things simple let’s just call them
Qualitative Research
Quantitative Research
Data Modeling
Analysis
Strategy
If we look at comprehensive research methods in-depth, what we find is an array of tools being used. We’re cobbling things together. In most cases, we’re using human intervention (with subjectivity and biases built in) to fill in the gaps between Google Docs and Excel spreadsheets.
No two projects studying the same thing (at different points in time) seem to be structured the same way even though jobs are stable over time. This is due to rules being changed (for sound reasons, or convenience) and competing biases between people. When it comes to people, there is no single source of truth, because we all have egos. I’m as guilty of this as anyone.
One of the goals I set a number of years ago was to find ways to make Jobs-to-be-Done research repeatable by people who don’t have the time or money to invest in overly laborious and secretive methods. I also wanted it to be scalable. This could mean doing research at scale (horizontally or vertically) so we could zoom in as needed instead of designing the research with a narrow focus (think Prezi). This could open the door to faster innovation cycles with more predictable results.
What we can’t do today is answer an Executive’s key questions in a timely manner. Let’s be honest, many times they are asking the wrong questions. But what if we could point them to the right questions and have answers for them in our back pocket?
One way I thought we could do this would be to build a portfolio of universal jobs, prioritize them, and develop unique analyses - all in advance. This could elevate the professional who took the time and expense to do this to a lofty advisory position to top executives within an industry. Instead of selling projects (old school) you would be selling answers. Not data. Actual answers.
That may still be a better approach than what we currently have in the niche world of innovation consulting. Imagine the time and expense saved, even if the research has slightly less fidelity than the supposed best-of-breed approaches with bespoke modeling.
Ask yourself, has the success rate of product-launches really improved? Do corporate clients of sophisticated innovation methods actually use the results? Some do, but you could certainly question whether they are able to replicate the exacting detail and mental gymnastics required on their own internal projects, especially in light of the cultural inertia working against them. I’ve seen that with my own eyes. Consulting options are not scalable due to time and cost.
Cutting out the friction
A significant challenge with the adoption of new, complicated approaches (to anything) is getting large groups of people to think and act differently. This is especially true when the results from these efforts are still well over the horizon. They are accountable now and will almost always revert back to an established norm. I’ve tried making the case that converting tools like job maps (with metrics) into capability models is a path forward. However, I would still need to teach corporate groups how to develop these models differently than before. The past has always worked for them, so why would they listen to me?
If we get back to the concept of answers, how might we ensure that we have answers to every question before they are even asked? That would be expensive…even if we constrain ourselves to a specific industry. The answer might lie with AI.
Renato Caliari posted a small portfolio of apps he created using cookup.ai on LinkedIn recently.
I didn’t hold out hope that they would do much, but I was actually slightly amazed at the results. While this only helps us to establish a basic framework, the front end of this process is sometimes where people get stuck the most, so it’s a good start. Here are a couple of examples for you to consider (everything below is AI output, not mine):
User Input
Restore Blood Flow to an Artery
AI Output
MAINTENANCE JOB (it's ongoing) | JOB OF ACTION (within executor's control)
--------------------------------------------------
PREPARE:
Gather tools and equipment needed
Calculate the right medication dosage
Identify suitable access points for the artery
Get informed about the anatomy of the artery
EXECUTE:
Administer the medication
Cut the artery
Insert the tube
Check the pressure
Repair the artery
MONITOR:
Measure the pressure in the artery
Monitor the patient's vital signs
Monitor the patient's response to the treatment
CONCLUSION:
Remove the tube
Close the artery
Apply special dressing to the cut
DOCUMENT:
Record patient's vital signs
Document the medical intervention
Report the results of the procedure
Note: it seems to have excluded a planning step. In the case of this job, that could be disastrous! :)
User Input
Get breakfast on the go
AI Output
MAINTENANCE JOB | JOB OF ACTION (WITHIN EXECUTOR'S CONTROL)
--------------------------------------------------
PREPARE AND PLAN:
RESEARCH BREAKFAST OPTIONS:
Identify available breakfast options.
Analyze time and cost of different options.
Compare nutritional values of different breakfast options.
Identify locations of breakfast options.
ESTABLISH CRITERIA:
Determine the most important criteria for breakfast.
Establish non-negotiable criteria.
EXECUTE:
SELECT BREAKFAST OPTION:
Select breakfast option based on criteria.
Identify the location of the breakfast option.
PAY FOR BREAKFAST OPTION:
Calculate the cost of breakfast option.
Collect payment for breakfast option.
COLLECT BREAKFAST OPTION:
Locate the breakfast option.
Gather the breakfast option.
MONITOR AND ADAPT:
CONSUME BREAKFAST OPTION:
Eat the breakfast option.
Monitor the taste, temperature and satisfaction.
CONCLUSION:
DISPOSE OF BREAKFAST OPTION:
Discard the breakfast option.
Clean up the area.
DOCUMENT AND PREPARE:
REVIEW BREAKFAST OPTION:
Assess the breakfast option.
Record the experience.
Prepare for the next breakfast option.
Are these perfect? No. But then again, I have only seen a few job maps that I thought were perfect (no, not mine). It really doesn’t matter since this is just a starting point and those involved will need to learn how to balance between usefulness and perfection. No one invested in secretive and complex methods are usually interested in disrupting themselves. 😜
Getting more of the Job done on a single platform
What would we need a solution to do to get the entire job done? It probably won’t exist for some time, but we can certainly think about it - and perhaps some dynamic minds will put some energy into moving this forward just like what we’ve seen above. Here are some things we might focus on:
Generate a list of potential jobs to consider based on a problem or key questions
Generate a job map for a selected job (core functional, consumption)
Develop sets of performance metrics relevant to each step in the job
Construct a survey mechanism (what if AI could scour the Internet to essentially replicate what we would get from an end user survey?)
Consume survey (or other) results and construct a data model
Use machine learning to analyze the data model to answer key questions, or questions that AI thinks should also be answered, i.e., what are the top n things we have the capability to do that will have the largest business impact?
Use machine learning to run a large number of segmentation experiments (that humans typically get lazy with) to find the most accessible opportunities - also explain the segments in layman's terms.
Develop unbiased strategies and tactics to leverage our new knowledge
It’s a lot to ask. Say hi to your new best friend👇
Someone needs to teach AI the difference between solutions and the core functional job. The job map for the restore blood flow is entirely in the solution space - not the JTBD space. Sure, you can use JTBD on a product but you are then limited to the innovation on the product, not beyond the product (or the solution set used today). Really enjoy your blogs!