Accelerating JTBD Research with AI
by Jim Kalbach (Feb 2024)
AI and JTBD research complement each other well. AI can significantly enhance research efforts by providing a wealth of information without compromising the validity of results or interactions with real customers.
First, the JTBD framework’s solution-agnostic nature allows AI to draw from a vast array of resources. Services like ChatGPT offer information that isn’t confined to a single product or service, broadening the research coverage.
But also, JTBD provides a precise language for articulating insights gathered in the field, with distinct categories and formulation rules. Large language models, due to their structured approach to language, can easily generate consistent results.
The effectiveness of AI largely depends on the prompts. While I am not a prompt engineering expert, I have found the following general prompt structure and flow to be effective:
- Frame: Begin by stating your objectives and providing context for your inquiry.
- Definition: Offer a clear definition of the concept you’re exploring.
- Example: Provide concrete examples for AI to base its responses on.
- Request: Clearly articulate the desired output, its format, the number of items needed, and any other relevant factors.
Here’s an example prompt, in this case for identifying job steps after defining a target job, using the JTBD framework:
EXAMPLE PROMPT:
- Frame: I’m working with the jobs to be done (JTBD) framework and would like to create a job map for insured patients (job performers) trying to find a healthcare provider (target job). I’d like to understand how job performers get that job done.
- Definition: A “job map” is a chronological sequence of steps. The steps are the sub-objectives that are required in order to complete the job.
- Example: For instance, in the field of buying and selling homes, the job map for a home buyer trying to shop for a new home might look like this: determine new home criteria, identify a source to shop, learn about the market, seek available homes, compare available homes, select a home, place a bid.
- Request: Please list 12 job steps in bold followed by a one-sentence description of that job step in that field. Each job step should begin with a first-person verb and be singular, with no conjunctions (i.e., no “ANDs” or “ORs”).
Keep in mind that using ChatGPT is iterative. You can chat back and forth to refine the outcomes using a critical eye to judge accuracy and appropriateness. Question the categorizations and formulations, and request re-generation of results as necessary.
You may also need to adapt AI-generated information for use in your data collection and models. Tweaks and modifications are usually necessary, but the speed with which AI can produce a comprehensive data set is remarkable.
The JTBD Canvas 2.0 provides the key elements in the JTBD framework you’re looking for. Our new online course, “Foundations of JTBD,” teaches you in detail how to use the canvas along with our all-new 5-stage process.
Phase 1: Scoping your JTBD landscape
The canvas’s upper section focuses on top-down decision-making elements. Given your current situation, project and strategy, it’s up to you to determine the target job to focus on. This sets the stage for your research.
While AI can add valuable insights to team discussions, it’s crucial to also bring key stakeholders together for alignment. Do not solely rely on AI for this step.
A precursor to using AI for JTBD research is to consider your field in terms of the core benefit you provide. This is generally different than an industry category. For example, in “real estate,” it’s better to define the field as “buying and selling homes.” Or, for HR professionals, the field could be expressed as “recruiting and retaining employee talent.”
Identifying a target job then involves three steps:
- STEP 1: Identify potential job performers and select one to focus on.
- STEP 2: List the objectives these performers have (referred to as related jobs on the canvas) and choose one as the target job.
- STEP 3: Consider the aspirations associated with the target job.
Below are three prompts created for ChatGPT 4.0 that correspond to these steps, as outlined in the JTBD Canvas 2.0’s upper section. Each prompt adheres to the structure described earlier, with variables indicated by double brackets for you to customize with your specific context.
PROMPT 1: Define the job performers
Who are job performers are in my field?
I’m working with the jobs to be done (JTBD) framework and would like to find job performers in the field of [[field]].The job performers are the key actors or roles that have objectives to accomplish within that field.
For instance, in the field of buying and selling homes, job performers would include, but are not limited to: home buyers, home sellers, real estate agents, lawyers, inspectors, loan officers, mortgage lenders, titling agents, and neighbors.
Please list [[n]] job performers in bold followed by a one sentence description of their role in that field. Each job performer should be singular in nature, with no “ANDs.”
PROMPT 2: Explore related jobs and select a target job
What jobs are people trying to get done in my field?
I’m working with the jobs to be done (JTBD) framework and would like to find the top jobs that [[job performers]] are trying to get done in the field of [[field]].A “job” is an objective that the job performer strive to achieve in that field.
For instance, in the field of buying and selling homes, job would include, but are not limited to: shop for a new home, bid on a new home, close on a new home, move homes, sell old home, renovate new home, and settle into new home.
Please list [[n]] jobs in bold followed by a one sentence description of that job in that field. Each job should begin with a first person verb and be singular in nature, with no conjunctions (i.e., no “ANDs” or “ORs”).
PROMPT 3: Consider aspirations
What aspirations do job performers have while completely their job?
I’m working with the jobs to be done (JTBD) framework and would like to find aspirations that [[job performers]] are trying to [[target job]].An “aspiration” is what we call a “be” goal, or what the job performer wants to become by getting the job done. They are typically not measurable and have many solutions that could contribute to achieve them over time.
For instance, in the field of buying and selling homes, aspirations might include, but are not limited to: Become a proud homeowner, be more content with home life, Provide a better home life for my family.
Please list [[n]] aspirations in bold followed by a one sentence description of that aspiration for the target job. Begin each with the verb “be” or “become” where possible, but other starting verbs are possible. Aspirations should be singular with no conjunctions (i.e., no “ANDs” or “ORs”).
Phase 2: Investigating the elements of JTBD
After scoping your investigation, the next step involves filling out the bottom portion of the JTBD Canvas 2.0. This section includes four key elements: job steps, outcomes, emotional and social factors, and job differentiators.
It’s crucial to remember that the output from AI should not be accepted at face value. For example, when generating job steps, AI might misclassify information or mistakenly include an outcome within a step. The phrasing may also be imprecise.
More importantly, it’s essential not to bypass direct engagement with job performers. AI cannot substitute for firsthand insights gained in the field by meeting people face-to-face. This interaction also presents an opportunity to involve colleagues, allowing them to hear directly from individuals about their experiences in accomplishing a job. Failing to engage with the people you serve would undermine the goal of becoming a more human-centric organization.
Nevertheless, AI significantly aids the research process in several ways. First, it equips you to be a more knowledgeable investigator. You’re not starting from scratch but using AI-generated hypotheses as a springboard for your fieldwork, though it’s important to validate these hypotheses.
Second, AI can reduce the number of required qualitative interviews. Instead of needing to conduct a dozen interviews, you might find that six or fewer suffice. Tasks that could take weeks might now be completed in days.
Below are prompts corresponding to the aforementioned elements, located in the lower part of the JTBD Canvas 2.0. You will need to fill in the specific variables denoted by double brackets.
PROMPT 4: Create a job map
How do job performers get the job done?
I’m working with the jobs to be done (JTBD) framework and would like to create a job map for [[job performers]] trying to get [[target job]] done.A “job map” is a chronological sequence of steps. The steps are the sub-objectives that are required in order to complete the job.
For instance, in the field of buying and selling homes, the job map for a home buyer trying to shop for a new home might look like this: determine new home criteria, identify source to shop, learn about the market, seek available homes, compare available homes, select a home, place a bid.
Please list 12 job steps in bold followed by a one-sentence description of that job step in that field. Each job step should begin with a first-person verb and be singular in nature, with no conjunctions (i.e., no “ANDs” or “ORs”).
PROMPT 5: Outcome statements
How do job performers determine the success of getting the job done?
I’m working with the jobs to be done (JTBD) framework and would like to generate outcome statements for [[job performers]] trying to get [[target job]] done.An outcome statement is a personal success metric defined and decided on by the job performer.
For instance, in the field of buying and selling homes, outcomes the job performer is looking for might include: “minimize the time it takes to identify new homes to purchase” and “minimize the chance of having to do renovations before moving in.
Please list 50 outcome statements in bold. Outcome statement must have the following structure:
- Begin with “minimize” followed be an object (i.e., the unit of analysis). and then after that a qualifier as needed to provide enough context for the statement to be specific and understandable.
- Don’t include conjunctions (i.e., no “ANDs” or “ORs”),
- No reference to technologyPROMPT 6: Emotional and social aspects
How do job performers feel while getting the job done? How do they wish to appear to other?
I’m working with the jobs to be done (JTBD) framework and would like to generate emotional and social aspects for [[job performers]] trying to get [[target job]] done.An emotional aspect reflects the feelings a job performer has while executing the target job. A social aspect is how they wish to be seen while getting the job done.
For instance, in the field of buying and selling homes, emotions the job performer has might include: “Feel confident about new home selection” or “Avoid confusion with the seller.” Social aspect might include “Appear competent” or “Avoid appearing overly eager.”
Please list [[n]] top emotions. Formulate them with the following rules:
- Begin with “Feel…” or “Avoid feeling…”
- Don’t include conjunctions (i.e., no “ANDs” or “ORs”),
- No reference to technologyThe list [[n]] social aspects of the target job. Formulate them with the following rules:
- Begin with “Appear…” or “Avoid appearing…”
- Don’t include conjunctions (i.e., no “ANDs” or “ORs”),
- No reference to technologyPROMPT 7: Job differentiators
What factors make the biggest differences in getting the job done?I’m working with the jobs to be done (JTBD) framework and would like to generate job differentiators for [[job performers]] trying to get [[target job]] done.
Job differentiators are those factors that make a difference in how the job might get done by different people under different conditions. They are the contextual descriptors of the target job.
For instance, in the field of buying and selling homes, job differentiators might include, but are not limited to: if buying a home for self versus with a family, if buying close by (e.g., in same or neighboring city) versus far away (e.g., cross county or internationally)
Please list [[n]] job differentiators. Formulate them with the following rules:
Start each with “if”
Indicate the conditions that exist on a scale with two end-points separated by the word “versus”
JTBD Canvas GPTs
The prompts provided can be employed individually based on the specific needs of your project. For instance, if the task at hand involves creating a job map for a team discussion or workshop, the job map prompt alone may suffice.
To assist you in navigating both the scoping and investigation phases of your JTBD research with the canvas, we have developed two separate GPT models using ChatGPT 4.0. Access to these models requires a subscription to ChatGPT 4.0; however, the prompts mentioned earlier are compatible with ChatGPT 3.5 and other AI tools.
JTBD Canvas GPT 1: Scoping Your JTBD Landscape
JTBD Canvas GPT 2: Investigating a Target Job
Currently, these GPT models are in a beta phase and are available for use at no cost. We plan to continue their development based on user feedback and advancements in the field
.The primary goal of these tools is not to eliminate direct interaction with the people in your target markets. Rather, these GPT models are designed to expedite the process of completing the JTBD Canvas 2.0, setting the stage for more informed fieldwork.
The prompts are also fairly general under the assumption that you’ll be able to discern the output with a critical eye and re-formulate elements as needed. Ultimately, the integration of AI-generated insights with field data is essential to develop a comprehensive understanding of your research subject.
We are eager to engage with our user community and benefit from your experiences, whether they are successes or challenges. We encourage you to share your feedback and findings with us at hello@jtbdtoolkit.com.
— — — — — — — — — — -
See our new online course, “Foundations of JTBD”, for more details on the JTBD Canvas 2.0 and our all-new core process for the JTBD framework.
