Jobs To Be Done Within the 4 Walls
Jobs To Be Done Within the 4 Walls
Applying the JTTB framework to jobs that cross departments requires details on how teams work together. Sounds obvious, but it's not that simple. Read: "Everything Starts Out Looking Like a Toy" #196
Hi, I’m Greg 👋! I write weekly product essays, including system “handshakes”, the expectations for workflow, and the jobs to be done for data. What is Data Operations? was the first post in the series.
This week’s toy: an endless, auto-generated “city”. The obvious uses for this tech is to produce realistic screen savers, background polygons or scenery for games, or other simulacra. What’s weird is that the level of detail possible now in a “background” process that builds itself is … pretty amazing. Add generative transformers to something like this and the possibilities for world generation in games or art is impressive. Edition 196 of this newsletter is here - it’s April 29, 2024.
Feeling lost in a sea of sales tools? Apollo.io has your entire GTM stack covered.
A true end-to-end sales platform with a huge (and very accurate) lead database, enrichment, outreach, call intelligence, scoring, calendaring, and email automation. They integrate seamlessly with your CRM and existing workflows. The quality of Apollo’s data is unmatched; they are ranked #1 for contact and company data accuracy on G2 with over 6800 reviews and a 4.8 star rating. Book a meeting to get started for free.
If you have a comment or are interested in sponsoring, hit reply.
The Big Idea
A short long-form essay about data things
⚙️ Jobs to be Done within the 4 walls
We recently had an integration fail at work for unknown reasons. One provider that was supposed to get webhooks from another system we used simply stopped working. Talking to their support staff revealed that they didn’t know why the system stopped working and responding to webhooks either.
The fix? We repointed the webhooks to another, similar integration bus and built a new workflow to handle the underlying job that needed to be done: notifying sales teams about the details of incoming lead conversations.
This little story points to one of the challenges of building solutions and integration to solve business problems: you have to know the details of the problem you’re trying to solve to make a meaningful difference in the outcome provided by that integration.
Clayton Christensen has famously called this “Jobs to Be Done”, writing:
People don’t simply buy products or services; they pull them into their lives to make progress. We call this progress the “job” they are trying to get done, and understanding this opens a world of innovation possibilities. (Christensen Institute)
Building integrations requires codifying the work of the organization into a series of tasks and steps that result in the right work getting done at the right time. That means that a key requirement to building an effective process is observing how work is getting done today and determining which things we want to keep doing and which things we want to change.
Knowing How We Work
We need to know how we work today before we can identify or imagine how we want to change that in the future. It’s tempting to start coding before you know what’s needed. When I do, I re-read this essay on programming that reminds us that you need to spend 11/12ths of the total time thinking about what needs to be done before the actual work gets done.
If you can write the pseudo-code for what your integration does and you have feedback from the people who will use it, you are a lot closer to a workable solution.
“Knowing how we work” today also means comparing how people talk about the process to the quality of the data captured by the process. When you ask about duplicate accounts in your CRM, is there a standard way to identify them automatically or are you relying on humans to follow a process to find dupes?
Be careful what you automate. If you over-index on changing fields from one behavior to another without first thinking about the behavior you want to reinforce, you may get high data quality without getting behavior changes.
“Within the 4 Walls” refers to the interactions that people have with one another in a company. Now that many companies include a virtual component talking about this happening within a building sounds a little anachronistic but it also gets you thinking about how people work together across tasks and time zones.
In programming terms we would refer to these issues as “race conditions” or “sequencing” - people doing the same thing at the same time - and the results as “data quality”. Most people describe it as a “critical path to action” and the end state of the information.
Breaking tasks down into sequences
Whether you end up doing this with a machine process or people, you’re designing sequences that start with a Job to Be Done and end with data (and objects) in a different state. Building these tasks for jobs that cross departments and functions is similar to building a task for a single function.
When designing the Jobs to be Done “within the four walls”, i.e. inside the company, you need the regular steps you would take:
Determine the job to be done
Document what’s going on where the data flows
Writing a standard operation procedure
Testing the usability of the result (you might need to edit)
Step four highlights the importance of the human in the loop. When you automate a process, you often find out that the humans want to change it or have a part of the task they forgot to document.
This context is crucial for understanding what to do next. If you automatically assign accounts to a team and miss the informal discussion that happens on that team to divide up assignments, you might disrupt something that’s working perfectly. In this case, automation is great for making sure that something doesn’t fall through the cracks but isn’t the best first step.
Why is this important for AI
Increasingly, almost all jobs to be done will contain steps completed by machine learning. The catch? You typically won’t see the parts of the job done by machine learning. But you need to assume that ML or AI are affecting the results you’re seeing in most products even if you don’t explicitly use them.
What to do? Use your design thinking, of course. Read about designing prompts rather than engineering them.
Expect that you will be using AI eventually to do some of the more computationally intensive tasks of your workflows, especially the ones that involve fuzzy matching of tasks and inference that takes into account past interactions.
For example, when you are comparing duplicate accounts, you'll want to look beyond the name of the account and also consider other factors. To see if the person who started the account has ever interacted with you before, you may need to consider the nearness of their email account with other emails they have used.
In this example, you don’t want AI to invent new people, and you do want to comb through many similar examples and provide a similarity score to aid in human decision-making.
If you start with a strong model for the outcome of your process, you’ll have a better idea of the steps where AI can assist and where it’s a pretty bad idea today to delegate human decisions to a deterministic process.
(Note for future you: multi-shot approaches or agentic processes for AI may change this assessment drastically. Keep an eye on this space and check back in a year or two.)
What’s the takeaway? The context for Jobs to Be Done that cross departmental boundaries is difficult to establish and critical for good integration. Start with the human first, and then consider how the capabilities of AI can help.
Links for Reading and Sharing
These are links that caught my 👀
1/ Finding Deep Fakes - There are some tells in “deep fake” information that’s being shared online, but the fakes are getting harder to detect. Here are some common AI detection tools along with suggestions about why they are missing on some common content.
2/ How do visual transformers work? - This is a clever way of demonstrating how visual transformers encode, store, and retrieve visual tokens for use in AI.
3/ The next generation of CarPlay - Casper Kessels has a few really interesting ideas on the future of Apple’s CarPlay and how to maintain brand equity in a world where everyone wants to control the user experience. It’s a paradox: how do you deliver an Apple-like experience in someone else’s wrapper without taking over the entire function?
What to do next
Hit reply if you’ve got links to share, data stories, or want to say hello.
The next big thing always starts out being dismissed as a “toy.” - Chris Dixon