|
A Follows B: Automatic Processes
Let’s return to Optimize Enterprises’ process for producing its daily
content. Unlike the examples we’ve just been considering, this
process includes no task boards or review meetings. Indeed, there
are almost no interactions or decisions being made at all. Once Brian
Johnson puts a new lesson idea onto the shared spreadsheet, it
moves from one status to the next like clockwork. At each phase, the
relevant people know exactly what’s expected of them.
This style of automatic production process plays an important
role in many knowledge work settings. Not all processes, however,
can be made automatic. For this strategy to apply, the process in
question must produce some output in a highly repeatable fashion,
where the same steps are implemented, in the same order, by the
same people, each time. The types of processes optimized with task
boards, by contrast, are more diverse and dynamic, requiring
collaborative decision making to figure out what tasks to tackle next
and who should be responsible for them.
Consider, for example, the task of putting together a quarterly
budget for your team. This is probably something that can be
reduced to a series of unambiguous steps that are executed the same
way and in the same order each quarter, making the task a good
candidate for automation. Updating your company website, on the
other hand, is probably a project that’s less well defined and will
require more discussion and planning to get right, making it better
suited for a task board approach. The process for adding new client
testimonials to the website, however, probably could be automated,
as it’s highly repeatable. And so on.
Once you’ve identified a process that does seem like a good
candidate for automation, the following guidelines will help you
succeed with the transformation:
1. Partitioning: Split the process into a series of well-defined
phases that follow one after the other. For each phase, clearly
specify what work must be accomplished and who is
responsible.
2. Signaling: Put in place a signaling or notification system that
tracks the current phase of each output being generated by the
process, allowing those involved to know when it’s their turn to
take over the work.
3. Channeling: Institute clear channels for delivering the
relevant resources and information from one phase to the next
(such as files in shared directories).
The daily lesson production process at Optimize clearly follows
these guidelines. It’s divided into well-defined phases, uses a shared
spreadsheet to signal each lesson’s current status, and makes use of
shared directories to transfer files. Automatic processes, however,
don’t necessarily have to rely on software systems. Over the years
I’ve been a professor, for example, I’ve optimized the automatic
process I use to work with my teaching assistants to grade problem
sets for my larger courses. When I write my problem sets, I
concurrently write detailed sample solutions for each of the
problems. I also add some rough thoughts on grading the problems,
capturing what I think would deserve full credit, versus partial credit,
versus no credit.
9
On the day that I post the problem set for my
students, I send these documents to my TAs.
The students hand in their problem sets at the beginning of class,
and I bring them back to my office and leave them in a mail sorter I
had installed on the wall in the hallway beside my door. The TAs will
later come to pick them up. I don’t have to tell them to do this, as
they already know the class schedule and therefore know the days
that problem sets are returned. Once the TAs have the problem sets,
they can start grading. As they assess the students’ answers, they
might update my grading notes to reflect common issues they come
across or particular grading heuristics they decide to apply.
10
When they’re done, the TAs enter the students’ grades into a
shared grading spreadsheet that I set up at the beginning of the
semester, and put the problem sets back in the mail sorter by my
door. On the day I plan to hand back the assignments, I use the
spreadsheet to generate statistics about the problem set scores (e.g.,
average and median scores), which I paste onto a document that also
includes the sample solutions and the grading notes updated by my
TAs. (I discovered through trial and error that detailed sample
solutions and grading notes significantly decreased the number of
students complaining about their scores.) I print out the sample
solutions right before class, then distribute them along with the
graded assignments.
This process always more or less follows the guidelines specified
above. The phases are well defined, the current phase is clear to the
people involved, and we have set channels in place for moving the
relevant resources—problem sets, grading notes, solutions, grades—
where they need to be. Unlike the Optimize example, however, a lot
of this process is physical—involving actual pieces of paper being
moved back and forth. This detail turns out not to matter much. As
long as the phases and communication channels are clear, the
process can be effective.
Like any good automatic process, my approach to problem set
grading eliminates basically all unscheduled communication
between me and my TAs about grading. After I’m done writing a
problem set, my only interaction with the material is to bring the
problem sets to my office mail sorter after the students hand them in,
and then to carry them back to the classroom along with sample
solutions once the TAs are done grading them. The only email
involved in this entire process is when I send the sample solutions to
my TAs (though even this step could be further automated by giving
my TAs access to a shared directory where I keep these solutions).
None of my cognitive energy is dissipated worrying about logistics or
trying to arrange meetings. This might sound superficial—like
someone trying to avoid work—but the reality is that the energy and
attention saved from administrative wrangling can be invested into
activities that actually improve the quality of the class, like polishing
lectures or answering student questions. This advantage is true of
most automatic processes: eliminating unnecessary coordination
does not just reduce frustration, but also increases resources to
invest in the activities that really matter.
Most organizations or teams have some processes that are good
candidates for automation. This is not, however, a transformation to
take lightly, as the overhead in working out all the details of these
processes can be substantial. (It took me a couple of years of
tinkering, for example, to arrive at the process I now use for problem
set grading.) A good approach to figuring out whether this effort is
warranted is to apply the 30x rule. As explained by the management
consultant Rory Vaden, in its original form, this rule states: “You
should spend 30x the amount of time training someone to do a task
than it would take you to do the task yourself one time.”
11
We can
loosely adapt this rule to automatic process construction: if your
team or organization produces a given type of result thirty times a
year or more, and it’s possible to transform its production into an
automatic process, the transformation is probably worth the effort.
Do'stlaringiz bilan baham: |
