17: Ducks in a row
In recent emails, I’ve been sharing the thinking behind designing a performance guidance system. A system that tells us what to do, when to do it and how it should be done. Our satnav for work.
And what we’ve uncovered throughout the conversation are two types of workflows:
conditional workflows i.e., workflows triggered by external conditions, and
cyclical workflows i.e, workflows triggered by the passing of time.
The examples I shared were ‘making a video’ (conditional on a client requesting a video), and ‘submitting my tax return’ (cyclical based on three months elapsing).
Now, regardless of which type of workflow we use, defining workflow steps is not only helpful to improve our own performance, but also to help others.
And taking it one step further - workflows also give us the ability to delegate.
We often hear people complaining about being busy, but when asked why they haven’t delegated their work, they reply “it’s just faster to do it myself” - but that’s only because the workflow hasn’t been defined!
And yes, it does take some time up front to do this (did I mention I took a week off work to build my system?! I dread to think how much time I’ve spent tweaking - but probably not as much as it’s saved me).
So, I completely sympathise - in a world that’s speeding up, who has time to document their workflows?
But avoiding this will no longer be an option.
As we move into a world in which AI can do many of our tasks, the business will require high value workflows to be defined so AI can be leveraged.
Point in case - agents. To build an agent, you have to tell it what to do. You must give it specific instructions, context and guardrails for each task, so it delivers satisfactory outcomes.
And this is the key mistake people are making - they assume AI will magically figure out their workflows.
But they won’t.
They can’t!
Think about it. Think about everything that sits inside a task, but that isn’t documented. All the micro-decisions, exceptions, trade-offs, judgement calls, nuance, contextual hints. AI can guess - and it can guess pretty well. But if we want reliable outcomes, all of this stuff needs to be made explicit.
And so, not only is what we’re discussing here essential for improving human performance, it’s also essential in the design of a hybrid-agentic future - which is what every business leader is being told they need to design for.
Which is why I’m so stoked to be involved in L&D right now.
There will be disruption.
But for those of us who can see what’s unfolding, and who’ve committed to doubling down on the first principles of performance improvement, it’s a very exciting time to be a performance improvement professional.
And I’m excited to be on the journey with you.
Yours,
- Ant
