In my early twenties, I ran audio for concerts at a dive bar. The venue was small, the speaker stacks were large, and our philosophy pretty much boiled down to “The louder, the better(opens in a new tab),” often to mixed results.
One lasting result was hearing loss. My ears still ring with tinnitus, and in loud, crowded places I often find myself cupping my hands to hear the person next to me (or bowing out of the conversation completely).
There’s so much noise, and so little signal, that the room might as well be silent.
It’s a bit on the nose, then, that after years of straining to hear through the noise, I now spend my days at Justworks making sure our product doesn’t do the same to our customers.
Not every message matters
Justworks sends a lot of communications to our users. And there’s plenty to say: timecards need approving, health insurance needs renewing, and a new law just took effect in California that needs parsing — because keeping up with California labor law is a job in itself.
But sending communications doesn’t mean we’re communicating effectively. To our customers, the barrage of messages might feel pretty similar to that crowded place: there may be a thousand interesting conversations going on in the world of HR and payroll platforms, but if they can’t figure out which ones are relevant to them, then they’ll probably just leave.
As one of our customers recently lamented:
I wish I had time for it, but I can’t stay up-to-date on ‘all the new trends and best practices related to HR.’ Ideally I’d like to stay up-to-date on only the news, trends, and practices relevant to my business […] My staffing model, my states, my countries, my industry, my HR situation, my payroll situation, my benefits situation […]. The signal ratio would be theoretically as close to perfect as you can get.
We’ve got a lot to say, and people want to hear it — as long as it applies to them.
So how do we send the right message to the right person at the right time through the right medium? That’s the role of the communication layer: the dynamic part of the product that actively speaks to users, layered on top of its static structure.
A shared language for system communications
We can’t design a thoughtful communication layer without first asking a basic question: what do we even mean by “communication”?
To answer that, we started by collecting examples from across the product: an alert to approve a PTO request, a banner reminding admins to enroll employees in FSAs, a launch notification in the “What’s New” drawer.
They looked different, but they shared a common trait: each was triggered by a condition. A condition might be a user action, a date passing, an info change, or a feature release. When a condition occurs, system logic applies an if/then rule to determine which communication to send.
For example: if a newly-onboarded employee’s profile is missing a social security number, then the employee and their admin both receive a system communication to fill in the missing info.
That chain — a condition triggering the product to send a user a message on its own behalf — is what defines a system communication. Collectively, they make up the product’s communication layer.
Building a taxonomy
When we looked across all the system communications we’d gathered, one pattern stood out: does the message require the user to take action, or not?
If the answer was yes, we labeled it a task. If no action was required, we labeled it a notice.
Tasks explicitly require the user to take an action.
Notices share information without requiring action.*
*While notices can include actions, those actions aren’t required for that user.
From there, we identified a handful of useful subtypes within each category, which form the building blocks of our taxonomy:
Task subtypes:
Simple workflow
: A task where the user needs to take action to unblock operations (i.e., approve an employee’s time off)
Complex workflow
: A task where the user needs to create objects or update their attributes (i.e., set up company contributions for benefits)
Notice subtypes:
Awaiting action
: Informs the user when progress is dependent on another person (i.e., an employee is missing personal information)
Announcement
: Lets the user know about a scheduled, active, or recent change or event (i.e., a new employment law has gone into effect)
Value driver
: Makes the user aware of a capability that is available or newly relevant (i.e., we launched a new way to manage cash tips)
Incentive
: Motivates a user by offering a reward (i.e., tell us how your onboarding experience went)
Each type of communication then has attributes:
Trigger
: What is the condition that triggers it?
Action
: Is there an associated action?
Priority
: How urgent is it?
Density
: How much information is there?
Location
: Where does it show up?
Dismissibility
: Can the user dismiss it?
Expiration
: Does it expire without being explicitly dismissed?
Persistence
: Will it be accessible even after it is dismissed or expires?
The taxonomy helps us categorize any system communication and begin shaping the communication layer into something more intentional. But categories alone can’t clear the noise in a crowded room. How do we make sure the messages that matter actually get heard?
Trust sharpens the signal
If the taxonomy helps us define what a system communication is, the trust framework helps us understand how much weight it carries, and whether we’ve earned the right to send it.
Trust is built by understanding the customer’s context. Where are they in their company lifecycle? What are they doing in the product? What other communications have they already received, and through which channels (email, in-app, SMS)? Context gives us the signal for when and how to show up.
Based on what we know about that customer, we can set up conditions that trigger the most relevant communications for them. The more consistently a customer receives messages that feel timely and relevant, the more trust they have in us.
Picture a health bar in a video game. Every time we deliver something useful — a reminder that prevents a missed deadline, an update that saves time — the bar fills. Every time we interrupt with irrelevant stuff, the bar drains.
As we build that bar, the customer then has more trust to spend when we surface bigger opportunities: check out this new feature, expand to more countries, refer a friend. Something that might have felt like an annoying pop-up now feels like a natural, well-timed suggestion, like an invitation from an old friend.
Communications differ in tone and persona. We made this more concrete by anchoring each trust level in a human archetype:
Trust Level 1: Foundational. Like a patient teacher, they offer ever-present, step-by-step guidance.
Trust Level 2: Contextual. Like a hotel concierge, they anticipate what you need and point you in the right direction.
Trust Level 3: Advisory. Like a good life coach, they help you see the big picture, motivate you, and prepare you for what’s next.
Trust Level 4: Aspirational. Like a trusted personal shopper, they know your taste and only surface opportunities that fit your needs and goals.
Trust Level 5: Relational. Like a best friend, they show up to celebrate milestones and occasionally ask for something in return, because the relationship goes both ways.
As trust levels rise, so do the stakes: a PTO reminder carries little risk, while a referral request carries much more. Framing our communication layer through these personas keeps us rooted in the fact that it isn’t neutral but relational.
But frameworks and taxonomies are only useful if they actually shape what we build next. How do we put this into practice today, and how do we evolve it tomorrow?
Where we go from here
We think about this work in three horizons: near (what we’re already planning to do), medium (what we’d confidently do now, if we could), and far (our best idea of the long-term path, right now).
Right now, we’re focused on the near horizon. Using the taxonomy and trust framework, we’re making in-app communications more consistent and more respectful of users’ attention. That means standardizing tone, behavior, and visuals across the product, and establishing these norms in Home (currently known as Dashboard), where the to do / to know distinction shows up most clearly.
The medium-horizon builds on that foundation with infrastructure. A centralized communications API will let us coordinate across channels — extending to email, which is one of the biggest sources of noise today — and unlock user configurability for channel and frequency preferences. It also pushes us to structure communications content more systematically, the kind of groundwork that lets us use LLMs to shape and route messages with more nuance.
The far horizon is naturally less clear, but we imagine a system that understands a customer’s lifecycle holistically and adapts communications intelligently. It’d be flexible enough to use structured content and emerging AI capabilities, but always in service of sending fewer, better messages to the right people, at the right time, in the right medium.
Sending a signal through the noise
Here’s what we’ve mapped out so far. System communications are the messages a product sends on its own behalf. Our taxonomy sorts them into tasks (to do) and notices (to know), giving us clarity. The trust framework helps us gauge how much weight each message should carry, and whether we’ve earned the right to send it. Together, these elements form the beginnings of a communication layer — a dynamic layer that brings the static structure of the product to life.
Now comes the hard part: making this real for our customers. As I learned working at the dive bar, noise is easy; clarity is hard. At Justworks, our communication layer work is about turning noise into meaning by speaking more thoughtfully (and perhaps turning down the volume).
Every message we send is part of an ongoing relationship. If we can say the right thing at the right time, we don’t just make the product easier to use. We build the trust that makes people want to listen at all.
Do you want to design products that help entrepreneurs and small businesses grow with confidence? We’re hiring across our Technology teams, come build with us! Check out our Careers page.
