Skip to content

Workflow Gainsharing ​

Workflow Gainsharing is a proposed program. An employee runs the work on their Task Cards, improves it with help from a coach in the app, and the company shares the measured gain with them.

WARNING

Planned. This page describes how the program is designed to work. The Earn code has a time-saved bonus rate with a cap; the run records, the coach, the gain check and the payout choice on this page are not built. The Status box at the end says which part is which. All numbers on this page are examples, not results.

How it fits with Employee Knowledge Compensation ​

Employee Knowledge Compensation pays an employee for describing their work in an accepted Task Card. Workflow Gainsharing pays the employee for making that work better afterwards.

Employee Knowledge CompensationWorkflow Gainsharing
Pays forKnowledge written into an accepted Task CardA measured gain on a Task Card the employee runs
EvidenceThe accepted card and a check periodRun records before and after, confirmed by the person who receives the work
WhenOnce per card, in two partsEach period, while the gain holds
The employee's question"Why should I write down what I know?""Why should I keep using this, and get faster?"

Gainsharing is an established pay practice: a company shares a measured saving with the people who produced it. This page applies that practice to Task Cards.

Why an employee would want this ​

An employee who documents three to five Task Cards and runs them every week becomes the person who keeps those workflows running. Many steps still need a human decision. The employee who runs several workflows well is harder to replace, not easier.

The program gives that employee three things:

  1. A private record of their own progress. They see how long each run took and how that changes.
  2. A coach. An AI coach in the app knows their role and their Task Cards. It shows which step is slow and where a bot could do a step.
  3. A share of the gain. When the same output takes less time, or the employee produces more output, the company pays a share of that value.

How a run is measured ​

A run is one time the employee does the work on one Task Card. DutyGraph needs a start and an end. The design avoids a visible ticking clock.

mermaid
flowchart LR
    S["Start<br/>employee taps<br/>'Start this task'"] --> W["Work<br/>the steps on the card"]
    W --> H["Hand-off<br/>the result goes to<br/>the next person"]
    H --> C["Confirm<br/>the next person<br/>has it and it is right"]
    C --> R["Run record<br/>work time, wait time,<br/>output count"]
    classDef planned fill:#eeeeee,stroke:#888,color:#333,stroke-dasharray: 4 3;
    class S,W,H,C,R planned;
  • Start. The employee taps Start this task. The card shows one quiet line, for example "Started 9:12". It does not count up on the screen.
  • End. The run ends at the hand-off: the employee sends the result to the next person named on the card. No second tap is needed.
  • If the employee forgets. The next time they open the app, the coach asks one question: "Did you finish the weekly numbers? About how long did it take?" The employee picks an answer. DutyGraph marks that run as an estimate.
  • Confirmation. The person who receives the work confirms that it arrived and that it is right. An unconfirmed run does not count toward pay.

Work time and wait time ​

Each hand-off gives two numbers. Work time is the time from start to hand-off. Wait time is the time the result waits before the next person starts their own card.

mermaid
flowchart LR
    A["Check the weekly numbers<br/>work: 25 min"] -- "waits 10 min" --> B["Call the accounts<br/>work: 40 min"]
    B -- "waits 2 days" --> D["Approve the credit<br/>work: 5 min"]
    D -- "waits 20 min" --> E["Update the forecast<br/>work: 15 min"]
    classDef slow fill:#fff3cd,stroke:#a07800,color:#321;
    class D slow;
    linkStyle 1 stroke:#a07800,stroke-width:3px;

In this example every card is fast, but the work waits two days before one approval. The wait between links shows where the constraint is. This uses the same idea as the Work Map's possible constraints (Theory of Constraints, a method that improves the slowest step first). A company that knows the wait times can fix the slowest link first, which is often not the link anyone suspected.

Wait time is a fact about the chain. DutyGraph does not use it to rank people.

From a run to a payment ​

mermaid
flowchart TD
    B["Step 1: Baseline<br/>set before the program starts,<br/>from records the employee<br/>cannot change"] --> R["Step 2: Runs<br/>confirmed run records<br/>over a check period"]
    R --> G["Step 3: Gain<br/>baseline minus now,<br/>after a quality check"]
    G --> CH{"Step 4: The employee<br/>chooses"}
    CH -- "Time back" --> T["Keep the freed time,<br/>under the company's rules"]
    CH -- "More output" --> O["Use the freed time<br/>for more paid output"]
    G --> P["Step 5: Payout<br/>a share of the gain,<br/>through the company's payroll"]
    classDef planned fill:#eeeeee,stroke:#888,color:#333,stroke-dasharray: 4 3;
    class B,R,G,CH,T,O,P planned;

Step 1

Set the baseline ​

Before the company announces the program, DutyGraph records the starting numbers for each card: runs per month, time per run, and output per period. The numbers come from company records where they exist, such as orders entered or invoices sent. Otherwise they come from the "how long" and "how often" answers that the employee and the manager confirmed when the card was accepted.

Step 2

Record confirmed runs ​

The employee runs the card. Each run ends at the hand-off, and the receiver confirms it. Runs are collected over a check period, for example eight weeks.

Step 3

Compute the gain ​

The gain is the difference between the baseline and the check period. It counts only if quality held: the receiver confirmed the work, and rework did not go up.

Step 4

The employee chooses how to use the freed time ​

The employee chooses between time back and more paid output, within the rules the company set. See Two ways to take the gain.

Step 5

The company pays a share ​

The company pays the employee a share of the gain through its own payroll. DutyGraph keeps the record. It does not hold the money.

The formula ​

time gain per year   = (baseline hours per run − current hours per run)
                       × runs per year × loaded hourly rate

output gain per year = (current output per period − baseline output per period)
                       × periods per year × value per unit of output

employee's share     = gain per year × share rate, up to the company's cap

The company sets the share rate, the cap, the check period and the value of a unit of output.

Example (fictional numbers) ​

One employee, one Task Card, "Check the weekly numbers":

ItemCalculationAmount
Baseline25 minutes per run, 48 runs a year20 hours a year
After the change18 minutes per run, confirmed over 8 weeks14.4 hours a year
Time gain5.6 hours × $40 loaded hourly rate$224 a year
Employee's share25% of $224$56 a year

One card is a small amount. The amount grows with the number of cards and with how often each card runs. An employee with five cards that each run daily has a much larger gain. The company also learns where the work waits, which can be worth more than the time saved on any single card.

Two ways to take the gain ​

Freed time has value only when someone uses it. The program lets the employee choose, within the company's rules:

ChoiceWhat happensFits
Time backThe employee finishes their agreed output and keeps the freed timeSalaried roles with a clear, agreed output
More outputThe employee uses the freed time for more work, and the output gain raises their payRoles where more output has a clear value

WARNING

Pay and working-time rules differ by state and country, and they are stricter for hourly employees. An hourly employee who leaves early is normally paid less, so "time back" may not be an option for them. A company should have an employment lawyer review its program before launch.

How people could misuse the program, and the checks ​

MisuseCheck
Work slowly at first, so the baseline looks slowThe baseline is set before the program is announced, from company records or from the numbers confirmed when the card was accepted
Report a shorter time than the real onePay is based on output that the receiver confirms, not on time the employee reports. Estimated runs are marked and count for less
Work faster and let quality dropA gain counts only if the receiver confirmed the work and rework did not go up
Skip a step to save timeThe Task Card's steps and its "done means" stay the standard. A changed step needs the manager's acceptance
Hold back, because a faster time becomes the new targetThe baseline is frozen for a set period, for example 12 months. A gain does not raise the employee's quota during that period
One person slows the chain for othersA part of the share can go to the team that owns the chain, so everyone gains when the whole chain is faster
Work for the number instead of the workThe gain is one input to pay. It is not the only input to a performance review

What employees see, and what they do not ​

  • The employee owns their run times. Run times are private to the employee. The manager sees a gain only when the employee submits it for the check.
  • No stopwatch on the screen. The card shows the start time. It does not count up.
  • No ranking. DutyGraph does not show a leaderboard, activity time, keystrokes or a comparison of people. This is the same rule as for KAR in Employee Knowledge Compensation.
  • Rules first. The employee sees the share rate, the cap, the check period and the baseline for each card before they start.
  • Off by default. The program is a company setting. A company turns it on when it has agreed the rules.

The coach ​

The coach is an AI assistant in the employee's app. It uses the role's profile and the employee's own Task Cards.

  1. The employee opens a saved Task Card and asks, "Help me make this faster."
  2. The coach shows which step takes the most time.
  3. It marks each step as "You" or "A bot could". It never says a bot is running when none is.
  4. It proposes one change to try on the next run.
  5. After the run it shows the result next to the employee's usual time. Only the employee sees it.

When a step becomes an agent, the rules in Employees own the agents built from their cards apply.

Rollout order ​

  1. Run records first. Record starts, hand-offs and confirmations, with pay off. The company gets work time and wait time for each link in the chain.
  2. Private progress. Show each employee their own run times and trend.
  3. The coach. Add the "make this faster" conversation.
  4. The gain check. Add baselines, the check period and the quality check.
  5. Payout and the choice. Add the share rate, the cap, the choice between time back and more output, and the payroll export.

INFO

Status. Built (Partial): the Earn code lets a company set a time-saved bonus rate with a cap and pays only through a payroll file the company runs itself; live payouts are not connected. A confirmed Task Card records "how long" and "how often". Planned: run records with a start and a hand-off, confirmation by the receiver, work time and wait time per link, the employee's private progress view, the coach, baselines frozen for a period, the quality check, the choice between time back and more output, and the team share. A front-end mockup of the phone screens exists; it is not connected to the server.

Source: gitbook/product/roadmap/workflow-gainsharing.md in the DutyGraph repository.

DutyGraph. Pages label each feature Built, In review or Planned.