Automation for small and mid-sized business

Automation that learns how your business already works.

Describe the work between your tools. DsEarn builds and tests the connections, then runs approved steps as code. Your team handles the exceptions.

Early access for operations teams.

ORDER TO INVOICEIllustrative workflow

One order. Every handoff.

From a new order to a sent invoice.

  1. Read the orderBring order details into the workflow
    Input
  2. Check and reserve stockApply the inventory rules you approved
    Code
  3. Stock doesn’t match?Ask your team before substituting an item
    Human
  4. Send the invoiceContinue when the exception is resolved
    Code

Your employees are the glue between your software.

Small businesses run on tools that don't talk to each other, and people fill the gaps by hand. DsEarn learns how information moves between your systems, and the recurring work your team does to move it.

OrdersNew order received
InventoryStock checked
FulfillmentReady to ship
People fill the gaps
Re-typing the same data, checking progress, chasing every exception by hand.
Custom connections need maintenance
Each connection needs someone to build, test and maintain it.
Repeated reasoning adds cost
Workflows that call a model at every step accumulate model costs on every run.
Automation tools still need setting up
Someone translates the process into rules — then keeps them true as you change.

You describe the work. DsEarn does the integration.

Start with one recurring process, the tools it touches, and the decisions your team needs to keep.

  1. 1

    Describe

    You talk through how the work happens. DsEarn asks about the parts that are ambiguous, and maps what connects to what.

  2. 2

    Build

    It writes the connections between your systems and tests every one in an isolated sandbox.

  3. 3

    Run

    Approved steps run as plain code. Anything needing judgment goes to whoever should make the call.

  4. 4

    Improve

    Review run history and recurring exceptions to decide which parts of the workflow to change next.

DsEarn

Decide

Today

Inbox

Observe

Processes

Runs

Insights

Your processes

Harbor & Thread — order to close Healthy
Runs / wk
50+
Clean
98%
Human steps
4
Supplier invoices — intake 2 to decide
Runs / wk
31
Clean
94%
Human steps
2
Illustrative product view · Demo data, not customer results.

Build once. Reuse the work.

A model helps build the workflow. Compiled code handles routine runs. Explore how run volume and model exceptions affect estimated infrastructure costs.

Harbor & Thread · order → invoice Estimate
  1. Read the order
  2. Check stock
  3. Reserve the items
  4. Price the shipping
  5. Raise the invoice
  6. Email the customer
  7. Close the order
Model on every run / month $64.00
Compiled workflow / month $0.50

127× lower estimated runtime cost

At 500 runs per month. Illustrative infrastructure model, to be measured in pilots. Excludes the platform plan, human labor and third-party fees. See customer pricing.

500
5% of runs
Estimated monthly cost by run rate
Runs per monthAI agentDsEarn
10$1.28$0.01
100$12.80$0.10
1,000$128$1.01
10,000$1,280$10.07
50,000$6,400$50.35

$0.13

Model on every run · estimate

$0.0010

Compiled run · estimate

7 weeks

Estimated infrastructure break-even

How these numbers are derived

Estimated, and to be measured in pilots. Assumes an agent making 8 model calls per run on Sonnet 5 at $2.00 and $10.00 per million input and output tokens; DsEarn compiling each of seven steps once on Opus 5 at $5.00 and $25.00 per million, then running them as code, with the share of runs that escalate to one model call set by the second control above. Real figures depend on how often your workflow calls a model, and on model choice.

Compiling your workflow is our own one-time cost — $105, not a charge to you. Running compiled code instead of a model costs us so much less per run that our build cost is offset within 827 runs.

Why DsEarn

Learn the process once.
Put it to work every day.

DsEarn maps the handoffs, generates connector code and tests the workflow before it runs. Routine work reuses that code. Your team keeps the judgment calls.

BUILD ONCE
You describeHow the work
gets done.
One tested workflowConnections. Rules. Approvals.
RUN AS CODE
Order 001 ✓Order 002 ✓Order 003 ✓

The same workflow, ready for the next order.

An exception? Your team decides before work continues.

A flat plan. Then it grows with what you run.

Past the plan you pay for what actually happened — each action DsEarn took, and each judgement call it handed to a person instead of guessing.

Every month

$500

the plan

Covers the platform, 10,000 actions and a $50 token allowance.

Past the plan

$0.02

an action

Tokens past the allowance at our cost plus 20%.

When a person is needed

$0.50

a decision

A judgement call routed to your team, with the context attached.

The infrastructure estimates above are our operating costs, not your bill. Pilot pricing is agreed per business before work begins.

Where we are

DsEarn is pre-launch. We're recruiting a small pilot cohort to test real workflows with operations teams.

Maps the process
Apps, APIs, data handoffs and decision rules.
Writes the integration
Generates the workflow and the connector code.
Tests before launch
Validates every connection in an isolated sandbox.
Runs it
Compiles the workflow and executes it on demand.

What you get

Work directly with the founders to scope a first workflow, connect the tools it needs and review the results together.

What we ask

Time with the people who do the work, and honesty about what isn't landing. We're in Philadelphia, Miami and Phoenix, so we can sit with you.

What the cohort will answer

  1. Which work delivers the most value once automated.
  2. What businesses will pay for that.
  3. How to onboard a business repeatably.

Who's building it

Friends since kindergarten, building software together since middle school — through different high schools and different colleges.

Brendan Hirshorn

Brendan Hirshorn

Cofounder

University of Pennsylvania, Mechanical Engineering and Applied Mechanics · Regeneron Science Talent Search Scholar

Applied machine learning and multi-objective optimization to airfoil design and explainable superalloy design. Spearheaded automation at iRocket, building tools for autonomous rocket design.

Chase Ende

Chase Ende

Cofounder

University of Miami, Applied Physics · International Science and Engineering Fair finalist

Studied protein–DNA interactions through computational biology research at Columbia University, and was recognised for computational drug design.

Start with one workflow.

Join the waitlist

Tell us where the work gets stuck. We’ll get in touch about early access.

A few words is plenty. The more specific, the better.

We'll only email you about DsEarn. Privacy.

Built with you

Your tools.
Your process.
Your first workflow.

Work with the founders to connect one recurring process and see how it performs.