What happens on the fit call?

45 minutes. We pick the workflow, check the access and data it needs, and tell you honestly whether it's a fit. If it isn't, we say so and suggest what to do first.

Do we need a developer or an IT team?

No. We do the build. You need one person who owns the workflow and two or three people to review the AI's output in the second week.

Which AI models do you use?

Whichever does your task best at the lowest cost, chosen by testing on your own examples. That is usually a mix of models from Anthropic, OpenAI or Google, or an open-weight model, with routine steps sent to a smaller, cheaper one.

What does the money-back guarantee cover?

If the AI isn't live on your workflow by day 14, you don't keep the invoice — we refund the $3,000.

Is our data used to train AI models?

No. We use model providers under terms that exclude training on your data. Access starts read-only, and write access is added only where you approve it.

What happens after day 14?

You get the report. If the numbers work, the agent keeps running as a Packaged Agent from $500 a month (the setup fee is waived), or under AI Operations if it's custom. Bigger or customer-facing work moves to an Agent Build. If the numbers don't work, you keep the report and the cleaned-up process.

AI Agent Pilot · Start here

One workflow. One AI agent. Fourteen days.

You already know which part of your business bleeds hours — chasing follow-ups, retyping documents, triaging the inbox, compiling the weekly numbers. Pick that one workflow. We put a working AI agent on it inside the tools you already use, measured against your own baseline — not a demo.

See it on your workflow Estimate your payback

Price: $3,000; Fixed. No day rates.
Timeline: 14 days; Three checkpoints
Access: Read-only; Write access only where you approve
Guarantee: Money back; If it doesn't go live

Pick your workflow

See what changes on the workflow you pick.

These are the workflows growing teams pick most. Same price, same 14 days, same guarantee — whichever one you choose.

Sales follow-ups & CRM hygiene

Follow-ups drafted every morning and the CRM kept current, inside HubSpot or Salesforce.

Today

Reps write follow-ups from memory after calls
CRM fields updated days late, if at all
Stalled deals noticed at the pipeline review

With the pilot

Follow-up drafts ready each morning from call notes and email threads
Deal stage, contacts and next step written back for one-click approval
Deals with no activity flagged to the owner in Slack

Example run

Illustrative

07:02Read 23 call notes and 41 email threads since yesterday
07:03Drafted 19 follow-ups in each rep's tone
07:034 drafts mention pricing — held for rep review
07:0431 deal stages and next steps proposed for approval
07:042 deals idle 14+ days — owners notified in Slack

Systems it touches

HubSpot or SalesforceGmail or OutlookCall notesSlack

We measure

Time to follow up
CRM fields complete
Rep hours per week

Human checkpoint

Every email is a draft. A rep reviews and sends it.

Also available as a Packaged Agent

Don't see yours? If it happens on a screen, repeatedly, and eats hours, it's probably a fit. Bring it to the fit call and we'll tell you honestly.

What you get

Five things in your hands by day 14.

Not a slide deck. A working system on your real work, and the evidence to decide what happens next.

Hours / week

Turnaround

Error rate

Day 2

Baseline scorecard

Hours, turnaround and error rate on your real operation, plus a plain read on what your data can and can't support.

Live on real work

CRM · read + draft
Inbox · read
Slack · notify

Day 5

A working AI in your tools

Connected through official APIs or MCP connectors, running on live work while your team reviews every output.

draft #56Approved

draft #42Edited

draft #84Needs review

Days 5–10

Review queue and audit log

Every action logged. Anything uncertain routed to a person, and the AI tuned daily on their feedback.

Before

After

Day 14

Before-and-after report

Baseline vs. actual, adoption, cost per task and a straight go/no-go recommendation.

1Expand to quotes

2Clean vendor master data

3Move to AI Operations

Day 14

Next-step plan

What to expand next, what it would cost, and what to fix in your process or data first.

How it plugs in

Your tools stay. The AI works between them.

Nothing to migrate and no new app for your team to learn. The pilot reads from the tools the workflow already lives in and writes back only where you approve.

01
Your tools
Where the work already happens
CRMEmailDocs & drivesAccounting / ERPSlack or TeamsSpreadsheets
02
Connectors
Read-only first
Official APIsMCP serversWebhooksScoped credentials
03
The AI agent
One workflow, clear limits
Step-by-step workflowTool permissionsRetries & fallbacksConfidence thresholds
04
Models
The smallest one that passes
AnthropicOpenAIGoogleOpen-weightRouted by cost

Across every layer

Human review queue

Uncertain or risky items wait for a person.

Across every layer

Evals & tracing

Every run logged and scored against your examples.

The 14 days

Two weeks, three checkpoints, one number.

The pilot exists to answer one question with evidence: does this pay for itself on your operation? That's why the baseline comes first.

D10

D11

D12

D13

D14

Connect & measure

Read-only access; baseline on your real work

D1–2

Build & connect

Workflow, connectors, review queue

D3–4

Live with your team

Real work, every output reviewed, tuned daily

D5–10

The receipts

Baseline vs. actual, go/no-go

D11–14

Checkpoints

Baseline scorecard

Live on real work

Report & decision

D1–2Connect & measureRead-only access; baseline on your real work
D3–4Build & connectWorkflow, connectors, review queue
D5–10Live with your teamReal work, every output reviewed, tuned daily
D11–14The receiptsBaseline vs. actual, go/no-go

Payback

What would it be worth on your workflow?

Move the sliders to your numbers for a rough payback estimate. The pilot replaces the estimate with the real figure, measured on your own work.

Hours per week your team spends on this workflow15 h

Loaded cost of an hour of that work$45

Share of the work the AI takes on50%

An estimate to frame the decision, not a promise. The pilot measures the real figure on your own work in days 1–2 and 11–14.

Hours back per month: 32
Value per month: $1,461
Pilot pays for itself in: 9 weeks; Against the $3,000 fixed price

Who does what

What we need from you, and what we take off your plate.

You provide

One workflow owner, about two hours a week
Read-only access to the tools the workflow lives in
10–20 real examples of the work, good and bad
Two or three reviewers in the second week

We provide

A senior engineer on the build from start to finish
Connectors, prompts, guardrails and the review queue
Baseline and day-14 measurement
Daily tuning from your team's feedback

Not included

More than one workflow (that's an Agent Build)
Customer-facing apps or custom software
Large data clean-up or migration projects
Running costs after day 14 (see Packaged Agents)

Is this you?

Built for teams that run a business, not a dev team.

A good fit if…

You have 30–250 employees and no dev team
The workflow lives in tools like a CRM, inbox, spreadsheets or an ERP
It's repetitive, it happens on a screen, and it eats hours every week
Someone owns it and wants proof before a bigger AI bet

Not the right start if…

You need a customer-facing agent or app — start with an Agent Build
The process changes every week or isn't written down anywhere
The data lives on paper or in systems with no API or export
You need a regulated-data approval (HIPAA, PCI) before anyone can touch the data

After the pilot

The day-14 report is the map for what's next.

If the numbers work, the agent keeps running from $500 a month, and bigger work moves to an Agent Build. If they don't, you keep the report, the cleaned-up process and your money. How we build every system

FAQ

Questions we hear on fit calls

What happens on the fit call?: 45 minutes. We pick the workflow, check the access and data it needs, and tell you honestly whether it's a fit. If it isn't, we say so and suggest what to do first.
Do we need a developer or an IT team?: No. We do the build. You need one person who owns the workflow and two or three people to review the AI's output in the second week.
Which AI models do you use?: Whichever does your task best at the lowest cost, chosen by testing on your own examples. That is usually a mix of models from Anthropic, OpenAI or Google, or an open-weight model, with routine steps sent to a smaller, cheaper one.
What does the money-back guarantee cover?: If the AI isn't live on your workflow by day 14, you don't keep the invoice — we refund the $3,000.
Is our data used to train AI models?: No. We use model providers under terms that exclude training on your data. Access starts read-only, and write access is added only where you approve it.
What happens after day 14?: You get the report. If the numbers work, the agent keeps running as a Packaged Agent from $500 a month (the setup fee is waived), or under AI Operations if it's custom. Bigger or customer-facing work moves to an Agent Build. If the numbers don't work, you keep the report and the cleaned-up process.

$3,000 fixed · 14 days

Two weeks from now, you'll have the numbers either way.

$3,000 fixed. Money-back guarantee. A free fit call to pick the right workflow before you spend a dollar.

Book the fit call