The workflow your team does every day runs itself.

Record it once. We build the automation in 72 hours — tested on your exact portals, reviewed by an engineer, running in production. Your team keeps every decision. Type-2 keeps the repetition.

Windows 10+ · 7 companies · June start · Applications close May 31

19 shipments. 3 carrier portals.
One workflow. Zero APIs.

Reads an Excel list of container numbers, logs into ANL/MSC/Maersk portals, downloads documents, and dispatches to clients via Outlook. 19 jobs running in parallel. Every email reviewed before it sends.

Chapters Click to jump
0:00 19 shipments. 3 carriers. The Monday problem.
0:06 One trigger — 19 jobs dispatched in parallel
0:20 Two machines, one workflow, running live
0:49 Document retrieval — PDF downloaded, carrier 1
1:03 Carrier 2: different portal, same automation
1:21 Shipments 3–17 at 7× speed
2:27 HITL queue — 19 emails waiting, nothing sent
2:39 Human reads, approves, requests German draft
3:15 Outlook opens — sending the approved email
3:37 PDF attached, matched to shipment ID
4:45 18 emails dispatching at scale
7:25 Output tracker — all 19 done, exceptions flagged
Multi-carrier portal automation — semantic recognition, not pixel coordinates
0:00 The Monday problem
0:06 19 jobs dispatched
0:20 Running live
0:49 PDF downloaded
1:03 Carrier 2
1:21 7× speed
2:27 HITL queue
2:39 German draft
3:15 Outlook opens
3:37 PDF attached
4:45 18 emails scale
7:25 Output tracker
What did your last automation vendor say they couldn't touch?
That's usually where we start.
92 %+
Accuracy in production

ANL · MSC · Maersk — live production. Same portals, same credentials, same clients. Timestamps are real.

* tested production workflows only
72 h
Recording to production

Engineer-reviewed before the first run. That's why it holds.

19
Parallel jobs

From one trigger. Scale doesn't require new recordings.

100 %
Human approval

19 emails in the queue. Every one reviewed before it sent. Your team decides — the automation waits.

0
APIs required

Screen visible? We automate it. Legacy ERP included.

Safe in production

If something unexpected happens, the job stops and flags it — it doesn't guess and keep going. Your team decides what happens next.

How it works

Record once.
The system builds it.
An engineer ships it.

A workflow tested on your exact environment before it runs once.

1 You do

Record your workflow. Once. Exactly as you do it today.

Open the agent trainer and do the workflow on your screen — logging into portals, filling fields, downloading documents, sending emails. The recorder captures every click, every keystroke, every scroll, with a screenshot at each moment. No scripting. No technical setup. When you stop recording, your part is done.

30 minutes of your time. Nothing else required from your team.

Agent trainer · recording session · live
The same shipment workflow from the demo above. 30 minutes. No scripting.
2 System builds

The system reads your recording and generates a workflow template automatically.

Every action you took — captured with exact coordinates, values, and context — is used by the system to generate a structured workflow template. Every step mapped. Every decision point identified. The system reads your screen the way a person does: understanding what each element is, not just where it sits. That's what makes it resilient when portals update their UI.

Most automation tools are black boxes. Before our system builds anything, you can see exactly what it captured — every click, every value, every screen state. If something looks wrong, you catch it here, before it runs.

The system generates a capable template. Our engineer then validates it against your exact environment before it runs — that's the quality gate.

⚡ auto-generated What the system captured — click data, typed values, screenshots
This is the input the system builds from. Every action, every moment.
3 We ship

We don't ship you what the system generated. We ship you what an engineer signed off on.

Our engineer takes the system-generated template, tests it against your exact environment, handles edge cases, adds error recovery, and validates it end to end. This is not a review in name only — it's a run against your actual portals, your actual credentials, your actual data.

What gets deployed is not what the system generated — it's what survived an engineer's hands-on review of it.

This is why the completion rate is 92%+* — not 60%. The system gets you there fast. The engineer gets you there reliably.

✓ engineer reviewed Production workflow template · final
This is what a production-ready workflow template looks like. The system auto-generates the starting version. Our engineer validates and hardens it. Your recording produces the equivalent — scoped to your workflow, your environment.
"30 minutes of your time. 72 hours to production. An engineer signed off on every line."

Your ERP has no API.
Your portal has no connector.
We've automated both.

Human in the loop
Nothing executes
without approval

19 emails queued. Nothing sent. Every one reviewed before dispatch. You saw this in the demo above — that's not a feature flag, it's the architecture.

Scale
One template.
N machines. Live.

Jobs dispatch to whatever machine is available. Proven live: 19 shipments across 3 carrier portals running in parallel.

Enterprise control
LLM guardrails you configure

Auto-generated prompts for every LLM step. You set the model, its flexibility, and exactly which decisions it can make. Our engineers refine these during the build review. The LLM does precisely what you authorised — nothing more.

Data control
Data stays where you decide

Execution runs on your machine by default. Full local deployment — zero data leaves your environment — is available for enterprises that require it. Every data routing decision is deliberate and auditable.

Rules engine
Your business logic

Decision trees, approval thresholds, pricing rules, routing logic — configured during the build, runs inside every job.

Integrations
Your existing APIs

APIs your team already uses are the preferred path — cleaner and more reliable than screen automation where one exists.

Agent bridge
Your AI thinks.
Type-2 acts.

Your agents reason and decide — but stop when there's no API. Type-2 is the execution layer that carries the decision through to the screen.

Universal reach
Any portal, any app

Carrier portals, supplier portals, legacy ERPs, Outlook, Excel. Semantic recognition — not pixel coordinates.

All capabilities ship in production — not demos, not roadmap items.

Workflows we've built or scoped

Two live demos.
More being built.

If your team does this manually, we've probably seen it.

Logistics
Multi-carrier portal document fetch + email dispatch ANL · MSC · Maersk. Reads 19 container numbers from Excel, downloads PDFs across 3 portals, dispatches via Outlook — human approval before every send.
~2.5 hrs/day saved
RevOps
LinkedIn prospect research + personalised outreach Reads a lead list, visits each profile, extracts context, drafts personalised messages — HITL approval before every send.
~3 hrs/day saved
Demo live June 05
Finance
Supplier portal invoice data entry into legacy ERP Logs into supplier portals, extracts invoice fields, enters into ERP with no API. We've scoped 4 workflows like this.
~3 hrs/day saved
Procurement
Purchase order creation across vendor portals Creates POs across multiple vendor portals from a single internal list. No API required — any portal, any format.
~2 hrs/day saved
Coming soon
Any
Yours If it runs on screen and repeats, we can scope it. Send us a Loom or book a call — we'll tell you in 24 hours if we can automate it.
You tell us

Not the right fit if Zapier already solves it.

The boring math

We don't promise ROI percentages.
We show you the arithmetic.

Most automation vendors show you a multiplied percentage. We show you the actual labor hours, the actual hourly rate, and the actual monthly cost — then compare it to what Type-2 costs. You do the math.

Logistics · freight coordinator Daily shipment document fetch — 3 carrier portals
Calculation Monthly
Labor cost $22/hr × 3 hrs/day × 22 days $1,452
Type-2 cost Setup (scoped on call) amortized + usage ~$132
Monthly savings Labor freed for higher-value work ~$1,320
Finance · accounts payable Invoice data entry — supplier portals to ERP
Calculation Monthly
Labor cost $22/hr × 2.5 hrs/day × 22 days $1,210
Type-2 cost Setup (scoped on call) amortized + usage ~$101
Monthly savings AP clerk time back on exception handling ~$1,109
RevOps · sales operations CRM enrichment — LinkedIn + custom sources
Calculation Monthly
Labor cost $32/hr × 2 hrs/day × 22 days $1,408
Type-2 cost Setup (scoped on call) amortized + usage ~$179
Monthly savings Sales ops on pipeline — not data entry ~$1,229
Your numbers
Calculate your own labor cost

Adjust the inputs. The math updates live. Type-2 estimate is based on typical single-workflow pricing.

$ / hr
Include benefits — typically 1.25× base salary
hrs / day
Estimate across the team for this specific workflow
people
Everyone who touches this workflow regularly
Monthly labor cost $1,100
Type-2 estimate ~$120
Monthly savings ~$980
Payback period 3–5 days

Type-2 estimate is illustrative — setup and monthly costs are scoped individually on the workflow call.

These are labor-cost estimates, not guaranteed savings. Actual numbers depend on your workflow, team size, and what people do with freed time. We're not in the business of inflating ROI claims — that's why we show you the arithmetic instead.

Scoped to your workflow.
Fixed before you pay.

Every deployment is priced individually — because every workflow is different. You'll know the exact number before you commit to anything.

What every deployment includes

Four commitments. Every workflow. No negotiation.

01
Built on your recording — not a template

Every click path, every portal, built for your exact environment. If it doesn't work on your machine, we rebuild it.

Tested on your exact portals before delivery
0
generic templates shipped
02
An engineer runs it before you do

Not a staging environment, not a demo account — your actual portals. We don't ship until it passes.

Personal review on every deployment
72h
recording to running
03
Running in production in 72 hours

Others quote 6–12 weeks. We quote 72 hours. No config on your end. No debugging. You click Run.

Median delivery across all tested workflows
6–12wk
industry avg — we're not that
04
90 days. Anything breaks, we fix it.

Portals update. ERPs push patches. First 90 days, if it ran before and doesn't now — we fix it. No ticket, no charge.

Included in every deployment, no exceptions
90
days covered, no charge

What determines your price

Three variables. We tell you which bucket your workflow falls into before you get on a call.

Variable What it means for your workflow Effect
Workflow complexity
steps, decisions, exceptions
A 15-step linear document fetch is a different build from an 80-step branching procurement workflow with exception handling. Both are automatable. The engineering effort is different. ↑ More steps↓ Linear flows
Target system stability
how brittle your portals are
A modern SaaS portal is stable. A 15-year-old ERP or a carrier portal that changes layout quarterly requires more engineering to maintain 92%+ accuracy. We'll tell you which category yours is in. ↑ Legacy / brittle↓ Modern SaaS
Integrations you connect
APIs, agents, tools
APIs your team already uses are the preferred path — cleaner and faster to build. Custom agent handoffs and tool connections add scoping time. If you have them, bring them — they often reduce the screen-automation scope. ↑ Custom handoffs↓ Standard APIs

You don't need to know which category you're in before the call. That's what the pre-call assessment is for.

How you get an exact number

Four steps. You're the subject of every one.

01
You do
Submit a 2-minute application

Describe your workflow in plain English — what it does, how often, which systems it touches. No technical knowledge required.

02
You receive
A written assessment within 24 hours

Complexity tier, systems assessment, and whether we think it's a fit. You get this before any call — no call required to get a verdict.

03
You get
A fixed quote on a 20-minute call

Not a range — a number. Setup cost, monthly estimate, maintenance terms — all confirmed before the call ends.

04
You decide
Go or no-go — on your terms

You have everything you need to decide. No follow-up if you say no. No charge until you say yes.

On monthly costs
Usage-based. Scoped individually. No markup on AI.

Monthly usage scales with how many jobs you run and which tools you connect. LLM and external API costs pass through at cost — no markup. You choose which models you use; we tell you what they add before you commit. Your monthly estimate is confirmed on the scoping call alongside the setup fee.

After the included 90-day maintenance window, ongoing maintenance is billed at a flat hourly rate you'll know upfront. Most UI updates take under an hour. Major scope changes — new steps, new systems — are quoted separately, and we'll always tell you the difference before doing any work.

June 2026 batch

7 workflows. Production-deployed by end of June. Recording to production in 72 hours.

Good fit if
  • You're on Windows — required for now
  • Your team does the same workflow repeatedly — portals, Excel, desktop apps, or legacy ERPs
  • Someone can spend 30 minutes recording the workflow
  • You want it running in production — not a pilot, not a prototype
  • You have existing agents, APIs, or tools you want connected
Not a fit if
  • Zapier or Make already solves your problem
  • You need a free pilot before committing
  • You're on Mac only — Windows is required
  • Your workflow runs sub-second or needs real-time execution
  • Enterprise procurement requires 3+ months of legal review
June Batch Closes May 31
7 spots
applications open now
Scoped to your workflow · Fixed quote before you pay
Early-batch pricing, locked for 12 months
Applications close May 31. Written assessment sent within 24 hours of applying.
Apply for June Batch →

Pre-call assessment included — no call required to get a complexity verdict.

June 2026 Batch · 7 spots · Closes May 31

Apply for a June workflow deployment

Submit a 2-minute application. You'll receive a pre-call assessment within 24 hours — whether it's a fit or not.

Try: "My team does [action] every [frequency] using [systems]."

We review every application within 24 hours and send a written assessment — whether it's a fit or not. No call required to get that far.

Questions

The questions everyone actually asks

Real. The 19 shipments are real container numbers. The ANL, MSC, and Maersk portals are live production portals. The emails dispatched to real clients. Nothing was staged, scripted, or run on a test environment.

The sections labelled 7× are compressed — we told you that. Everything else ran at normal human speed, unedited. What you watched is what your team would see running on a Tuesday morning.

Most automation tools work by connecting to software through APIs — structured data pipelines that developers build. When there's no API (and most carrier portals, legacy ERPs, and internal tools don't have one), those tools stop. Your team goes back to doing it manually.

Type-2 works the way a person works — by navigating the screen. No API required. If someone on your team can log in and do the workflow, we can automate it. That's the gap we fill — not a better version of Zapier, but a different category entirely.

The second reason automations break is that nobody owns them after launch. We include 90 days of maintenance — when the portal updates its layout and something changes, we fix it. That's not an upsell. It's in every deployment.

Your part is about 30 minutes. One person on your team does the workflow on their screen while our recorder captures it — exactly as they do it normally. No scripting, no technical knowledge, no preparation required.

After that, we take over. Our engineers build the automation, test it on your exact environment, and deploy it. The next time your team touches it is to click Run — and review the jobs that need a human decision before they execute.

The automation doesn't crash — it stops safely and flags the problem. If a document is missing, a portal behaves unexpectedly, or any step falls outside what the automation was built to handle, that job pauses and goes into a review queue. The rest of the jobs keep running.

Your team sees exactly what happened and handles that one manually. Nothing gets silently wrong — every exception is visible, labelled, and waiting for a human decision. You saw this in the demo: the output tracker at the end shows completed jobs and flagged ones separately.

Your data stays on your machine. The automation runs locally on your Windows desktop — your credentials, your documents, your portal sessions never leave your environment and don't pass through our servers.

During the build, our engineers see the structure of your workflow — which steps happen in which order — not the data flowing through it. We don't store outputs. We don't have access to your portals after deployment. For teams with stricter requirements, full local deployment is available where there is zero external connection of any kind.

You are, at every step that matters. Any action that has a consequence — an email sent, a form submitted, a document filed — goes into a human review queue before it executes. Your team sees what the automation is about to do and approves it. Nothing sensitive happens without a human seeing it first.

The automation handles the repetitive navigation, data entry, and retrieval. Your team handles every decision. That boundary is not configurable by accident — it's the architecture. The demo shows this clearly: 19 emails queued, nothing sent, waiting for a human to review each one.

From your recording session to a working automation in production: 72 hours. Not a pilot, not a prototype — a working automation, tested on your exact environment by an engineer, ready to run.

Pricing has two parts: a one-time setup fee and a monthly usage cost. Both are scoped to your workflow — the complexity, the systems involved, and what you want connected. You get a fixed quote on a 20-minute call before you pay anything. We assess your workflow, tell you the exact number, and you decide. If it's not a fit, we tell you on the call — no charge.

The honest answer: Type-2 is built for workflows that are repetitive, screen-based, and happen on software that has no API or integration support. If your team does the same multi-step process daily — navigating portals, copying data between systems, sending templated emails — and no existing tool automates it, that's our territory.

It's not the right fit if Zapier or Make already solves your problem — those tools cost a fraction of what we do and you should use them. It's also not right if you're Mac-only (we're Windows for now), or if your workflow requires sub-second real-time execution. If you're not sure, describe your workflow in two sentences and send it to founder@type2.io — we'll tell you within 24 hours whether it's a fit, no call required.

Most automation tools locate elements by where they are on screen — pixel coordinates, CSS selectors, or fixed positions. When a portal updates and a button moves two inches to the left, the automation breaks because it's looking at a location, not an element.

Our system understands what each element is, not just where it sits — the same way a person reads a page. When ANL updated its portal layout last year, the automation kept running because it recognised "the download button for this shipment" rather than "the button at position 847, 392." This is what's included in the 90-day maintenance window — and why most updates take under an hour to address when they do break.

Bring us your workflow.

If your team does a repetitive workflow that no integration tool can automate — record it once and send it to us. We will tell you in 24 hours whether we can automate it, what it takes, and what it costs.

Or email founder@type2.io with a Loom